diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,140521 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 23416,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4224751066856332e-08,
+      "loss": 1.375,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8449502133712664e-08,
+      "loss": 1.3496,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2674253200569e-08,
+      "loss": 1.3516,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.689900426742533e-08,
+      "loss": 1.3711,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.112375533428165e-08,
+      "loss": 1.3232,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.5348506401138e-08,
+      "loss": 1.3662,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.957325746799431e-08,
+      "loss": 1.2842,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1379800853485066e-07,
+      "loss": 1.292,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.28022759601707e-07,
+      "loss": 1.2568,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.422475106685633e-07,
+      "loss": 1.3066,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5647226173541965e-07,
+      "loss": 1.3584,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.70697012802276e-07,
+      "loss": 1.3447,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.849217638691323e-07,
+      "loss": 1.3438,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9914651493598863e-07,
+      "loss": 1.2861,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1337126600284494e-07,
+      "loss": 1.3262,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.275960170697013e-07,
+      "loss": 0.3574,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4182076813655766e-07,
+      "loss": 1.3184,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.56045519203414e-07,
+      "loss": 1.292,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.702702702702703e-07,
+      "loss": 1.3545,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.844950213371266e-07,
+      "loss": 1.2627,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.987197724039829e-07,
+      "loss": 1.3389,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.129445234708393e-07,
+      "loss": 1.3242,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.271692745376956e-07,
+      "loss": 1.2803,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.41394025604552e-07,
+      "loss": 1.2646,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.556187766714083e-07,
+      "loss": 0.3568,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.698435277382646e-07,
+      "loss": 1.2861,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8406827880512093e-07,
+      "loss": 1.3447,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9829302987197725e-07,
+      "loss": 1.2773,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.1251778093883357e-07,
+      "loss": 1.3311,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.267425320056899e-07,
+      "loss": 1.3115,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.409672830725463e-07,
+      "loss": 1.2812,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.551920341394026e-07,
+      "loss": 1.3027,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6941678520625894e-07,
+      "loss": 1.3438,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.836415362731153e-07,
+      "loss": 1.2686,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.978662873399716e-07,
+      "loss": 1.3447,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.12091038406828e-07,
+      "loss": 1.3711,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.263157894736843e-07,
+      "loss": 1.2197,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.405405405405406e-07,
+      "loss": 1.2842,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.547652916073969e-07,
+      "loss": 1.2227,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.689900426742532e-07,
+      "loss": 1.3096,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.832147937411095e-07,
+      "loss": 1.2246,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.974395448079659e-07,
+      "loss": 1.2144,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.116642958748222e-07,
+      "loss": 1.291,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.258890469416786e-07,
+      "loss": 1.2139,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.401137980085349e-07,
+      "loss": 1.1865,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.543385490753912e-07,
+      "loss": 1.2637,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.685633001422475e-07,
+      "loss": 1.2344,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.82788051209104e-07,
+      "loss": 1.2324,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.970128022759602e-07,
+      "loss": 1.2666,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.112375533428166e-07,
+      "loss": 0.3076,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.254623044096728e-07,
+      "loss": 1.2188,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.396870554765292e-07,
+      "loss": 1.2666,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.539118065433856e-07,
+      "loss": 1.1592,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.681365576102419e-07,
+      "loss": 1.1416,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.823613086770983e-07,
+      "loss": 1.2607,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.965860597439545e-07,
+      "loss": 1.2153,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.108108108108109e-07,
+      "loss": 1.1807,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.250355618776671e-07,
+      "loss": 0.2351,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.392603129445236e-07,
+      "loss": 1.2051,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.534850640113798e-07,
+      "loss": 1.292,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.677098150782362e-07,
+      "loss": 1.207,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.819345661450926e-07,
+      "loss": 1.1846,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.961593172119488e-07,
+      "loss": 0.438,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.103840682788053e-07,
+      "loss": 1.2383,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.246088193456615e-07,
+      "loss": 1.1958,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.388335704125179e-07,
+      "loss": 1.2246,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.530583214793741e-07,
+      "loss": 1.124,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.672830725462306e-07,
+      "loss": 1.249,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.815078236130868e-07,
+      "loss": 1.2178,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.957325746799433e-07,
+      "loss": 1.1924,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0099573257467995e-06,
+      "loss": 1.2061,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.024182076813656e-06,
+      "loss": 1.1973,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0384068278805121e-06,
+      "loss": 1.167,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0526315789473685e-06,
+      "loss": 1.1523,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0668563300142247e-06,
+      "loss": 1.2129,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0810810810810812e-06,
+      "loss": 1.1523,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0953058321479376e-06,
+      "loss": 1.1406,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1095305832147938e-06,
+      "loss": 1.1865,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1237553342816502e-06,
+      "loss": 1.2285,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1379800853485064e-06,
+      "loss": 1.145,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1522048364153629e-06,
+      "loss": 1.1562,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.166429587482219e-06,
+      "loss": 1.1455,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1806543385490755e-06,
+      "loss": 1.1953,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1948790896159317e-06,
+      "loss": 1.144,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2091038406827881e-06,
+      "loss": 1.1221,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2233285917496443e-06,
+      "loss": 1.1719,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2375533428165008e-06,
+      "loss": 1.1875,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2517780938833572e-06,
+      "loss": 1.2373,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2660028449502134e-06,
+      "loss": 1.1035,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2802275960170698e-06,
+      "loss": 1.2031,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2944523470839262e-06,
+      "loss": 1.1641,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3086770981507825e-06,
+      "loss": 1.1729,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3229018492176387e-06,
+      "loss": 1.1619,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.337126600284495e-06,
+      "loss": 1.1929,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3513513513513515e-06,
+      "loss": 1.1416,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.365576102418208e-06,
+      "loss": 1.1738,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3798008534850641e-06,
+      "loss": 1.208,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3940256045519204e-06,
+      "loss": 1.2188,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4082503556187768e-06,
+      "loss": 1.1982,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4224751066856332e-06,
+      "loss": 1.2334,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4366998577524894e-06,
+      "loss": 1.2266,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4509246088193456e-06,
+      "loss": 1.1016,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.465149359886202e-06,
+      "loss": 1.2061,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4793741109530585e-06,
+      "loss": 0.3242,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4935988620199149e-06,
+      "loss": 1.1567,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.507823613086771e-06,
+      "loss": 1.103,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5220483641536273e-06,
+      "loss": 1.105,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5362731152204837e-06,
+      "loss": 1.1426,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5504978662873402e-06,
+      "loss": 1.1484,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5647226173541966e-06,
+      "loss": 1.1719,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5789473684210526e-06,
+      "loss": 1.1621,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.593172119487909e-06,
+      "loss": 1.2197,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6073968705547654e-06,
+      "loss": 1.1543,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6216216216216219e-06,
+      "loss": 1.1558,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.635846372688478e-06,
+      "loss": 1.1748,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6500711237553343e-06,
+      "loss": 1.1187,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6642958748221907e-06,
+      "loss": 1.1729,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6785206258890471e-06,
+      "loss": 1.165,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6927453769559035e-06,
+      "loss": 1.1299,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7069701280227595e-06,
+      "loss": 1.1572,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.721194879089616e-06,
+      "loss": 1.1689,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7354196301564724e-06,
+      "loss": 1.1558,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7496443812233288e-06,
+      "loss": 1.1177,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7638691322901852e-06,
+      "loss": 1.1445,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7780938833570412e-06,
+      "loss": 1.1436,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7923186344238977e-06,
+      "loss": 1.1484,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.806543385490754e-06,
+      "loss": 1.1309,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8207681365576105e-06,
+      "loss": 1.1816,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8349928876244665e-06,
+      "loss": 1.1128,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.849217638691323e-06,
+      "loss": 1.1934,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8634423897581793e-06,
+      "loss": 1.1265,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8776671408250358e-06,
+      "loss": 1.1479,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8918918918918922e-06,
+      "loss": 1.1074,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9061166429587482e-06,
+      "loss": 1.1089,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9203413940256046e-06,
+      "loss": 1.1602,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9345661450924613e-06,
+      "loss": 1.125,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9487908961593175e-06,
+      "loss": 1.189,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9630156472261737e-06,
+      "loss": 1.082,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.97724039829303e-06,
+      "loss": 1.1079,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9914651493598865e-06,
+      "loss": 1.1885,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0056899004267427e-06,
+      "loss": 1.1177,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.019914651493599e-06,
+      "loss": 1.1318,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.034139402560455e-06,
+      "loss": 1.0786,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.048364153627312e-06,
+      "loss": 1.105,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.062588904694168e-06,
+      "loss": 1.1172,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0768136557610242e-06,
+      "loss": 1.0664,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.091038406827881e-06,
+      "loss": 1.1426,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.105263157894737e-06,
+      "loss": 1.1724,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1194879089615933e-06,
+      "loss": 1.103,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1337126600284495e-06,
+      "loss": 1.0571,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.147937411095306e-06,
+      "loss": 1.1089,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1621621621621623e-06,
+      "loss": 1.0908,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1763869132290185e-06,
+      "loss": 1.1748,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.190611664295875e-06,
+      "loss": 1.084,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2048364153627314e-06,
+      "loss": 1.1226,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2190611664295876e-06,
+      "loss": 1.1133,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2332859174964442e-06,
+      "loss": 1.1206,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2475106685633004e-06,
+      "loss": 1.1099,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2617354196301567e-06,
+      "loss": 1.1592,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.275960170697013e-06,
+      "loss": 1.1187,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2901849217638695e-06,
+      "loss": 1.0464,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3044096728307257e-06,
+      "loss": 1.0869,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.318634423897582e-06,
+      "loss": 1.1055,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.332859174964438e-06,
+      "loss": 1.0762,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3470839260312948e-06,
+      "loss": 1.0601,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.361308677098151e-06,
+      "loss": 1.1104,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.375533428165007e-06,
+      "loss": 1.1816,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3897581792318634e-06,
+      "loss": 1.0986,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.40398293029872e-06,
+      "loss": 1.1309,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4182076813655762e-06,
+      "loss": 1.104,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.432432432432433e-06,
+      "loss": 1.1309,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4466571834992887e-06,
+      "loss": 1.1611,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4608819345661453e-06,
+      "loss": 1.126,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4751066856330015e-06,
+      "loss": 1.0693,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.489331436699858e-06,
+      "loss": 1.0444,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5035561877667144e-06,
+      "loss": 1.1816,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5177809388335706e-06,
+      "loss": 1.105,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5320056899004268e-06,
+      "loss": 1.0127,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5462304409672834e-06,
+      "loss": 1.0889,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5604551920341396e-06,
+      "loss": 1.126,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5746799431009963e-06,
+      "loss": 1.0879,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5889046941678525e-06,
+      "loss": 1.0698,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6031294452347083e-06,
+      "loss": 1.1689,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.617354196301565e-06,
+      "loss": 1.1572,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.631578947368421e-06,
+      "loss": 1.0903,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6458036984352773e-06,
+      "loss": 1.061,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.660028449502134e-06,
+      "loss": 1.0947,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.67425320056899e-06,
+      "loss": 1.1074,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.688477951635847e-06,
+      "loss": 1.0256,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.702702702702703e-06,
+      "loss": 1.0952,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7169274537695597e-06,
+      "loss": 1.1396,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.731152204836416e-06,
+      "loss": 1.1514,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7453769559032716e-06,
+      "loss": 1.0967,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7596017069701283e-06,
+      "loss": 1.0786,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7738264580369845e-06,
+      "loss": 1.0908,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7880512091038407e-06,
+      "loss": 1.1182,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8022759601706973e-06,
+      "loss": 1.1641,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8165007112375536e-06,
+      "loss": 1.0718,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.83072546230441e-06,
+      "loss": 1.1211,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8449502133712664e-06,
+      "loss": 1.0796,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8591749644381226e-06,
+      "loss": 1.0581,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.873399715504979e-06,
+      "loss": 0.9883,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.887624466571835e-06,
+      "loss": 0.9814,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9018492176386912e-06,
+      "loss": 1.1211,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.916073968705548e-06,
+      "loss": 1.0684,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.930298719772404e-06,
+      "loss": 1.0767,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9445234708392607e-06,
+      "loss": 0.4392,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.958748221906117e-06,
+      "loss": 1.0864,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9729729729729736e-06,
+      "loss": 1.0835,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9871977240398298e-06,
+      "loss": 1.0938,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0014224751066856e-06,
+      "loss": 1.0449,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.015647226173542e-06,
+      "loss": 1.1108,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0298719772403984e-06,
+      "loss": 1.0615,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0440967283072546e-06,
+      "loss": 1.0806,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0583214793741113e-06,
+      "loss": 1.0596,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0725462304409675e-06,
+      "loss": 1.1113,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.086770981507824e-06,
+      "loss": 1.0386,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1009957325746803e-06,
+      "loss": 1.0928,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1152204836415365e-06,
+      "loss": 1.0649,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.129445234708393e-06,
+      "loss": 1.0537,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.143669985775249e-06,
+      "loss": 1.0884,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.157894736842105e-06,
+      "loss": 1.0688,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.172119487908962e-06,
+      "loss": 1.0732,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.186344238975818e-06,
+      "loss": 1.103,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2005689900426746e-06,
+      "loss": 1.0269,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.214793741109531e-06,
+      "loss": 1.1475,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2290184921763875e-06,
+      "loss": 1.1328,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2432432432432437e-06,
+      "loss": 1.0464,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2574679943101e-06,
+      "loss": 1.0654,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.271692745376956e-06,
+      "loss": 1.1387,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2859174964438123e-06,
+      "loss": 1.0845,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3001422475106685e-06,
+      "loss": 1.0947,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.314366998577525e-06,
+      "loss": 1.0732,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3285917496443814e-06,
+      "loss": 1.043,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.342816500711238e-06,
+      "loss": 1.083,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3570412517780942e-06,
+      "loss": 1.0342,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3712660028449505e-06,
+      "loss": 1.1553,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.385490753911807e-06,
+      "loss": 0.2842,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3997155049786633e-06,
+      "loss": 1.0776,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.413940256045519e-06,
+      "loss": 1.0503,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4281650071123757e-06,
+      "loss": 1.0981,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.442389758179232e-06,
+      "loss": 1.0908,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4566145092460886e-06,
+      "loss": 1.0796,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4708392603129448e-06,
+      "loss": 1.1152,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4850640113798014e-06,
+      "loss": 1.019,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4992887624466576e-06,
+      "loss": 0.2668,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.513513513513514e-06,
+      "loss": 1.0957,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5277382645803705e-06,
+      "loss": 1.1206,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5419630156472263e-06,
+      "loss": 1.0298,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5561877667140825e-06,
+      "loss": 1.1147,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.570412517780939e-06,
+      "loss": 1.0669,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5846372688477953e-06,
+      "loss": 1.0059,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.598862019914652e-06,
+      "loss": 1.1216,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.613086770981508e-06,
+      "loss": 1.0781,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6273115220483644e-06,
+      "loss": 1.0586,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.641536273115221e-06,
+      "loss": 1.0869,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6557610241820772e-06,
+      "loss": 1.0791,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.669985775248933e-06,
+      "loss": 1.0107,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6842105263157896e-06,
+      "loss": 1.1548,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.698435277382646e-06,
+      "loss": 1.0498,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7126600284495025e-06,
+      "loss": 1.0962,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7268847795163587e-06,
+      "loss": 1.0742,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7411095305832153e-06,
+      "loss": 1.1089,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7553342816500715e-06,
+      "loss": 1.0908,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7695590327169278e-06,
+      "loss": 1.1011,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7837837837837844e-06,
+      "loss": 1.0532,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7980085348506406e-06,
+      "loss": 1.0586,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8122332859174964e-06,
+      "loss": 1.0059,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.826458036984353e-06,
+      "loss": 1.0283,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.840682788051209e-06,
+      "loss": 1.0854,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.854907539118066e-06,
+      "loss": 1.0449,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8691322901849225e-06,
+      "loss": 1.0835,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.883357041251778e-06,
+      "loss": 1.0103,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.897581792318635e-06,
+      "loss": 1.0415,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9118065433854916e-06,
+      "loss": 1.0664,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.926031294452347e-06,
+      "loss": 1.0508,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.940256045519203e-06,
+      "loss": 1.0322,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.95448079658606e-06,
+      "loss": 1.0293,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.968705547652916e-06,
+      "loss": 1.0259,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.982930298719773e-06,
+      "loss": 1.0107,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.997155049786629e-06,
+      "loss": 1.0762,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0113798008534855e-06,
+      "loss": 1.123,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.025604551920342e-06,
+      "loss": 1.0293,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.039829302987198e-06,
+      "loss": 1.0659,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0540540540540545e-06,
+      "loss": 1.0806,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.06827880512091e-06,
+      "loss": 1.0723,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.082503556187767e-06,
+      "loss": 1.084,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.096728307254624e-06,
+      "loss": 1.0513,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.110953058321479e-06,
+      "loss": 1.0889,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.125177809388336e-06,
+      "loss": 1.0454,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.139402560455193e-06,
+      "loss": 1.0215,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1536273115220484e-06,
+      "loss": 1.0508,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.167852062588905e-06,
+      "loss": 1.0737,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.182076813655762e-06,
+      "loss": 1.0493,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.196301564722618e-06,
+      "loss": 1.0098,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.210526315789474e-06,
+      "loss": 1.1035,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.22475106685633e-06,
+      "loss": 1.1084,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2389758179231865e-06,
+      "loss": 1.0356,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.253200568990043e-06,
+      "loss": 0.2501,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.267425320056899e-06,
+      "loss": 1.0342,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.281650071123756e-06,
+      "loss": 1.0898,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.295874822190612e-06,
+      "loss": 1.1221,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.310099573257469e-06,
+      "loss": 1.0005,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.324324324324325e-06,
+      "loss": 1.001,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3385490753911804e-06,
+      "loss": 0.9802,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.352773826458037e-06,
+      "loss": 1.02,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.366998577524894e-06,
+      "loss": 0.4558,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.38122332859175e-06,
+      "loss": 1.0444,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.395448079658606e-06,
+      "loss": 1.0518,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.409672830725463e-06,
+      "loss": 1.0479,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.423897581792319e-06,
+      "loss": 1.0479,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.438122332859175e-06,
+      "loss": 1.0513,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.452347083926032e-06,
+      "loss": 1.0532,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4665718349928885e-06,
+      "loss": 1.0977,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.480796586059744e-06,
+      "loss": 1.0986,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.495021337126601e-06,
+      "loss": 1.0713,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.509246088193457e-06,
+      "loss": 0.3102,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.523470839260313e-06,
+      "loss": 1.0835,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.53769559032717e-06,
+      "loss": 1.0962,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.551920341394026e-06,
+      "loss": 1.1191,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.566145092460882e-06,
+      "loss": 1.0996,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.580369843527739e-06,
+      "loss": 1.0317,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.594594594594596e-06,
+      "loss": 1.0825,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.608819345661451e-06,
+      "loss": 1.1152,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.623044096728307e-06,
+      "loss": 1.0317,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.637268847795164e-06,
+      "loss": 1.0068,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6514935988620205e-06,
+      "loss": 0.3975,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.665718349928876e-06,
+      "loss": 1.1006,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.679943100995733e-06,
+      "loss": 1.0825,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6941678520625895e-06,
+      "loss": 1.0283,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.708392603129446e-06,
+      "loss": 1.0693,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.722617354196302e-06,
+      "loss": 1.082,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.736842105263158e-06,
+      "loss": 1.0654,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.751066856330014e-06,
+      "loss": 1.1318,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.765291607396871e-06,
+      "loss": 1.0151,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.779516358463727e-06,
+      "loss": 1.0889,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7937411095305834e-06,
+      "loss": 1.1221,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.80796586059744e-06,
+      "loss": 1.0229,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.822190611664297e-06,
+      "loss": 1.02,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8364153627311525e-06,
+      "loss": 0.9805,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.850640113798009e-06,
+      "loss": 1.0332,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.864864864864866e-06,
+      "loss": 1.0459,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8790896159317216e-06,
+      "loss": 0.9834,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.893314366998577e-06,
+      "loss": 1.1074,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.907539118065434e-06,
+      "loss": 1.082,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.921763869132291e-06,
+      "loss": 1.0254,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.935988620199147e-06,
+      "loss": 1.0337,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.950213371266003e-06,
+      "loss": 1.0728,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.96443812233286e-06,
+      "loss": 1.1123,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.978662873399716e-06,
+      "loss": 1.0298,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.992887624466572e-06,
+      "loss": 1.0928,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.007112375533429e-06,
+      "loss": 1.0718,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.021337126600285e-06,
+      "loss": 1.021,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.035561877667141e-06,
+      "loss": 0.9873,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.049786628733998e-06,
+      "loss": 1.0215,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0640113798008536e-06,
+      "loss": 1.022,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.078236130867709e-06,
+      "loss": 1.0488,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.092460881934567e-06,
+      "loss": 1.0479,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.106685633001423e-06,
+      "loss": 1.0356,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.120910384068279e-06,
+      "loss": 1.0576,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.135135135135135e-06,
+      "loss": 1.0928,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.1493598862019925e-06,
+      "loss": 1.1523,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.163584637268848e-06,
+      "loss": 1.0874,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.177809388335705e-06,
+      "loss": 1.0835,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.192034139402561e-06,
+      "loss": 1.0874,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2062588904694165e-06,
+      "loss": 1.0508,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.220483641536274e-06,
+      "loss": 1.0122,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.23470839260313e-06,
+      "loss": 1.0449,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2489331436699864e-06,
+      "loss": 1.0615,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.263157894736842e-06,
+      "loss": 0.957,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.277382645803699e-06,
+      "loss": 1.019,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.291607396870555e-06,
+      "loss": 1.0264,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.305832147937412e-06,
+      "loss": 1.0879,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.320056899004268e-06,
+      "loss": 1.0664,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.334281650071124e-06,
+      "loss": 1.0752,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.34850640113798e-06,
+      "loss": 1.0718,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.362731152204836e-06,
+      "loss": 1.021,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.376955903271694e-06,
+      "loss": 1.0156,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.391180654338549e-06,
+      "loss": 1.0396,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.405405405405406e-06,
+      "loss": 1.0278,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.419630156472262e-06,
+      "loss": 0.9985,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.433854907539119e-06,
+      "loss": 1.0449,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.448079658605975e-06,
+      "loss": 1.0469,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.462304409672832e-06,
+      "loss": 1.1035,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4765291607396875e-06,
+      "loss": 0.9976,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.490753911806543e-06,
+      "loss": 0.9873,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5049786628734e-06,
+      "loss": 0.9961,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5192034139402566e-06,
+      "loss": 1.0928,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.533428165007113e-06,
+      "loss": 1.0386,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.547652916073969e-06,
+      "loss": 1.0522,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.561877667140826e-06,
+      "loss": 1.1221,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.576102418207681e-06,
+      "loss": 1.126,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.590327169274539e-06,
+      "loss": 1.0645,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.604551920341395e-06,
+      "loss": 1.0684,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6187766714082505e-06,
+      "loss": 1.0913,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.633001422475107e-06,
+      "loss": 1.0811,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.647226173541963e-06,
+      "loss": 1.0464,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.66145092460882e-06,
+      "loss": 1.0366,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.675675675675676e-06,
+      "loss": 1.0513,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.689900426742533e-06,
+      "loss": 1.0381,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.704125177809389e-06,
+      "loss": 0.9702,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.718349928876245e-06,
+      "loss": 1.001,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.732574679943102e-06,
+      "loss": 1.1108,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.746799431009958e-06,
+      "loss": 0.9839,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.761024182076814e-06,
+      "loss": 0.9307,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.77524893314367e-06,
+      "loss": 1.0439,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.789473684210527e-06,
+      "loss": 1.0317,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8036984352773825e-06,
+      "loss": 1.0601,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.81792318634424e-06,
+      "loss": 1.0186,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.832147937411096e-06,
+      "loss": 1.0542,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.846372688477952e-06,
+      "loss": 1.0239,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.860597439544808e-06,
+      "loss": 1.0908,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.874822190611664e-06,
+      "loss": 0.9932,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8890469416785214e-06,
+      "loss": 1.0352,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.903271692745377e-06,
+      "loss": 1.0371,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.917496443812234e-06,
+      "loss": 0.3362,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.93172119487909e-06,
+      "loss": 1.1157,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.945945945945947e-06,
+      "loss": 1.0508,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.960170697012803e-06,
+      "loss": 1.0708,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9743954480796596e-06,
+      "loss": 0.3328,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.988620199146515e-06,
+      "loss": 1.0269,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.002844950213371e-06,
+      "loss": 1.063,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.017069701280228e-06,
+      "loss": 1.0332,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.031294452347084e-06,
+      "loss": 0.9463,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.045519203413941e-06,
+      "loss": 1.041,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.059743954480797e-06,
+      "loss": 1.0615,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0739687055476535e-06,
+      "loss": 1.0176,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.088193456614509e-06,
+      "loss": 1.0693,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.102418207681367e-06,
+      "loss": 1.041,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.1166429587482225e-06,
+      "loss": 1.043,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.130867709815079e-06,
+      "loss": 1.0073,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.145092460881935e-06,
+      "loss": 1.04,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.159317211948791e-06,
+      "loss": 1.0425,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.173541963015648e-06,
+      "loss": 1.0034,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.187766714082504e-06,
+      "loss": 1.0703,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.201991465149361e-06,
+      "loss": 0.3364,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.2162162162162164e-06,
+      "loss": 1.0649,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.230440967283073e-06,
+      "loss": 1.0454,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.24466571834993e-06,
+      "loss": 0.2655,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.258890469416786e-06,
+      "loss": 1.0269,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.273115220483642e-06,
+      "loss": 1.0117,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.287339971550498e-06,
+      "loss": 1.0435,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3015647226173545e-06,
+      "loss": 1.0923,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.31578947368421e-06,
+      "loss": 0.9839,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.330014224751068e-06,
+      "loss": 1.001,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.344238975817924e-06,
+      "loss": 0.9854,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.35846372688478e-06,
+      "loss": 0.9663,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.372688477951636e-06,
+      "loss": 1.0674,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3869132290184935e-06,
+      "loss": 1.0171,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.401137980085349e-06,
+      "loss": 1.0967,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.415362731152205e-06,
+      "loss": 1.0088,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.429587482219062e-06,
+      "loss": 1.0977,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4438122332859175e-06,
+      "loss": 1.002,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.458036984352775e-06,
+      "loss": 0.999,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.472261735419631e-06,
+      "loss": 1.0659,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.486486486486487e-06,
+      "loss": 1.0381,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.500711237553343e-06,
+      "loss": 1.042,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.5149359886202e-06,
+      "loss": 1.0835,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.529160739687056e-06,
+      "loss": 0.9585,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.543385490753912e-06,
+      "loss": 1.0171,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.557610241820769e-06,
+      "loss": 0.9858,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.571834992887625e-06,
+      "loss": 1.0503,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.586059743954481e-06,
+      "loss": 1.0117,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.600284495021337e-06,
+      "loss": 0.9868,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.614509246088195e-06,
+      "loss": 0.9985,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.62873399715505e-06,
+      "loss": 1.084,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.642958748221907e-06,
+      "loss": 1.0771,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.657183499288763e-06,
+      "loss": 1.0908,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.671408250355619e-06,
+      "loss": 0.9829,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.685633001422476e-06,
+      "loss": 1.0825,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.699857752489332e-06,
+      "loss": 1.0054,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.7140825035561885e-06,
+      "loss": 0.9961,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.728307254623044e-06,
+      "loss": 0.9795,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.742532005689901e-06,
+      "loss": 1.0186,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.7567567567567575e-06,
+      "loss": 1.0439,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.770981507823614e-06,
+      "loss": 1.0576,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.78520625889047e-06,
+      "loss": 1.0361,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.799431009957327e-06,
+      "loss": 1.043,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.813655761024182e-06,
+      "loss": 0.9824,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.827880512091038e-06,
+      "loss": 1.1035,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.842105263157896e-06,
+      "loss": 1.0571,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8563300142247514e-06,
+      "loss": 1.021,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.870554765291608e-06,
+      "loss": 1.0581,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.884779516358464e-06,
+      "loss": 1.04,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.899004267425321e-06,
+      "loss": 1.0044,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.913229018492177e-06,
+      "loss": 1.0771,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.927453769559034e-06,
+      "loss": 1.0098,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.9416785206258896e-06,
+      "loss": 0.9917,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.955903271692745e-06,
+      "loss": 1.0571,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.970128022759603e-06,
+      "loss": 1.0142,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.984352773826459e-06,
+      "loss": 1.0342,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.998577524893315e-06,
+      "loss": 1.0342,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.012802275960171e-06,
+      "loss": 1.0625,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.027027027027028e-06,
+      "loss": 1.0444,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0412517780938835e-06,
+      "loss": 0.9717,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.055476529160741e-06,
+      "loss": 1.0083,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.069701280227597e-06,
+      "loss": 1.0073,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0839260312944525e-06,
+      "loss": 1.0298,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.098150782361309e-06,
+      "loss": 1.0435,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.112375533428165e-06,
+      "loss": 1.0308,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.126600284495022e-06,
+      "loss": 1.0195,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.140825035561878e-06,
+      "loss": 1.0132,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.155049786628735e-06,
+      "loss": 0.9688,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.169274537695591e-06,
+      "loss": 1.0034,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.183499288762448e-06,
+      "loss": 0.9927,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.197724039829304e-06,
+      "loss": 1.1479,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.21194879089616e-06,
+      "loss": 1.0386,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.226173541963016e-06,
+      "loss": 0.9883,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.240398293029872e-06,
+      "loss": 0.9805,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.254623044096729e-06,
+      "loss": 1.0215,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.268847795163585e-06,
+      "loss": 0.998,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.283072546230442e-06,
+      "loss": 0.9946,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.297297297297298e-06,
+      "loss": 0.98,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.3115220483641544e-06,
+      "loss": 1.0356,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.32574679943101e-06,
+      "loss": 0.9556,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.339971550497866e-06,
+      "loss": 1.0137,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.3541963015647235e-06,
+      "loss": 0.9902,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.368421052631579e-06,
+      "loss": 0.9814,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.382645803698436e-06,
+      "loss": 1.0474,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.396870554765292e-06,
+      "loss": 1.0698,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.411095305832149e-06,
+      "loss": 1.0088,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.425320056899005e-06,
+      "loss": 1.0093,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.439544807965862e-06,
+      "loss": 1.0405,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.453769559032717e-06,
+      "loss": 1.0547,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.467994310099573e-06,
+      "loss": 0.9634,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.482219061166431e-06,
+      "loss": 1.0356,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.4964438122332865e-06,
+      "loss": 1.0679,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.510668563300143e-06,
+      "loss": 1.0103,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.524893314366999e-06,
+      "loss": 1.0532,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.5391180654338555e-06,
+      "loss": 0.9868,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.553342816500711e-06,
+      "loss": 1.0737,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.567567567567569e-06,
+      "loss": 1.0688,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.5817923186344246e-06,
+      "loss": 0.9927,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.596017069701281e-06,
+      "loss": 0.9912,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.610241820768137e-06,
+      "loss": 1.0293,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.624466571834993e-06,
+      "loss": 1.0239,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.63869132290185e-06,
+      "loss": 1.0742,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.652916073968705e-06,
+      "loss": 0.9966,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.667140825035564e-06,
+      "loss": 1.0059,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.681365576102418e-06,
+      "loss": 1.0405,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.695590327169275e-06,
+      "loss": 0.9546,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.709815078236132e-06,
+      "loss": 1.043,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.724039829302988e-06,
+      "loss": 1.0664,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.738264580369845e-06,
+      "loss": 0.9893,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.7524893314367e-06,
+      "loss": 1.0566,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.766714082503557e-06,
+      "loss": 1.0396,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.780938833570413e-06,
+      "loss": 1.0049,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.79516358463727e-06,
+      "loss": 1.0576,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.809388335704125e-06,
+      "loss": 1.0835,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.823613086770983e-06,
+      "loss": 1.0176,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.837837837837838e-06,
+      "loss": 1.0376,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.852062588904695e-06,
+      "loss": 1.0298,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.866287339971551e-06,
+      "loss": 1.0332,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.880512091038406e-06,
+      "loss": 0.9307,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.894736842105265e-06,
+      "loss": 1.0625,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.90896159317212e-06,
+      "loss": 1.04,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.923186344238976e-06,
+      "loss": 1.042,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.937411095305833e-06,
+      "loss": 1.0215,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.95163584637269e-06,
+      "loss": 1.0347,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.965860597439546e-06,
+      "loss": 1.0298,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.980085348506403e-06,
+      "loss": 0.9761,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.994310099573258e-06,
+      "loss": 0.9526,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.008534850640114e-06,
+      "loss": 1.0059,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.022759601706971e-06,
+      "loss": 0.9795,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.036984352773826e-06,
+      "loss": 1.061,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.051209103840684e-06,
+      "loss": 0.9727,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.065433854907539e-06,
+      "loss": 1.0186,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.079658605974396e-06,
+      "loss": 1.124,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.093883357041252e-06,
+      "loss": 1.0396,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.108108108108109e-06,
+      "loss": 1.0498,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.122332859174966e-06,
+      "loss": 1.0063,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.13655761024182e-06,
+      "loss": 1.0317,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.150782361308677e-06,
+      "loss": 1.0698,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.165007112375534e-06,
+      "loss": 0.9722,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.17923186344239e-06,
+      "loss": 0.9976,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.193456614509247e-06,
+      "loss": 1.0327,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.207681365576104e-06,
+      "loss": 0.939,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.221906116642959e-06,
+      "loss": 1.0029,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.236130867709815e-06,
+      "loss": 0.998,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.250355618776672e-06,
+      "loss": 1.0249,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.264580369843529e-06,
+      "loss": 0.9805,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.278805120910385e-06,
+      "loss": 1.0337,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.29302987197724e-06,
+      "loss": 0.9893,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.307254623044097e-06,
+      "loss": 1.0063,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.321479374110953e-06,
+      "loss": 1.0405,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.33570412517781e-06,
+      "loss": 1.0244,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.349928876244667e-06,
+      "loss": 1.0703,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.364153627311523e-06,
+      "loss": 1.0342,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.378378378378378e-06,
+      "loss": 0.98,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.392603129445237e-06,
+      "loss": 1.0537,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.406827880512092e-06,
+      "loss": 1.0898,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.421052631578948e-06,
+      "loss": 1.0532,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.435277382645805e-06,
+      "loss": 1.0288,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.44950213371266e-06,
+      "loss": 1.0195,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.463726884779518e-06,
+      "loss": 1.021,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.477951635846373e-06,
+      "loss": 1.0327,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.49217638691323e-06,
+      "loss": 0.9756,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.506401137980086e-06,
+      "loss": 0.9976,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.520625889046943e-06,
+      "loss": 0.9917,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.534850640113798e-06,
+      "loss": 1.0396,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.549075391180655e-06,
+      "loss": 1.0454,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.563300142247511e-06,
+      "loss": 0.959,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.577524893314368e-06,
+      "loss": 1.0,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.591749644381224e-06,
+      "loss": 1.0371,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.60597439544808e-06,
+      "loss": 1.0088,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.620199146514938e-06,
+      "loss": 0.9668,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.634423897581793e-06,
+      "loss": 0.9917,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.64864864864865e-06,
+      "loss": 1.0166,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.662873399715506e-06,
+      "loss": 1.0205,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.677098150782361e-06,
+      "loss": 1.0513,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.69132290184922e-06,
+      "loss": 1.0347,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.705547652916074e-06,
+      "loss": 0.981,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.71977240398293e-06,
+      "loss": 1.0513,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.733997155049787e-06,
+      "loss": 0.9834,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.748221906116644e-06,
+      "loss": 0.9971,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.7624466571835e-06,
+      "loss": 0.9927,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.776671408250357e-06,
+      "loss": 1.1084,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.790896159317212e-06,
+      "loss": 1.043,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.805120910384069e-06,
+      "loss": 1.0308,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.819345661450926e-06,
+      "loss": 0.9976,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.83357041251778e-06,
+      "loss": 0.9634,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.847795163584639e-06,
+      "loss": 1.0327,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.862019914651494e-06,
+      "loss": 0.3149,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.87624466571835e-06,
+      "loss": 0.9902,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.890469416785207e-06,
+      "loss": 1.0732,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.904694167852064e-06,
+      "loss": 1.0283,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.91891891891892e-06,
+      "loss": 1.0259,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.933143669985777e-06,
+      "loss": 1.0278,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 1.0439,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.961593172119488e-06,
+      "loss": 1.0605,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.975817923186345e-06,
+      "loss": 1.0366,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.990042674253202e-06,
+      "loss": 1.0454,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.004267425320058e-06,
+      "loss": 0.9629,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.018492176386913e-06,
+      "loss": 1.0688,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.03271692745377e-06,
+      "loss": 1.0376,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.046941678520627e-06,
+      "loss": 1.0166,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.061166429587483e-06,
+      "loss": 0.9404,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.07539118065434e-06,
+      "loss": 0.2821,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.089615931721195e-06,
+      "loss": 1.0776,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.103840682788051e-06,
+      "loss": 1.0303,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.118065433854908e-06,
+      "loss": 1.0195,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.132290184921765e-06,
+      "loss": 1.0459,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.146514935988621e-06,
+      "loss": 1.0317,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.160739687055478e-06,
+      "loss": 1.0195,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.174964438122333e-06,
+      "loss": 0.3699,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.189189189189191e-06,
+      "loss": 1.0156,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.203413940256046e-06,
+      "loss": 1.0093,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.217638691322903e-06,
+      "loss": 1.0068,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.23186344238976e-06,
+      "loss": 1.0259,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.246088193456614e-06,
+      "loss": 0.9697,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.260312944523471e-06,
+      "loss": 1.0391,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.274537695590328e-06,
+      "loss": 0.998,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.288762446657184e-06,
+      "loss": 0.9717,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.302987197724041e-06,
+      "loss": 1.0234,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.317211948790898e-06,
+      "loss": 0.3805,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.331436699857753e-06,
+      "loss": 1.0264,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.34566145092461e-06,
+      "loss": 1.0015,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.359886201991466e-06,
+      "loss": 1.0205,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.374110953058322e-06,
+      "loss": 1.0083,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.388335704125179e-06,
+      "loss": 1.001,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.402560455192034e-06,
+      "loss": 1.0503,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.416785206258892e-06,
+      "loss": 0.9927,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.431009957325747e-06,
+      "loss": 1.0024,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.445234708392604e-06,
+      "loss": 1.0615,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.45945945945946e-06,
+      "loss": 1.0537,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.473684210526315e-06,
+      "loss": 0.328,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.487908961593174e-06,
+      "loss": 1.0176,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.502133712660029e-06,
+      "loss": 1.0376,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.516358463726885e-06,
+      "loss": 0.9702,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.530583214793742e-06,
+      "loss": 0.9727,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.544807965860599e-06,
+      "loss": 1.0078,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.559032716927454e-06,
+      "loss": 0.9844,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.573257467994312e-06,
+      "loss": 1.0342,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.587482219061167e-06,
+      "loss": 1.0708,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.601706970128024e-06,
+      "loss": 1.0146,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.61593172119488e-06,
+      "loss": 0.9922,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.630156472261735e-06,
+      "loss": 1.0381,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.644381223328593e-06,
+      "loss": 0.9927,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.658605974395448e-06,
+      "loss": 1.0449,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.672830725462305e-06,
+      "loss": 1.0518,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.687055476529162e-06,
+      "loss": 1.0181,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.701280227596018e-06,
+      "loss": 1.0649,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.715504978662875e-06,
+      "loss": 1.0234,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.729729729729732e-06,
+      "loss": 0.9922,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.743954480796586e-06,
+      "loss": 0.9448,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.758179231863443e-06,
+      "loss": 1.0176,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.7724039829303e-06,
+      "loss": 1.0342,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.786628733997155e-06,
+      "loss": 1.0205,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.800853485064013e-06,
+      "loss": 0.9941,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.815078236130868e-06,
+      "loss": 1.0488,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.829302987197725e-06,
+      "loss": 0.9775,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.843527738264581e-06,
+      "loss": 1.0586,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.857752489331438e-06,
+      "loss": 1.0366,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.871977240398294e-06,
+      "loss": 0.9468,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.88620199146515e-06,
+      "loss": 0.265,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900426742532006e-06,
+      "loss": 1.0005,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914651493598863e-06,
+      "loss": 1.0117,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92887624466572e-06,
+      "loss": 1.002,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.943100995732576e-06,
+      "loss": 1.0073,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.957325746799433e-06,
+      "loss": 1.0034,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971550497866288e-06,
+      "loss": 0.9868,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985775248933144e-06,
+      "loss": 1.0103,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 1.0273,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999995217106e-06,
+      "loss": 1.0415,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999808684244e-06,
+      "loss": 1.0332,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999569539551e-06,
+      "loss": 0.9536,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999234736987e-06,
+      "loss": 0.9712,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998804276559e-06,
+      "loss": 1.0005,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998278158273e-06,
+      "loss": 0.9282,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999997656382144e-06,
+      "loss": 0.9937,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999693894818e-06,
+      "loss": 1.0928,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999996125856395e-06,
+      "loss": 0.9565,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999995217106805e-06,
+      "loss": 1.0093,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999994212699427e-06,
+      "loss": 1.04,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999311263428e-06,
+      "loss": 1.0488,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999191691139e-06,
+      "loss": 1.0415,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999062553077e-06,
+      "loss": 0.9824,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999989238492454e-06,
+      "loss": 1.0337,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999987755796464e-06,
+      "loss": 1.0176,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999986177442829e-06,
+      "loss": 0.9839,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99998450343158e-06,
+      "loss": 1.0283,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999982733762748e-06,
+      "loss": 1.0312,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999980868436368e-06,
+      "loss": 0.9951,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999978907452474e-06,
+      "loss": 1.041,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999976850811104e-06,
+      "loss": 1.0522,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999974698512298e-06,
+      "loss": 0.9482,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999972450556097e-06,
+      "loss": 1.0132,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999970106942547e-06,
+      "loss": 1.0127,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999967667671687e-06,
+      "loss": 0.9849,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999965132743566e-06,
+      "loss": 0.2924,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999962502158235e-06,
+      "loss": 1.1133,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999959775915741e-06,
+      "loss": 0.9976,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99995695401614e-06,
+      "loss": 1.0288,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999954036459481e-06,
+      "loss": 1.0273,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999951023245822e-06,
+      "loss": 1.0278,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999947914375222e-06,
+      "loss": 1.0415,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99994470984774e-06,
+      "loss": 1.0225,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999941409663436e-06,
+      "loss": 1.0586,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999938013822373e-06,
+      "loss": 0.9414,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999934522324619e-06,
+      "loss": 0.4357,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999930935170237e-06,
+      "loss": 0.9316,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999927252359299e-06,
+      "loss": 1.0542,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999923473891871e-06,
+      "loss": 0.9551,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99991959976803e-06,
+      "loss": 0.9932,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99991562998785e-06,
+      "loss": 1.0088,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9999115645514e-06,
+      "loss": 0.9585,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999907403458767e-06,
+      "loss": 0.9858,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999903146710025e-06,
+      "loss": 0.9487,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999898794305258e-06,
+      "loss": 1.0435,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999894346244547e-06,
+      "loss": 1.0146,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999889802527978e-06,
+      "loss": 1.0156,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99988516315564e-06,
+      "loss": 1.0435,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99988042812762e-06,
+      "loss": 0.98,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999875597444007e-06,
+      "loss": 1.0234,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999870671104894e-06,
+      "loss": 0.9922,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999865649110379e-06,
+      "loss": 1.0464,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999860531460555e-06,
+      "loss": 0.9937,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999855318155518e-06,
+      "loss": 0.9541,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999850009195372e-06,
+      "loss": 1.0166,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999844604580216e-06,
+      "loss": 1.021,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999839104310155e-06,
+      "loss": 1.0186,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999833508385291e-06,
+      "loss": 0.9971,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999827816805734e-06,
+      "loss": 0.9829,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999822029571593e-06,
+      "loss": 1.0161,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999816146682977e-06,
+      "loss": 1.0215,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999810168139998e-06,
+      "loss": 0.9946,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999804093942774e-06,
+      "loss": 1.0249,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99979792409142e-06,
+      "loss": 0.9858,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99979165858605e-06,
+      "loss": 1.0327,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999785297426788e-06,
+      "loss": 0.9302,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999778840613754e-06,
+      "loss": 0.9795,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999772288147074e-06,
+      "loss": 1.0439,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99976564002687e-06,
+      "loss": 0.9697,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999758896253272e-06,
+      "loss": 1.0454,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999752056826406e-06,
+      "loss": 0.9521,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999745121746405e-06,
+      "loss": 1.0469,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999738091013401e-06,
+      "loss": 0.9946,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99973096462753e-06,
+      "loss": 0.9551,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999723742588924e-06,
+      "loss": 0.9502,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999716424897727e-06,
+      "loss": 0.9531,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999709011554078e-06,
+      "loss": 0.9844,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999701502558113e-06,
+      "loss": 1.0034,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99969389790998e-06,
+      "loss": 1.0029,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999686197609827e-06,
+      "loss": 0.9419,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999678401657797e-06,
+      "loss": 0.999,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99967051005404e-06,
+      "loss": 1.0347,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99966252279871e-06,
+      "loss": 1.0137,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999654439891954e-06,
+      "loss": 0.9243,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999646261333933e-06,
+      "loss": 1.0386,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9996379871248e-06,
+      "loss": 0.9668,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999629617264712e-06,
+      "loss": 1.0474,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999621151753834e-06,
+      "loss": 1.0083,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999612590592323e-06,
+      "loss": 1.0308,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999603933780344e-06,
+      "loss": 0.9482,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999595181318064e-06,
+      "loss": 1.0034,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999586333205651e-06,
+      "loss": 0.9922,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999577389443273e-06,
+      "loss": 1.0376,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9995683500311e-06,
+      "loss": 0.9712,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999559214969306e-06,
+      "loss": 1.0249,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999549984258065e-06,
+      "loss": 0.9712,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999540657897557e-06,
+      "loss": 0.9639,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999531235887957e-06,
+      "loss": 0.3835,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999521718229445e-06,
+      "loss": 0.998,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999512104922206e-06,
+      "loss": 0.9644,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999502395966422e-06,
+      "loss": 0.9912,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999492591362279e-06,
+      "loss": 1.0293,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999482691109962e-06,
+      "loss": 0.9102,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999472695209667e-06,
+      "loss": 1.0283,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999462603661578e-06,
+      "loss": 0.979,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999452416465892e-06,
+      "loss": 1.002,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999442133622803e-06,
+      "loss": 0.9834,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999431755132508e-06,
+      "loss": 1.0117,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999421280995205e-06,
+      "loss": 1.0161,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999410711211096e-06,
+      "loss": 1.0415,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99940004578038e-06,
+      "loss": 1.041,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999389284703265e-06,
+      "loss": 1.0576,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999378427979955e-06,
+      "loss": 1.0293,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999367475610658e-06,
+      "loss": 1.0117,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999356427595582e-06,
+      "loss": 0.9761,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999345283934938e-06,
+      "loss": 1.0098,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999334044628944e-06,
+      "loss": 1.0273,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99932270967781e-06,
+      "loss": 1.0688,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999311279081753e-06,
+      "loss": 1.0117,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999299752840996e-06,
+      "loss": 0.9937,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999288130955756e-06,
+      "loss": 1.0654,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999276413426255e-06,
+      "loss": 0.9819,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999264600252721e-06,
+      "loss": 0.9521,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999252691435375e-06,
+      "loss": 1.0459,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999240686974447e-06,
+      "loss": 1.0078,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999228586870168e-06,
+      "loss": 1.0029,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999216391122769e-06,
+      "loss": 1.0176,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99920409973248e-06,
+      "loss": 0.9956,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99919171269954e-06,
+      "loss": 0.9727,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999179230024185e-06,
+      "loss": 1.0034,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999166651706656e-06,
+      "loss": 1.0303,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999153977747187e-06,
+      "loss": 0.9932,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999141208146029e-06,
+      "loss": 0.9629,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999128342903418e-06,
+      "loss": 0.9648,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999115382019606e-06,
+      "loss": 0.9644,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99910232549484e-06,
+      "loss": 1.0107,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999089173329369e-06,
+      "loss": 0.9712,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999075925523442e-06,
+      "loss": 0.9824,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999062582077316e-06,
+      "loss": 0.9224,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999049142991246e-06,
+      "loss": 0.9951,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999035608265489e-06,
+      "loss": 1.0044,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999021977900303e-06,
+      "loss": 1.0044,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999008251895947e-06,
+      "loss": 1.0181,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998994430252688e-06,
+      "loss": 0.9458,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998980512970787e-06,
+      "loss": 0.9658,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998966500050512e-06,
+      "loss": 1.0566,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998952391492129e-06,
+      "loss": 1.0186,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998938187295912e-06,
+      "loss": 0.959,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99892388746213e-06,
+      "loss": 0.9565,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998909491991054e-06,
+      "loss": 0.9858,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998895000882964e-06,
+      "loss": 0.9736,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998880414138135e-06,
+      "loss": 1.0542,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998865731756846e-06,
+      "loss": 1.002,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998850953739378e-06,
+      "loss": 0.9663,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998836080086015e-06,
+      "loss": 1.0161,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99882111079704e-06,
+      "loss": 1.0811,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99880604587274e-06,
+      "loss": 1.0269,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998790885313403e-06,
+      "loss": 1.0396,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99877562911932e-06,
+      "loss": 1.0566,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998760277290783e-06,
+      "loss": 0.9751,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998744829828082e-06,
+      "loss": 1.0405,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998729286731517e-06,
+      "loss": 0.9946,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998713648001384e-06,
+      "loss": 1.021,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998697913637982e-06,
+      "loss": 0.999,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99868208364161e-06,
+      "loss": 0.9141,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998666158012575e-06,
+      "loss": 0.9727,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99865013675118e-06,
+      "loss": 0.9531,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99863401985773e-06,
+      "loss": 0.3091,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998617807332534e-06,
+      "loss": 0.3154,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998601499175902e-06,
+      "loss": 1.0298,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998585095388146e-06,
+      "loss": 1.0361,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998568595969583e-06,
+      "loss": 1.0195,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998552000920525e-06,
+      "loss": 0.9272,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998535310241289e-06,
+      "loss": 0.9536,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998518523932197e-06,
+      "loss": 1.0244,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99850164199357e-06,
+      "loss": 0.9399,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998484664425728e-06,
+      "loss": 1.0713,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998467591228998e-06,
+      "loss": 0.9897,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998450422403707e-06,
+      "loss": 0.9531,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998433157950182e-06,
+      "loss": 1.0601,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998415797868756e-06,
+      "loss": 1.0068,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998398342159757e-06,
+      "loss": 0.9248,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998380790823523e-06,
+      "loss": 0.9248,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998363143860387e-06,
+      "loss": 1.0073,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998345401270688e-06,
+      "loss": 1.0439,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998327563054765e-06,
+      "loss": 0.9639,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998309629212958e-06,
+      "loss": 1.0151,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998291599745613e-06,
+      "loss": 0.9707,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998273474653074e-06,
+      "loss": 0.9395,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998255253935686e-06,
+      "loss": 0.9912,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998236937593799e-06,
+      "loss": 1.0122,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998218525627764e-06,
+      "loss": 0.9604,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998200018037931e-06,
+      "loss": 1.042,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998181414824655e-06,
+      "loss": 1.0293,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998162715988293e-06,
+      "loss": 1.0479,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998143921529203e-06,
+      "loss": 1.0161,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998125031447743e-06,
+      "loss": 0.9917,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998106045744275e-06,
+      "loss": 1.0376,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998086964419162e-06,
+      "loss": 0.9424,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998067787472772e-06,
+      "loss": 0.9443,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998048514905465e-06,
+      "loss": 1.0112,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998029146717617e-06,
+      "loss": 1.0063,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998009682909595e-06,
+      "loss": 0.9883,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997990123481772e-06,
+      "loss": 1.0039,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997970468434522e-06,
+      "loss": 0.9761,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99795071776822e-06,
+      "loss": 1.0376,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997930871483246e-06,
+      "loss": 1.0371,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997910929579978e-06,
+      "loss": 0.9888,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997890892058799e-06,
+      "loss": 1.0205,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99787075892009e-06,
+      "loss": 0.979,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99785053016424e-06,
+      "loss": 1.0693,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997830205791631e-06,
+      "loss": 1.0005,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997809785802657e-06,
+      "loss": 0.9619,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997789270197704e-06,
+      "loss": 0.979,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997768658977168e-06,
+      "loss": 0.9883,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997747952141443e-06,
+      "loss": 0.9229,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997727149690923e-06,
+      "loss": 1.0205,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997706251626006e-06,
+      "loss": 1.0498,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997685257947096e-06,
+      "loss": 0.9521,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997664168654591e-06,
+      "loss": 0.978,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997642983748894e-06,
+      "loss": 0.9912,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997621703230412e-06,
+      "loss": 0.9756,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99760032709955e-06,
+      "loss": 1.0239,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99757885535672e-06,
+      "loss": 0.2911,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99755728800233e-06,
+      "loss": 0.9702,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997535625036795e-06,
+      "loss": 1.0186,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997513866460528e-06,
+      "loss": 1.0093,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997492012273946e-06,
+      "loss": 0.9873,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997470062477467e-06,
+      "loss": 1.0264,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99744801707151e-06,
+      "loss": 1.0063,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997425876056498e-06,
+      "loss": 0.9819,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997403639432855e-06,
+      "loss": 0.9741,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997381307201003e-06,
+      "loss": 0.9932,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997358879361373e-06,
+      "loss": 0.9976,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997336355914391e-06,
+      "loss": 0.9507,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997313736860492e-06,
+      "loss": 0.9854,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997291022200107e-06,
+      "loss": 1.0176,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997268211933668e-06,
+      "loss": 1.0181,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997245306061614e-06,
+      "loss": 1.0342,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997222304584383e-06,
+      "loss": 0.9766,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997199207502415e-06,
+      "loss": 1.022,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99717601481615e-06,
+      "loss": 0.9775,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997152726526034e-06,
+      "loss": 0.9487,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997129342632512e-06,
+      "loss": 0.981,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997105863136031e-06,
+      "loss": 1.0435,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99708228803704e-06,
+      "loss": 0.9595,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997058617335992e-06,
+      "loss": 0.9629,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997034851033339e-06,
+      "loss": 0.9375,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997010989129533e-06,
+      "loss": 0.9888,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996987031625031e-06,
+      "loss": 1.0312,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996962978520295e-06,
+      "loss": 0.9805,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996938829815783e-06,
+      "loss": 1.0352,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996914585511956e-06,
+      "loss": 0.9741,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99689024560928e-06,
+      "loss": 1.0835,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996865810108218e-06,
+      "loss": 0.9985,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99684127900924e-06,
+      "loss": 1.0513,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996816652312815e-06,
+      "loss": 1.0215,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996791930019412e-06,
+      "loss": 1.0103,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996767112129507e-06,
+      "loss": 1.0244,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99674219864357e-06,
+      "loss": 0.957,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996717189562084e-06,
+      "loss": 0.981,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996692084885524e-06,
+      "loss": 0.9438,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99666688461437e-06,
+      "loss": 0.9849,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996641588749104e-06,
+      "loss": 1.0708,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996616197290212e-06,
+      "loss": 0.9951,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996590710238176e-06,
+      "loss": 1.0244,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99656512759349e-06,
+      "loss": 1.0054,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996539449356637e-06,
+      "loss": 0.9717,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99651367552811e-06,
+      "loss": 1.0386,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996487806108404e-06,
+      "loss": 0.9883,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996461841098013e-06,
+      "loss": 1.0127,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996435780497434e-06,
+      "loss": 1.0088,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996409624307165e-06,
+      "loss": 0.9512,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996383372527704e-06,
+      "loss": 1.0444,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996357025159557e-06,
+      "loss": 1.0127,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996330582203229e-06,
+      "loss": 0.9751,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996304043659221e-06,
+      "loss": 1.0435,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996277409528046e-06,
+      "loss": 0.9771,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996250679810208e-06,
+      "loss": 1.0156,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996223854506225e-06,
+      "loss": 1.0146,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996196933616604e-06,
+      "loss": 0.4048,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996169917141864e-06,
+      "loss": 0.9839,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99614280508252e-06,
+      "loss": 0.9395,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99611559743909e-06,
+      "loss": 0.9702,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996088294212098e-06,
+      "loss": 1.0532,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996060895402062e-06,
+      "loss": 0.9741,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996033401009508e-06,
+      "loss": 1.0117,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996005811034964e-06,
+      "loss": 0.9585,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995978125478956e-06,
+      "loss": 0.9937,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995950344342014e-06,
+      "loss": 0.3467,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995922467624668e-06,
+      "loss": 0.3639,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995894495327453e-06,
+      "loss": 1.0303,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995866427450903e-06,
+      "loss": 0.9932,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995838263995557e-06,
+      "loss": 0.9702,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995810004961952e-06,
+      "loss": 1.0605,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99578165035063e-06,
+      "loss": 0.9922,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995753200162132e-06,
+      "loss": 0.9248,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995724654397002e-06,
+      "loss": 0.9956,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995696013055789e-06,
+      "loss": 1.0278,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995667276139037e-06,
+      "loss": 1.0366,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9956384436473e-06,
+      "loss": 0.957,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995609515581126e-06,
+      "loss": 0.9775,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99558049194107e-06,
+      "loss": 0.938,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995551372727688e-06,
+      "loss": 0.9814,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995522157941536e-06,
+      "loss": 1.021,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99549284758317e-06,
+      "loss": 1.0273,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995463441653157e-06,
+      "loss": 0.9448,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995433940152058e-06,
+      "loss": 1.0039,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995404343080433e-06,
+      "loss": 0.9106,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995374650438851e-06,
+      "loss": 1.0029,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995344862227883e-06,
+      "loss": 0.979,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995314978448094e-06,
+      "loss": 0.9897,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995284999100058e-06,
+      "loss": 1.0195,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995254924184348e-06,
+      "loss": 0.9365,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99522475370154e-06,
+      "loss": 0.9888,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995194487652212e-06,
+      "loss": 1.0127,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995164126036941e-06,
+      "loss": 0.3079,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99513366885631e-06,
+      "loss": 0.9927,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995103116110899e-06,
+      "loss": 1.0283,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995072467801294e-06,
+      "loss": 0.9834,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995041723928083e-06,
+      "loss": 0.9731,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995010884491852e-06,
+      "loss": 0.8984,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994979949493191e-06,
+      "loss": 0.3062,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994948918932695e-06,
+      "loss": 1.0254,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994917792810953e-06,
+      "loss": 0.9482,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994886571128561e-06,
+      "loss": 0.9912,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99485525388612e-06,
+      "loss": 0.9849,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994823841084227e-06,
+      "loss": 0.9961,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994792332723482e-06,
+      "loss": 1.001,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99476072880449e-06,
+      "loss": 1.0308,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994729029327854e-06,
+      "loss": 0.9272,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994697234294181e-06,
+      "loss": 0.4187,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99466534370408e-06,
+      "loss": 0.9873,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994633357558158e-06,
+      "loss": 0.9375,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994601275857032e-06,
+      "loss": 0.9482,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994569098601312e-06,
+      "loss": 0.9805,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994536825791615e-06,
+      "loss": 1.0308,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994504457428557e-06,
+      "loss": 1.001,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99447199351276e-06,
+      "loss": 0.9292,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994439434044843e-06,
+      "loss": 0.9937,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99440677902543e-06,
+      "loss": 0.939,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994374028455143e-06,
+      "loss": 1.0186,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994341182334612e-06,
+      "loss": 0.9888,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994308240664463e-06,
+      "loss": 0.9795,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99427520344533e-06,
+      "loss": 0.3207,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99424207067784e-06,
+      "loss": 0.9629,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99420884236263e-06,
+      "loss": 1.062,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994175518500334e-06,
+      "loss": 1.0293,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994142099091592e-06,
+      "loss": 0.9907,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994108584137041e-06,
+      "loss": 0.917,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994074973637322e-06,
+      "loss": 0.9824,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994041267593082e-06,
+      "loss": 1.0049,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99400746600496e-06,
+      "loss": 1.0479,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993973568873608e-06,
+      "loss": 1.0122,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993939576199669e-06,
+      "loss": 0.9756,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993905487983799e-06,
+      "loss": 0.9634,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993871304226648e-06,
+      "loss": 0.3547,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993837024928868e-06,
+      "loss": 0.938,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993802650091117e-06,
+      "loss": 0.9897,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993768179714052e-06,
+      "loss": 0.9849,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993733613798332e-06,
+      "loss": 0.9536,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99369895234462e-06,
+      "loss": 1.0,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993664195353576e-06,
+      "loss": 1.0259,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993629342825868e-06,
+      "loss": 1.0103,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993594394762162e-06,
+      "loss": 0.9106,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993559351163127e-06,
+      "loss": 0.9741,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99352421202943e-06,
+      "loss": 1.0176,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993488977361749e-06,
+      "loss": 0.9429,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993453647160753e-06,
+      "loss": 0.9912,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993418221427122e-06,
+      "loss": 1.0254,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993382700161531e-06,
+      "loss": 0.9922,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993347083364659e-06,
+      "loss": 0.9678,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99331137103719e-06,
+      "loss": 0.9546,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993275563179807e-06,
+      "loss": 0.9907,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993239659793193e-06,
+      "loss": 0.9688,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993203660878036e-06,
+      "loss": 0.9546,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993167566435024e-06,
+      "loss": 1.0898,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993131376464848e-06,
+      "loss": 0.9521,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993095090968202e-06,
+      "loss": 0.9756,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993058709945778e-06,
+      "loss": 0.96,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993022233398273e-06,
+      "loss": 1.0786,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992985661326384e-06,
+      "loss": 0.9448,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992948993730814e-06,
+      "loss": 0.9199,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992912230612258e-06,
+      "loss": 0.9863,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992875371971424e-06,
+      "loss": 1.0239,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992838417809019e-06,
+      "loss": 0.96,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992801368125745e-06,
+      "loss": 0.9883,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992764222922312e-06,
+      "loss": 0.9326,
+      "step": 1092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992726982199435e-06,
+      "loss": 1.0068,
+      "step": 1093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992689645957822e-06,
+      "loss": 0.9868,
+      "step": 1094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992652214198188e-06,
+      "loss": 0.9639,
+      "step": 1095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99261468692125e-06,
+      "loss": 0.9722,
+      "step": 1096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992577064127726e-06,
+      "loss": 1.0078,
+      "step": 1097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992539345818336e-06,
+      "loss": 0.3662,
+      "step": 1098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992501531993799e-06,
+      "loss": 0.9785,
+      "step": 1099
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992463622654843e-06,
+      "loss": 1.0317,
+      "step": 1100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992425617802189e-06,
+      "loss": 1.0078,
+      "step": 1101
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992387517436566e-06,
+      "loss": 1.0107,
+      "step": 1102
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992349321558703e-06,
+      "loss": 1.0254,
+      "step": 1103
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99231103016933e-06,
+      "loss": 0.9932,
+      "step": 1104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992272643269181e-06,
+      "loss": 0.9756,
+      "step": 1105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992234160858989e-06,
+      "loss": 1.0264,
+      "step": 1106
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992195582939491e-06,
+      "loss": 1.0396,
+      "step": 1107
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992156909511424e-06,
+      "loss": 0.9849,
+      "step": 1108
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99211814057553e-06,
+      "loss": 0.9937,
+      "step": 1109
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992079276132549e-06,
+      "loss": 0.9243,
+      "step": 1110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992040316183225e-06,
+      "loss": 0.9683,
+      "step": 1111
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992001260728303e-06,
+      "loss": 0.9927,
+      "step": 1112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991962109768529e-06,
+      "loss": 1.002,
+      "step": 1113
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991922863304655e-06,
+      "loss": 0.9634,
+      "step": 1114
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99188352133743e-06,
+      "loss": 0.9565,
+      "step": 1115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991844083867607e-06,
+      "loss": 0.9741,
+      "step": 1116
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99180455089594e-06,
+      "loss": 0.9541,
+      "step": 1117
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991764922423189e-06,
+      "loss": 0.2863,
+      "step": 1118
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991725198450105e-06,
+      "loss": 1.0308,
+      "step": 1119
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991685378977454e-06,
+      "loss": 0.958,
+      "step": 1120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991645464005994e-06,
+      "loss": 1.0137,
+      "step": 1121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991605453536493e-06,
+      "loss": 0.9644,
+      "step": 1122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991565347569711e-06,
+      "loss": 1.0156,
+      "step": 1123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99152514610642e-06,
+      "loss": 0.9458,
+      "step": 1124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991484849147388e-06,
+      "loss": 0.999,
+      "step": 1125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991444456693385e-06,
+      "loss": 1.0039,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991403968745182e-06,
+      "loss": 1.0112,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991363385303556e-06,
+      "loss": 0.9639,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991322706369285e-06,
+      "loss": 1.0552,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991281931943143e-06,
+      "loss": 0.98,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991241062025914e-06,
+      "loss": 1.0371,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991200096618379e-06,
+      "loss": 0.9658,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991159035721319e-06,
+      "loss": 0.5138,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991117879335523e-06,
+      "loss": 0.9497,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991076627461776e-06,
+      "loss": 0.9255,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991035280100869e-06,
+      "loss": 0.9629,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990993837253592e-06,
+      "loss": 1.0308,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990952298920736e-06,
+      "loss": 0.9873,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9909106651031e-06,
+      "loss": 0.9639,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990868935801477e-06,
+      "loss": 0.298,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990827111016668e-06,
+      "loss": 1.0156,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99078519074947e-06,
+      "loss": 0.9302,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990743175000688e-06,
+      "loss": 0.9702,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990701063771125e-06,
+      "loss": 0.9771,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990658857061586e-06,
+      "loss": 0.9678,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990616554872879e-06,
+      "loss": 0.9478,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990574157205813e-06,
+      "loss": 0.9751,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990531664061198e-06,
+      "loss": 0.9663,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99048907543985e-06,
+      "loss": 0.9585,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99044639134258e-06,
+      "loss": 0.9443,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990403611770207e-06,
+      "loss": 1.0449,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99036073672355e-06,
+      "loss": 1.0234,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990317766203429e-06,
+      "loss": 1.0435,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990274700210666e-06,
+      "loss": 1.0015,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99023153874608e-06,
+      "loss": 1.02,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990188281810504e-06,
+      "loss": 0.9902,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990144929404763e-06,
+      "loss": 1.0088,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990101481529684e-06,
+      "loss": 1.0093,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990057938186102e-06,
+      "loss": 0.9185,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990014299374848e-06,
+      "loss": 0.9336,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989970565096756e-06,
+      "loss": 0.9873,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989926735352665e-06,
+      "loss": 0.9575,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989882810143412e-06,
+      "loss": 0.9551,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98983878946984e-06,
+      "loss": 0.9492,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989794673332786e-06,
+      "loss": 1.0068,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989750461733096e-06,
+      "loss": 1.0005,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98970615467162e-06,
+      "loss": 0.9395,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9896617521492e-06,
+      "loss": 1.0229,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98961725416669e-06,
+      "loss": 0.916,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989572660724937e-06,
+      "loss": 0.9497,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989527971824799e-06,
+      "loss": 0.9639,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989483187467128e-06,
+      "loss": 0.9956,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989438307652778e-06,
+      "loss": 0.9805,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989393332382614e-06,
+      "loss": 0.9585,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989348261657492e-06,
+      "loss": 0.9478,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989303095478276e-06,
+      "loss": 1.0122,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98925783384583e-06,
+      "loss": 0.9487,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98921247676102e-06,
+      "loss": 0.9287,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989167024224711e-06,
+      "loss": 1.0161,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989121476237776e-06,
+      "loss": 0.3608,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989075832801087e-06,
+      "loss": 0.9824,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989030093915512e-06,
+      "loss": 0.3109,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988984259581934e-06,
+      "loss": 0.9785,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988938329801222e-06,
+      "loss": 1.0142,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988892304574257e-06,
+      "loss": 1.0063,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988846183901924e-06,
+      "loss": 0.9507,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9887999677851e-06,
+      "loss": 1.0029,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988753656224674e-06,
+      "loss": 0.9829,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988707249221527e-06,
+      "loss": 1.0283,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988660746776549e-06,
+      "loss": 1.0244,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988614148890631e-06,
+      "loss": 0.9468,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988567455564662e-06,
+      "loss": 1.0132,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988520666799538e-06,
+      "loss": 0.9946,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988473782596151e-06,
+      "loss": 1.0269,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988426802955401e-06,
+      "loss": 0.9976,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988379727878187e-06,
+      "loss": 1.0029,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988332557365404e-06,
+      "loss": 0.9546,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988285291417961e-06,
+      "loss": 0.9922,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98823793003676e-06,
+      "loss": 1.0107,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988190473222706e-06,
+      "loss": 0.9468,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988142920976708e-06,
+      "loss": 0.9917,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988095273299676e-06,
+      "loss": 1.0571,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988047530192521e-06,
+      "loss": 1.0415,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987999691656156e-06,
+      "loss": 0.9731,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987951757691496e-06,
+      "loss": 1.0439,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987903728299461e-06,
+      "loss": 0.2844,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987855603480968e-06,
+      "loss": 1.0562,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987807383236937e-06,
+      "loss": 0.9766,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987759067568288e-06,
+      "loss": 0.9023,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98771065647595e-06,
+      "loss": 0.9053,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987662149960848e-06,
+      "loss": 1.001,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987613548023911e-06,
+      "loss": 1.0317,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987564850666065e-06,
+      "loss": 0.9956,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987516057888245e-06,
+      "loss": 1.021,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987467169691382e-06,
+      "loss": 0.9854,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987418186076416e-06,
+      "loss": 0.9756,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987369107044279e-06,
+      "loss": 0.9202,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987319932595911e-06,
+      "loss": 0.9604,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987270662732255e-06,
+      "loss": 1.0215,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987221297454251e-06,
+      "loss": 0.9258,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987171836762846e-06,
+      "loss": 1.0366,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987122280658985e-06,
+      "loss": 1.02,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987072629143615e-06,
+      "loss": 0.958,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987022882217688e-06,
+      "loss": 0.9053,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986973039882152e-06,
+      "loss": 1.0273,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986923102137966e-06,
+      "loss": 0.9932,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986873068986082e-06,
+      "loss": 0.8735,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986822940427458e-06,
+      "loss": 1.0112,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986772716463053e-06,
+      "loss": 0.9873,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986722397093827e-06,
+      "loss": 1.0273,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986671982320744e-06,
+      "loss": 0.9448,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986621472144768e-06,
+      "loss": 1.0317,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986570866566865e-06,
+      "loss": 0.9971,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986520165588004e-06,
+      "loss": 0.9922,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986469369209155e-06,
+      "loss": 1.0396,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986418477431287e-06,
+      "loss": 0.9419,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986367490255378e-06,
+      "loss": 0.9697,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9863164076824e-06,
+      "loss": 0.9546,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986265229713332e-06,
+      "loss": 0.9487,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986213956349154e-06,
+      "loss": 1.0176,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986162587590844e-06,
+      "loss": 0.9443,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986111123439388e-06,
+      "loss": 0.9531,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98605956389577e-06,
+      "loss": 1.0254,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986007908960973e-06,
+      "loss": 0.9546,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985956158635988e-06,
+      "loss": 1.0,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985904312921806e-06,
+      "loss": 1.0356,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985852371819416e-06,
+      "loss": 0.9858,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985800335329815e-06,
+      "loss": 0.9014,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985748203453995e-06,
+      "loss": 0.9429,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985695976192957e-06,
+      "loss": 0.9839,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985643653547699e-06,
+      "loss": 0.9922,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98559123551922e-06,
+      "loss": 0.9937,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985538722108525e-06,
+      "loss": 1.0571,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98548611331662e-06,
+      "loss": 0.9873,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985433409144508e-06,
+      "loss": 0.9556,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985380609593198e-06,
+      "loss": 1.0054,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9853277146637e-06,
+      "loss": 1.0073,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985274724357028e-06,
+      "loss": 0.8929,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985221638674196e-06,
+      "loss": 0.9927,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985168457616217e-06,
+      "loss": 0.9575,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985115181184111e-06,
+      "loss": 0.3022,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985061809378894e-06,
+      "loss": 0.9907,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98500834220159e-06,
+      "loss": 1.0542,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984954779653222e-06,
+      "loss": 0.9839,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984901121734812e-06,
+      "loss": 1.0034,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98484736844739e-06,
+      "loss": 0.9795,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984793519791982e-06,
+      "loss": 0.9868,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984739575769619e-06,
+      "loss": 1.0049,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984685536381332e-06,
+      "loss": 0.283,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984631401628157e-06,
+      "loss": 0.9932,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984577171511127e-06,
+      "loss": 0.9443,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984522846031284e-06,
+      "loss": 0.8975,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984468425189662e-06,
+      "loss": 1.0562,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984413908987305e-06,
+      "loss": 0.9922,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984359297425255e-06,
+      "loss": 1.0435,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984304590504558e-06,
+      "loss": 0.979,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98424978822626e-06,
+      "loss": 0.3745,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984194890591409e-06,
+      "loss": 0.9614,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984139897601055e-06,
+      "loss": 1.0181,
+      "step": 1279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98408480925625e-06,
+      "loss": 0.957,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98402962555805e-06,
+      "loss": 0.9761,
+      "step": 1281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98397434650751e-06,
+      "loss": 0.9844,
+      "step": 1282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983918972105687e-06,
+      "loss": 0.9731,
+      "step": 1283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983863502353638e-06,
+      "loss": 0.9785,
+      "step": 1284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983807937252429e-06,
+      "loss": 0.9824,
+      "step": 1285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98375227680312e-06,
+      "loss": 0.9517,
+      "step": 1286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983696521006776e-06,
+      "loss": 0.978,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983640669864465e-06,
+      "loss": 1.0093,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983584723377255e-06,
+      "loss": 0.9346,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983528681546215e-06,
+      "loss": 0.9546,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983472544372419e-06,
+      "loss": 0.9829,
+      "step": 1291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983416311856939e-06,
+      "loss": 0.9741,
+      "step": 1292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983359984000853e-06,
+      "loss": 0.9951,
+      "step": 1293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983303560805239e-06,
+      "loss": 1.0034,
+      "step": 1294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983247042271174e-06,
+      "loss": 1.043,
+      "step": 1295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983190428399741e-06,
+      "loss": 0.9082,
+      "step": 1296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983133719192023e-06,
+      "loss": 0.9673,
+      "step": 1297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983076914649101e-06,
+      "loss": 1.0142,
+      "step": 1298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983020014772069e-06,
+      "loss": 0.9248,
+      "step": 1299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98296301956201e-06,
+      "loss": 1.0342,
+      "step": 1300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982905929020017e-06,
+      "loss": 0.959,
+      "step": 1301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982848743147182e-06,
+      "loss": 1.0088,
+      "step": 1302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982791461944597e-06,
+      "loss": 0.9585,
+      "step": 1303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98273408541336e-06,
+      "loss": 0.9604,
+      "step": 1304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98267661355457e-06,
+      "loss": 0.9233,
+      "step": 1305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982619046369321e-06,
+      "loss": 0.9893,
+      "step": 1306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982561383858721e-06,
+      "loss": 0.9736,
+      "step": 1307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98250362602387e-06,
+      "loss": 0.9985,
+      "step": 1308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982445772865873e-06,
+      "loss": 0.9204,
+      "step": 1309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982387824385835e-06,
+      "loss": 0.9951,
+      "step": 1310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98232978058487e-06,
+      "loss": 0.9995,
+      "step": 1311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982271641464083e-06,
+      "loss": 0.9634,
+      "step": 1312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982213407024588e-06,
+      "loss": 0.3892,
+      "step": 1313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982155077267502e-06,
+      "loss": 0.9985,
+      "step": 1314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982096652193936e-06,
+      "loss": 1.0552,
+      "step": 1315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982038131805009e-06,
+      "loss": 0.9902,
+      "step": 1316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981979516101845e-06,
+      "loss": 1.0244,
+      "step": 1317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981920805085561e-06,
+      "loss": 0.981,
+      "step": 1318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981861998757282e-06,
+      "loss": 0.9937,
+      "step": 1319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981803097118133e-06,
+      "loss": 0.9551,
+      "step": 1320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981744100169239e-06,
+      "loss": 0.9683,
+      "step": 1321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98168500791173e-06,
+      "loss": 0.8975,
+      "step": 1322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981625820346737e-06,
+      "loss": 1.0151,
+      "step": 1323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981566537475392e-06,
+      "loss": 0.9087,
+      "step": 1324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981507159298828e-06,
+      "loss": 0.978,
+      "step": 1325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981447685818184e-06,
+      "loss": 0.9268,
+      "step": 1326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981388117034596e-06,
+      "loss": 0.9814,
+      "step": 1327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981328452949204e-06,
+      "loss": 0.9448,
+      "step": 1328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981268693563147e-06,
+      "loss": 1.1035,
+      "step": 1329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981208838877572e-06,
+      "loss": 0.9995,
+      "step": 1330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981148888893622e-06,
+      "loss": 1.0322,
+      "step": 1331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981088843612446e-06,
+      "loss": 0.9927,
+      "step": 1332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98102870303519e-06,
+      "loss": 0.9912,
+      "step": 1333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980968467163003e-06,
+      "loss": 0.9331,
+      "step": 1334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980908135997043e-06,
+      "loss": 0.9795,
+      "step": 1335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980847709538461e-06,
+      "loss": 0.3198,
+      "step": 1336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980787187788414e-06,
+      "loss": 1.0093,
+      "step": 1337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980726570748061e-06,
+      "loss": 0.9722,
+      "step": 1338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980665858418558e-06,
+      "loss": 0.9458,
+      "step": 1339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980605050801068e-06,
+      "loss": 0.9434,
+      "step": 1340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980544147896756e-06,
+      "loss": 0.979,
+      "step": 1341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980483149706784e-06,
+      "loss": 1.0249,
+      "step": 1342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980422056232324e-06,
+      "loss": 0.9614,
+      "step": 1343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98036086747454e-06,
+      "loss": 0.978,
+      "step": 1344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980299583434606e-06,
+      "loss": 1.001,
+      "step": 1345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980238204113692e-06,
+      "loss": 0.9165,
+      "step": 1346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980176729512973e-06,
+      "loss": 1.0181,
+      "step": 1347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980115159633626e-06,
+      "loss": 1.0132,
+      "step": 1348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980053494476827e-06,
+      "loss": 0.8906,
+      "step": 1349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97999173404376e-06,
+      "loss": 0.959,
+      "step": 1350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9799298783356e-06,
+      "loss": 1.0117,
+      "step": 1351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979867927353536e-06,
+      "loss": 1.0376,
+      "step": 1352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979805881098752e-06,
+      "loss": 0.9155,
+      "step": 1353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979743739572433e-06,
+      "loss": 0.4269,
+      "step": 1354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979681502775769e-06,
+      "loss": 0.9487,
+      "step": 1355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979619170709952e-06,
+      "loss": 1.0093,
+      "step": 1356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979556743376174e-06,
+      "loss": 0.9771,
+      "step": 1357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979494220775624e-06,
+      "loss": 0.9912,
+      "step": 1358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979431602909508e-06,
+      "loss": 0.9521,
+      "step": 1359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979368889779018e-06,
+      "loss": 0.9722,
+      "step": 1360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979306081385354e-06,
+      "loss": 0.9878,
+      "step": 1361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979243177729717e-06,
+      "loss": 0.9263,
+      "step": 1362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979180178813312e-06,
+      "loss": 0.9956,
+      "step": 1363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979117084637345e-06,
+      "loss": 1.0103,
+      "step": 1364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979053895203022e-06,
+      "loss": 1.0103,
+      "step": 1365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978990610511551e-06,
+      "loss": 0.9653,
+      "step": 1366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978927230564146e-06,
+      "loss": 0.9829,
+      "step": 1367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978863755362014e-06,
+      "loss": 0.9756,
+      "step": 1368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978800184906376e-06,
+      "loss": 0.9443,
+      "step": 1369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978736519198442e-06,
+      "loss": 1.0107,
+      "step": 1370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978672758239434e-06,
+      "loss": 0.9956,
+      "step": 1371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978608902030571e-06,
+      "loss": 0.9497,
+      "step": 1372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978544950573075e-06,
+      "loss": 0.9976,
+      "step": 1373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978480903868167e-06,
+      "loss": 0.9893,
+      "step": 1374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978416761917076e-06,
+      "loss": 0.9937,
+      "step": 1375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978352524721027e-06,
+      "loss": 0.9844,
+      "step": 1376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978288192281247e-06,
+      "loss": 0.9229,
+      "step": 1377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978223764598972e-06,
+      "loss": 0.9937,
+      "step": 1378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97815924167543e-06,
+      "loss": 0.9863,
+      "step": 1379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978094623511858e-06,
+      "loss": 1.0137,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978029910109491e-06,
+      "loss": 0.9385,
+      "step": 1381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977965101469567e-06,
+      "loss": 1.0029,
+      "step": 1382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977900197593328e-06,
+      "loss": 0.9448,
+      "step": 1383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977835198482014e-06,
+      "loss": 0.9907,
+      "step": 1384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977770104136866e-06,
+      "loss": 0.9795,
+      "step": 1385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977704914559132e-06,
+      "loss": 0.9619,
+      "step": 1386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977639629750063e-06,
+      "loss": 0.9346,
+      "step": 1387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9775742497109e-06,
+      "loss": 0.9443,
+      "step": 1388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977508774442899e-06,
+      "loss": 0.3212,
+      "step": 1389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977443203947312e-06,
+      "loss": 1.0884,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977377538225393e-06,
+      "loss": 0.9175,
+      "step": 1391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977311777278399e-06,
+      "loss": 0.9487,
+      "step": 1392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977245921107586e-06,
+      "loss": 0.9795,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977179969714217e-06,
+      "loss": 0.9492,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97711392309955e-06,
+      "loss": 0.9966,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977047781264852e-06,
+      "loss": 0.9404,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976981544211388e-06,
+      "loss": 0.9683,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976915211940424e-06,
+      "loss": 1.0166,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976848784453229e-06,
+      "loss": 0.9814,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976782261751072e-06,
+      "loss": 1.0562,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97671564383523e-06,
+      "loss": 0.9897,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976648930706977e-06,
+      "loss": 0.9321,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976582122367585e-06,
+      "loss": 0.9609,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976515218818336e-06,
+      "loss": 0.959,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976448220060508e-06,
+      "loss": 0.936,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976381126095385e-06,
+      "loss": 0.9644,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97631393692425e-06,
+      "loss": 0.958,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976246652548384e-06,
+      "loss": 0.9556,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976179272969082e-06,
+      "loss": 0.9746,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976111798187626e-06,
+      "loss": 0.364,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97604422820531e-06,
+      "loss": 0.9658,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975976563023429e-06,
+      "loss": 0.9097,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975908802643273e-06,
+      "loss": 1.0205,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975840947066142e-06,
+      "loss": 0.9712,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975772996293334e-06,
+      "loss": 1.0098,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975704950326145e-06,
+      "loss": 0.9888,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97563680916588e-06,
+      "loss": 0.9995,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975568572813843e-06,
+      "loss": 0.9561,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97550024127134e-06,
+      "loss": 0.9839,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975431814539675e-06,
+      "loss": 1.0234,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975363292620159e-06,
+      "loss": 0.9663,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975294675514103e-06,
+      "loss": 1.0059,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97522596322282e-06,
+      "loss": 0.9521,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975157155747626e-06,
+      "loss": 1.041,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975088253089832e-06,
+      "loss": 0.9902,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975019255250764e-06,
+      "loss": 0.9272,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974950162231736e-06,
+      "loss": 0.9565,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974880974034073e-06,
+      "loss": 1.0288,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974811690659096e-06,
+      "loss": 0.9458,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974742312108132e-06,
+      "loss": 0.918,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974672838382511e-06,
+      "loss": 1.0425,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974603269483557e-06,
+      "loss": 0.9487,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974533605412604e-06,
+      "loss": 0.9839,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974463846170984e-06,
+      "loss": 0.9858,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974393991760032e-06,
+      "loss": 0.9712,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974324042181085e-06,
+      "loss": 0.979,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974253997435479e-06,
+      "loss": 1.0259,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974183857524557e-06,
+      "loss": 0.9756,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97411362244966e-06,
+      "loss": 1.0024,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974043292212129e-06,
+      "loss": 0.9648,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973972866813313e-06,
+      "loss": 0.937,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973902346254558e-06,
+      "loss": 0.9678,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973831730537214e-06,
+      "loss": 0.9624,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973761019662629e-06,
+      "loss": 0.9126,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973690213632158e-06,
+      "loss": 1.0498,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973619312447158e-06,
+      "loss": 0.978,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973548316108982e-06,
+      "loss": 0.9277,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973477224618991e-06,
+      "loss": 1.0,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973406037978541e-06,
+      "loss": 0.9644,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973334756188997e-06,
+      "loss": 0.4034,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973263379251723e-06,
+      "loss": 1.0234,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973191907168082e-06,
+      "loss": 1.0073,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973120339939446e-06,
+      "loss": 0.9819,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973048677567178e-06,
+      "loss": 0.9121,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972976920052654e-06,
+      "loss": 0.9683,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972905067397245e-06,
+      "loss": 0.9302,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972833119602325e-06,
+      "loss": 0.9482,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972761076669272e-06,
+      "loss": 0.936,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972688938599461e-06,
+      "loss": 0.9404,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972616705394279e-06,
+      "loss": 0.9277,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9725443770551e-06,
+      "loss": 0.978,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972471953583314e-06,
+      "loss": 0.9888,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972399434980302e-06,
+      "loss": 0.9614,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972326821247454e-06,
+      "loss": 0.9707,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972254112386158e-06,
+      "loss": 0.9302,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972181308397805e-06,
+      "loss": 0.9365,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97210840928379e-06,
+      "loss": 0.9331,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972035415045505e-06,
+      "loss": 0.9634,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971962325684348e-06,
+      "loss": 0.9941,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971889141201715e-06,
+      "loss": 0.9521,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971815861599011e-06,
+      "loss": 0.9756,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971742486877633e-06,
+      "loss": 0.9868,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97166901703899e-06,
+      "loss": 0.9912,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97159545208448e-06,
+      "loss": 1.0293,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971521792015516e-06,
+      "loss": 0.9121,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971448036833506e-06,
+      "loss": 1.0,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971374186539863e-06,
+      "loss": 0.9565,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971300241135996e-06,
+      "loss": 0.9624,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971226200623321e-06,
+      "loss": 0.9839,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971152065003257e-06,
+      "loss": 0.9995,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97107783427722e-06,
+      "loss": 0.9834,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971003508446629e-06,
+      "loss": 1.0195,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970929087512909e-06,
+      "loss": 1.0063,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970854571477482e-06,
+      "loss": 0.9878,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970779960341775e-06,
+      "loss": 0.9771,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970705254107212e-06,
+      "loss": 0.9854,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970630452775227e-06,
+      "loss": 0.9102,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970555556347248e-06,
+      "loss": 0.9551,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970480564824708e-06,
+      "loss": 0.9136,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970405478209043e-06,
+      "loss": 0.9487,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970330296501686e-06,
+      "loss": 0.9839,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970255019704082e-06,
+      "loss": 0.9434,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970179647817667e-06,
+      "loss": 0.999,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970104180843881e-06,
+      "loss": 1.0059,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970028618784171e-06,
+      "loss": 0.9702,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969952961639981e-06,
+      "loss": 0.9912,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969877209412762e-06,
+      "loss": 0.978,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969801362103958e-06,
+      "loss": 0.9268,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969725419715022e-06,
+      "loss": 0.9443,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96964938224741e-06,
+      "loss": 0.9575,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969573249702572e-06,
+      "loss": 0.3158,
+      "step": 1501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969497022081968e-06,
+      "loss": 0.9775,
+      "step": 1502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969420699387054e-06,
+      "loss": 0.9707,
+      "step": 1503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969344281619292e-06,
+      "loss": 1.0483,
+      "step": 1504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969267768780141e-06,
+      "loss": 0.9922,
+      "step": 1505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96919116087107e-06,
+      "loss": 0.9937,
+      "step": 1506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96911445789354e-06,
+      "loss": 1.022,
+      "step": 1507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96903765984902e-06,
+      "loss": 0.9644,
+      "step": 1508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96896076673898e-06,
+      "loss": 0.9717,
+      "step": 1509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968883778564891e-06,
+      "loss": 0.9561,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968806695328223e-06,
+      "loss": 0.98,
+      "step": 1511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968729517030455e-06,
+      "loss": 0.9268,
+      "step": 1512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96865224367306e-06,
+      "loss": 0.9824,
+      "step": 1513
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96857487525752e-06,
+      "loss": 0.9648,
+      "step": 1514
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968497411785311e-06,
+      "loss": 0.9873,
+      "step": 1515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96841985325792e-06,
+      "loss": 0.9639,
+      "step": 1516
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968342199676825e-06,
+      "loss": 0.9741,
+      "step": 1517
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968264451043517e-06,
+      "loss": 0.9053,
+      "step": 1518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96818660735948e-06,
+      "loss": 0.9839,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968108668626205e-06,
+      "loss": 0.9707,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96803063484518e-06,
+      "loss": 1.0483,
+      "step": 1521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.967952506017902e-06,
+      "loss": 0.9819,
+      "step": 1522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967874282145864e-06,
+      "loss": 1.0244,
+      "step": 1523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967795963230563e-06,
+      "loss": 0.9492,
+      "step": 1524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967717549273496e-06,
+      "loss": 1.0098,
+      "step": 1525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967639040276165e-06,
+      "loss": 0.9678,
+      "step": 1526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967560436240072e-06,
+      "loss": 0.9912,
+      "step": 1527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967481737166718e-06,
+      "loss": 0.9907,
+      "step": 1528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96740294305761e-06,
+      "loss": 0.3149,
+      "step": 1529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967324053914258e-06,
+      "loss": 0.9639,
+      "step": 1530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967245069738169e-06,
+      "loss": 0.9697,
+      "step": 1531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967165990530853e-06,
+      "loss": 0.9165,
+      "step": 1532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967086816293827e-06,
+      "loss": 0.9941,
+      "step": 1533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967007547028601e-06,
+      "loss": 0.8965,
+      "step": 1534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966928182736694e-06,
+      "loss": 0.9526,
+      "step": 1535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966848723419624e-06,
+      "loss": 0.9966,
+      "step": 1536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96676916907891e-06,
+      "loss": 0.9897,
+      "step": 1537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966689519716078e-06,
+      "loss": 0.3519,
+      "step": 1538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966609775332647e-06,
+      "loss": 0.9312,
+      "step": 1539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966529935930146e-06,
+      "loss": 1.0342,
+      "step": 1540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966450001510099e-06,
+      "loss": 0.8804,
+      "step": 1541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966369972074037e-06,
+      "loss": 0.8984,
+      "step": 1542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966289847623493e-06,
+      "loss": 0.9722,
+      "step": 1543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966209628159997e-06,
+      "loss": 0.9849,
+      "step": 1544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966129313685086e-06,
+      "loss": 0.3632,
+      "step": 1545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966048904200296e-06,
+      "loss": 0.9409,
+      "step": 1546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965968399707165e-06,
+      "loss": 0.9316,
+      "step": 1547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965887800207233e-06,
+      "loss": 0.9751,
+      "step": 1548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96580710570204e-06,
+      "loss": 0.9385,
+      "step": 1549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965726316193135e-06,
+      "loss": 0.9272,
+      "step": 1550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965645431682059e-06,
+      "loss": 0.9067,
+      "step": 1551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965564452170361e-06,
+      "loss": 0.9951,
+      "step": 1552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96548337765959e-06,
+      "loss": 0.9824,
+      "step": 1553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9654022081513e-06,
+      "loss": 0.9199,
+      "step": 1554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965320943647039e-06,
+      "loss": 0.9829,
+      "step": 1555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965239584148363e-06,
+      "loss": 0.9868,
+      "step": 1556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965158129656832e-06,
+      "loss": 0.9663,
+      "step": 1557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965076580173999e-06,
+      "loss": 0.9302,
+      "step": 1558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964994935701429e-06,
+      "loss": 0.9312,
+      "step": 1559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964913196240682e-06,
+      "loss": 0.9839,
+      "step": 1560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964831361793321e-06,
+      "loss": 0.9946,
+      "step": 1561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964749432360913e-06,
+      "loss": 0.9917,
+      "step": 1562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964667407945024e-06,
+      "loss": 0.3237,
+      "step": 1563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964585288547225e-06,
+      "loss": 1.0156,
+      "step": 1564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964503074169086e-06,
+      "loss": 0.9907,
+      "step": 1565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96442076481218e-06,
+      "loss": 0.9897,
+      "step": 1566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964338360478082e-06,
+      "loss": 0.27,
+      "step": 1567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964255861168368e-06,
+      "loss": 0.9253,
+      "step": 1568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964173266884618e-06,
+      "loss": 0.9478,
+      "step": 1569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964090577628409e-06,
+      "loss": 0.9399,
+      "step": 1570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964007793401325e-06,
+      "loss": 0.9546,
+      "step": 1571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96392491420495e-06,
+      "loss": 0.9385,
+      "step": 1572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96384194004087e-06,
+      "loss": 0.9194,
+      "step": 1573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963758870910672e-06,
+      "loss": 0.8804,
+      "step": 1574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963675706815942e-06,
+      "loss": 0.957,
+      "step": 1575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963592447758277e-06,
+      "loss": 0.9878,
+      "step": 1576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963509093739266e-06,
+      "loss": 0.9673,
+      "step": 1577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963425644760506e-06,
+      "loss": 0.9761,
+      "step": 1578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963342100823589e-06,
+      "loss": 0.9541,
+      "step": 1579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963258461930118e-06,
+      "loss": 0.9648,
+      "step": 1580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96317472808169e-06,
+      "loss": 0.9888,
+      "step": 1581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96309089927991e-06,
+      "loss": 1.0176,
+      "step": 1582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96300697552638e-06,
+      "loss": 0.9824,
+      "step": 1583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962922956822706e-06,
+      "loss": 0.9463,
+      "step": 1584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962838843170495e-06,
+      "loss": 1.0166,
+      "step": 1585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962754634571356e-06,
+      "loss": 1.0386,
+      "step": 1586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962670331026901e-06,
+      "loss": 0.981,
+      "step": 1587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962585932538742e-06,
+      "loss": 1.0078,
+      "step": 1588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962501439108494e-06,
+      "loss": 1.0635,
+      "step": 1589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962416850737775e-06,
+      "loss": 0.9521,
+      "step": 1590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9623321674282e-06,
+      "loss": 0.9194,
+      "step": 1591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962247389181391e-06,
+      "loss": 0.9702,
+      "step": 1592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962162515998972e-06,
+      "loss": 0.9844,
+      "step": 1593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962077547882563e-06,
+      "loss": 1.0005,
+      "step": 1594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961992484833792e-06,
+      "loss": 0.9463,
+      "step": 1595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961907326854286e-06,
+      "loss": 0.9443,
+      "step": 1596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961822073945672e-06,
+      "loss": 0.958,
+      "step": 1597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961736726109583e-06,
+      "loss": 1.0015,
+      "step": 1598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961651283347653e-06,
+      "loss": 1.0,
+      "step": 1599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961565745661515e-06,
+      "loss": 0.9136,
+      "step": 1600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961480113052806e-06,
+      "loss": 0.8862,
+      "step": 1601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961394385523164e-06,
+      "loss": 0.9702,
+      "step": 1602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96130856307423e-06,
+      "loss": 0.9785,
+      "step": 1603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961222645707643e-06,
+      "loss": 0.937,
+      "step": 1604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96113663342505e-06,
+      "loss": 0.9365,
+      "step": 1605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961050526228096e-06,
+      "loss": 0.9375,
+      "step": 1606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960964324118428e-06,
+      "loss": 1.0317,
+      "step": 1607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960878027097693e-06,
+      "loss": 1.0049,
+      "step": 1608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960791635167545e-06,
+      "loss": 0.9546,
+      "step": 1609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960705148329636e-06,
+      "loss": 0.9961,
+      "step": 1610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960618566585622e-06,
+      "loss": 0.9819,
+      "step": 1611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960531889937157e-06,
+      "loss": 0.9785,
+      "step": 1612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960445118385899e-06,
+      "loss": 0.9766,
+      "step": 1613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960358251933508e-06,
+      "loss": 0.939,
+      "step": 1614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960271290581648e-06,
+      "loss": 0.9307,
+      "step": 1615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960184234331985e-06,
+      "loss": 0.3676,
+      "step": 1616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960097083186177e-06,
+      "loss": 0.9438,
+      "step": 1617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960009837145898e-06,
+      "loss": 1.0103,
+      "step": 1618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959922496212815e-06,
+      "loss": 0.959,
+      "step": 1619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959835060388599e-06,
+      "loss": 0.9556,
+      "step": 1620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95974752967492e-06,
+      "loss": 0.9238,
+      "step": 1621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959659904073457e-06,
+      "loss": 1.0244,
+      "step": 1622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959572183585884e-06,
+      "loss": 0.9375,
+      "step": 1623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95948436821388e-06,
+      "loss": 0.4421,
+      "step": 1624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959396457959126e-06,
+      "loss": 1.0728,
+      "step": 1625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959308452823302e-06,
+      "loss": 0.9565,
+      "step": 1626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959220352808092e-06,
+      "loss": 0.9316,
+      "step": 1627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959132157915182e-06,
+      "loss": 0.9609,
+      "step": 1628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95904386814626e-06,
+      "loss": 0.9507,
+      "step": 1629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958955483503014e-06,
+      "loss": 0.9409,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958867003987133e-06,
+      "loss": 0.9746,
+      "step": 1631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958778429600315e-06,
+      "loss": 0.9971,
+      "step": 1632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958689760344251e-06,
+      "loss": 1.0166,
+      "step": 1633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958600996220638e-06,
+      "loss": 0.9785,
+      "step": 1634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958512137231173e-06,
+      "loss": 0.9766,
+      "step": 1635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958423183377557e-06,
+      "loss": 0.9546,
+      "step": 1636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958334134661493e-06,
+      "loss": 1.0034,
+      "step": 1637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958244991084684e-06,
+      "loss": 0.9785,
+      "step": 1638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958155752648834e-06,
+      "loss": 1.0288,
+      "step": 1639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958066419355652e-06,
+      "loss": 0.9932,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957976991206847e-06,
+      "loss": 0.9443,
+      "step": 1641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957887468204129e-06,
+      "loss": 0.9897,
+      "step": 1642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95779785034921e-06,
+      "loss": 0.9736,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957708137643807e-06,
+      "loss": 0.9766,
+      "step": 1644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957618330089633e-06,
+      "loss": 0.9937,
+      "step": 1645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95752842768841e-06,
+      "loss": 1.0146,
+      "step": 1646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957438430441855e-06,
+      "loss": 0.9443,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957348338351692e-06,
+      "loss": 0.9688,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957258151419642e-06,
+      "loss": 1.0054,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957167869647433e-06,
+      "loss": 0.9409,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957077493036791e-06,
+      "loss": 0.9067,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956987021589446e-06,
+      "loss": 0.9976,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956896455307126e-06,
+      "loss": 0.9458,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956805794191567e-06,
+      "loss": 0.9927,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956715038244502e-06,
+      "loss": 0.2528,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956624187467667e-06,
+      "loss": 0.9565,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956533241862803e-06,
+      "loss": 0.9443,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956442201431645e-06,
+      "loss": 0.9658,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956351066175937e-06,
+      "loss": 1.0049,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956259836097424e-06,
+      "loss": 0.9419,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95616851119785e-06,
+      "loss": 0.9429,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956077091478964e-06,
+      "loss": 1.02,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95598557694251e-06,
+      "loss": 0.9932,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955893967590245e-06,
+      "loss": 0.981,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955802263423919e-06,
+      "loss": 0.9517,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955710464445284e-06,
+      "loss": 0.9834,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955618570656101e-06,
+      "loss": 0.9902,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955526582058125e-06,
+      "loss": 0.9287,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955434498653117e-06,
+      "loss": 1.0244,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955342320442836e-06,
+      "loss": 0.9341,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95525004742905e-06,
+      "loss": 0.9824,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95515767961352e-06,
+      "loss": 1.0127,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955065216998018e-06,
+      "loss": 0.9854,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954972659584307e-06,
+      "loss": 0.9888,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954880007374164e-06,
+      "loss": 0.9824,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954787260369357e-06,
+      "loss": 0.9722,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954694418571661e-06,
+      "loss": 0.9307,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954601481982856e-06,
+      "loss": 0.9023,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954508450604716e-06,
+      "loss": 0.9595,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954415324439021e-06,
+      "loss": 0.981,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954322103487556e-06,
+      "loss": 0.9932,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954228787752098e-06,
+      "loss": 0.9692,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95413537723444e-06,
+      "loss": 0.978,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954041871936363e-06,
+      "loss": 0.9897,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953948271859659e-06,
+      "loss": 0.9028,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953854577006119e-06,
+      "loss": 0.98,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953760787377535e-06,
+      "loss": 0.9507,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9536669029757e-06,
+      "loss": 0.9673,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953572923802412e-06,
+      "loss": 0.9907,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953478849859468e-06,
+      "loss": 0.9746,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953384681148666e-06,
+      "loss": 0.9775,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953290417671812e-06,
+      "loss": 0.9688,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953196059430705e-06,
+      "loss": 0.9834,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953101606427154e-06,
+      "loss": 0.978,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953007058662963e-06,
+      "loss": 0.9873,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952912416139942e-06,
+      "loss": 0.9761,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952817678859901e-06,
+      "loss": 0.9331,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952722846824654e-06,
+      "loss": 1.0156,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952627920036014e-06,
+      "loss": 0.9209,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952532898495797e-06,
+      "loss": 0.9248,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952437782205822e-06,
+      "loss": 0.9746,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95234257116791e-06,
+      "loss": 0.9751,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952247265383879e-06,
+      "loss": 0.9873,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952151864855554e-06,
+      "loss": 0.9414,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95205636958476e-06,
+      "loss": 0.897,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951960779573323e-06,
+      "loss": 1.0557,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951865094823075e-06,
+      "loss": 0.981,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951769315335843e-06,
+      "loss": 1.0317,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951673441113462e-06,
+      "loss": 0.9927,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951577472157768e-06,
+      "loss": 0.8779,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951481408470591e-06,
+      "loss": 0.353,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951385250053772e-06,
+      "loss": 0.9658,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951288996909152e-06,
+      "loss": 0.9634,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951192649038571e-06,
+      "loss": 0.9658,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95109620644387e-06,
+      "loss": 0.9365,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9509996691269e-06,
+      "loss": 0.9961,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950903037089504e-06,
+      "loss": 1.0068,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95080631033353e-06,
+      "loss": 1.0215,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95070948886083e-06,
+      "loss": 0.9849,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950612572673255e-06,
+      "loss": 0.9771,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950515561772662e-06,
+      "loss": 0.9731,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950418456160905e-06,
+      "loss": 0.9604,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950321255839842e-06,
+      "loss": 0.9487,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950223960811332e-06,
+      "loss": 0.9355,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950126571077239e-06,
+      "loss": 1.0386,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950029086639423e-06,
+      "loss": 0.9868,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949931507499749e-06,
+      "loss": 0.9243,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949833833660087e-06,
+      "loss": 0.9146,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949736065122303e-06,
+      "loss": 0.939,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949638201888268e-06,
+      "loss": 0.2936,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949540243959854e-06,
+      "loss": 1.0205,
+      "step": 1731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949442191338939e-06,
+      "loss": 0.9863,
+      "step": 1732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949344044027392e-06,
+      "loss": 1.061,
+      "step": 1733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949245802027094e-06,
+      "loss": 0.9717,
+      "step": 1734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949147465339928e-06,
+      "loss": 0.9897,
+      "step": 1735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94904903396777e-06,
+      "loss": 1.0171,
+      "step": 1736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948950507912504e-06,
+      "loss": 0.9819,
+      "step": 1737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948851887176019e-06,
+      "loss": 0.9722,
+      "step": 1738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948753171760198e-06,
+      "loss": 0.979,
+      "step": 1739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94865436166693e-06,
+      "loss": 0.9946,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948555456898105e-06,
+      "loss": 0.9971,
+      "step": 1741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948456457455618e-06,
+      "loss": 0.978,
+      "step": 1742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94835736334136e-06,
+      "loss": 0.9497,
+      "step": 1743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948258174557228e-06,
+      "loss": 0.9517,
+      "step": 1744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94815889110512e-06,
+      "loss": 0.9795,
+      "step": 1745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948059512986936e-06,
+      "loss": 0.9629,
+      "step": 1746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947960040204576e-06,
+      "loss": 1.0034,
+      "step": 1747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947860472759943e-06,
+      "loss": 0.9517,
+      "step": 1748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947760810654941e-06,
+      "loss": 0.918,
+      "step": 1749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94766105389148e-06,
+      "loss": 0.3356,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947561202471465e-06,
+      "loss": 0.9629,
+      "step": 1751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947461256396808e-06,
+      "loss": 1.0068,
+      "step": 1752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947361215669422e-06,
+      "loss": 0.9761,
+      "step": 1753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947261080291219e-06,
+      "loss": 0.9272,
+      "step": 1754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947160850264116e-06,
+      "loss": 0.9873,
+      "step": 1755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947060525590031e-06,
+      "loss": 1.0146,
+      "step": 1756
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946960106270882e-06,
+      "loss": 0.959,
+      "step": 1757
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946859592308589e-06,
+      "loss": 0.9648,
+      "step": 1758
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946758983705079e-06,
+      "loss": 0.9917,
+      "step": 1759
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946658280462273e-06,
+      "loss": 0.9878,
+      "step": 1760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9465574825821e-06,
+      "loss": 1.002,
+      "step": 1761
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946456590066487e-06,
+      "loss": 0.9375,
+      "step": 1762
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946355602917366e-06,
+      "loss": 0.9526,
+      "step": 1763
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946254521136668e-06,
+      "loss": 0.9277,
+      "step": 1764
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946153344726324e-06,
+      "loss": 0.9336,
+      "step": 1765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946052073688273e-06,
+      "loss": 0.9946,
+      "step": 1766
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945950708024455e-06,
+      "loss": 1.0,
+      "step": 1767
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945849247736803e-06,
+      "loss": 1.0127,
+      "step": 1768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94574769282726e-06,
+      "loss": 0.9814,
+      "step": 1769
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945646043297774e-06,
+      "loss": 1.0,
+      "step": 1770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945544299150284e-06,
+      "loss": 0.9756,
+      "step": 1771
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945442460386738e-06,
+      "loss": 0.9678,
+      "step": 1772
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945340527009086e-06,
+      "loss": 0.9883,
+      "step": 1773
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945238499019276e-06,
+      "loss": 0.9766,
+      "step": 1774
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94513637641926e-06,
+      "loss": 0.9722,
+      "step": 1775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945034159210993e-06,
+      "loss": 0.978,
+      "step": 1776
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944931847396431e-06,
+      "loss": 1.0229,
+      "step": 1777
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944829440977529e-06,
+      "loss": 1.0107,
+      "step": 1778
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944726939956249e-06,
+      "loss": 1.0503,
+      "step": 1779
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94462434433455e-06,
+      "loss": 0.9365,
+      "step": 1780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944521654114397e-06,
+      "loss": 1.0049,
+      "step": 1781
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944418869297752e-06,
+      "loss": 0.9673,
+      "step": 1782
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944315989886582e-06,
+      "loss": 0.9507,
+      "step": 1783
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944213015882855e-06,
+      "loss": 1.0166,
+      "step": 1784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944109947288544e-06,
+      "loss": 0.9688,
+      "step": 1785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944006784105617e-06,
+      "loss": 0.9209,
+      "step": 1786
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943903526336051e-06,
+      "loss": 0.9277,
+      "step": 1787
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94380017398182e-06,
+      "loss": 0.9219,
+      "step": 1788
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943696727044901e-06,
+      "loss": 0.9629,
+      "step": 1789
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943593185527273e-06,
+      "loss": 0.9702,
+      "step": 1790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943489549430917e-06,
+      "loss": 0.939,
+      "step": 1791
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943385818757816e-06,
+      "loss": 0.937,
+      "step": 1792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943281993509955e-06,
+      "loss": 1.0083,
+      "step": 1793
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943178073689318e-06,
+      "loss": 0.916,
+      "step": 1794
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943074059297897e-06,
+      "loss": 0.9795,
+      "step": 1795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94296995033768e-06,
+      "loss": 0.9253,
+      "step": 1796
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942865746810659e-06,
+      "loss": 0.9663,
+      "step": 1797
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942761448718825e-06,
+      "loss": 0.9609,
+      "step": 1798
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942657056064177e-06,
+      "loss": 0.9355,
+      "step": 1799
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942552568848712e-06,
+      "loss": 0.9819,
+      "step": 1800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942447987074426e-06,
+      "loss": 0.9463,
+      "step": 1801
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942343310743323e-06,
+      "loss": 0.8955,
+      "step": 1802
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942238539857403e-06,
+      "loss": 0.9736,
+      "step": 1803
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942133674418672e-06,
+      "loss": 0.9175,
+      "step": 1804
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942028714429137e-06,
+      "loss": 0.9551,
+      "step": 1805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941923659890804e-06,
+      "loss": 0.9546,
+      "step": 1806
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941818510805685e-06,
+      "loss": 0.9912,
+      "step": 1807
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941713267175788e-06,
+      "loss": 0.9209,
+      "step": 1808
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941607929003132e-06,
+      "loss": 0.9189,
+      "step": 1809
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941502496289728e-06,
+      "loss": 0.938,
+      "step": 1810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941396969037594e-06,
+      "loss": 0.3468,
+      "step": 1811
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94129134724875e-06,
+      "loss": 1.0024,
+      "step": 1812
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941185630925216e-06,
+      "loss": 0.9458,
+      "step": 1813
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941079820069015e-06,
+      "loss": 0.9644,
+      "step": 1814
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94097391468217e-06,
+      "loss": 0.9756,
+      "step": 1815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940867914766709e-06,
+      "loss": 0.9805,
+      "step": 1816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940761820324658e-06,
+      "loss": 0.9658,
+      "step": 1817
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940655631358047e-06,
+      "loss": 0.9644,
+      "step": 1818
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94054934786891e-06,
+      "loss": 0.9443,
+      "step": 1819
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94044296985928e-06,
+      "loss": 0.9648,
+      "step": 1820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940336497331187e-06,
+      "loss": 0.9946,
+      "step": 1821
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940229930286673e-06,
+      "loss": 0.9751,
+      "step": 1822
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940123268727777e-06,
+      "loss": 0.9634,
+      "step": 1823
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940016512656536e-06,
+      "loss": 0.9727,
+      "step": 1824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939909662074996e-06,
+      "loss": 0.9302,
+      "step": 1825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9398027169852e-06,
+      "loss": 0.9424,
+      "step": 1826
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939695677389194e-06,
+      "loss": 0.8984,
+      "step": 1827
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939588543289026e-06,
+      "loss": 0.9448,
+      "step": 1828
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939481314686743e-06,
+      "loss": 0.9331,
+      "step": 1829
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939373991584401e-06,
+      "loss": 0.9604,
+      "step": 1830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93926657398405e-06,
+      "loss": 0.9946,
+      "step": 1831
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939159061887747e-06,
+      "loss": 0.8994,
+      "step": 1832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939051455297548e-06,
+      "loss": 0.9634,
+      "step": 1833
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93894375421551e-06,
+      "loss": 0.8979,
+      "step": 1834
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938835958643698e-06,
+      "loss": 0.9751,
+      "step": 1835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93872806858417e-06,
+      "loss": 0.9424,
+      "step": 1836
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938620084038992e-06,
+      "loss": 0.9854,
+      "step": 1837
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938512005010228e-06,
+      "loss": 0.9937,
+      "step": 1838
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938403831499948e-06,
+      "loss": 0.9985,
+      "step": 1839
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93829556351022e-06,
+      "loss": 0.9629,
+      "step": 1840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93818720104312e-06,
+      "loss": 0.9741,
+      "step": 1841
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938078744100713e-06,
+      "loss": 0.9736,
+      "step": 1842
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93797019268508e-06,
+      "loss": 0.9512,
+      "step": 1843
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937861546798296e-06,
+      "loss": 0.96,
+      "step": 1844
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937752806442438e-06,
+      "loss": 0.918,
+      "step": 1845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937643971619588e-06,
+      "loss": 1.0396,
+      "step": 1846
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93753504233183e-06,
+      "loss": 0.9541,
+      "step": 1847
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937426018581245e-06,
+      "loss": 0.9937,
+      "step": 1848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93731690036992e-06,
+      "loss": 0.8965,
+      "step": 1849
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93720768769994e-06,
+      "loss": 0.9741,
+      "step": 1850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9370983805734e-06,
+      "loss": 0.8975,
+      "step": 1851
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936988978992388e-06,
+      "loss": 1.0044,
+      "step": 1852
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936879482958997e-06,
+      "loss": 0.9395,
+      "step": 1853
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93676989247532e-06,
+      "loss": 0.8828,
+      "step": 1854
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936660207543457e-06,
+      "loss": 0.3274,
+      "step": 1855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936550428165504e-06,
+      "loss": 0.9888,
+      "step": 1856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936440554343563e-06,
+      "loss": 0.9546,
+      "step": 1857
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936330586079737e-06,
+      "loss": 0.9492,
+      "step": 1858
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936220523376125e-06,
+      "loss": 0.9678,
+      "step": 1859
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93611036623484e-06,
+      "loss": 0.9619,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93600011465798e-06,
+      "loss": 0.9546,
+      "step": 1861
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935889768647664e-06,
+      "loss": 0.9727,
+      "step": 1862
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935779328205997e-06,
+      "loss": 0.9062,
+      "step": 1863
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935668793335093e-06,
+      "loss": 1.0005,
+      "step": 1864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935558164037068e-06,
+      "loss": 0.9434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935447440314038e-06,
+      "loss": 0.9546,
+      "step": 1866
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935336622168122e-06,
+      "loss": 0.9014,
+      "step": 1867
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93522570960144e-06,
+      "loss": 0.8955,
+      "step": 1868
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93511470261611e-06,
+      "loss": 0.9897,
+      "step": 1869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93500360121426e-06,
+      "loss": 0.8916,
+      "step": 1870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934892405398016e-06,
+      "loss": 0.9971,
+      "step": 1871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934781115169503e-06,
+      "loss": 0.3314,
+      "step": 1872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934669730530852e-06,
+      "loss": 0.9155,
+      "step": 1873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934558251484192e-06,
+      "loss": 0.2832,
+      "step": 1874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934446678031656e-06,
+      "loss": 0.9844,
+      "step": 1875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93433501017538e-06,
+      "loss": 0.9482,
+      "step": 1876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9342232479175e-06,
+      "loss": 0.9336,
+      "step": 1877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934111391260153e-06,
+      "loss": 0.9937,
+      "step": 1878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933999440205481e-06,
+      "loss": 0.9688,
+      "step": 1879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933887394755625e-06,
+      "loss": 0.9351,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933775254912727e-06,
+      "loss": 0.9897,
+      "step": 1881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933663020678934e-06,
+      "loss": 0.9331,
+      "step": 1882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933550692056393e-06,
+      "loss": 0.9409,
+      "step": 1883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933438269047254e-06,
+      "loss": 0.9619,
+      "step": 1884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933325751653667e-06,
+      "loss": 0.9438,
+      "step": 1885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933213139877781e-06,
+      "loss": 0.9399,
+      "step": 1886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933100433721758e-06,
+      "loss": 0.9541,
+      "step": 1887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932987633187749e-06,
+      "loss": 0.9331,
+      "step": 1888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932874738277911e-06,
+      "loss": 0.978,
+      "step": 1889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932761748994409e-06,
+      "loss": 0.9438,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9326486653394e-06,
+      "loss": 0.9155,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932535487315049e-06,
+      "loss": 0.9312,
+      "step": 1892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932422214923521e-06,
+      "loss": 0.9741,
+      "step": 1893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932308848166985e-06,
+      "loss": 0.8926,
+      "step": 1894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932195387047606e-06,
+      "loss": 1.0024,
+      "step": 1895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932081831567561e-06,
+      "loss": 0.9531,
+      "step": 1896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931968181729016e-06,
+      "loss": 1.0059,
+      "step": 1897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931854437534147e-06,
+      "loss": 0.9985,
+      "step": 1898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931740598985133e-06,
+      "loss": 0.8975,
+      "step": 1899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93162666608415e-06,
+      "loss": 0.8647,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931512638833376e-06,
+      "loss": 1.001,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931398517234993e-06,
+      "loss": 0.9053,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931284301291189e-06,
+      "loss": 0.939,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931169991004145e-06,
+      "loss": 0.937,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931055586376046e-06,
+      "loss": 0.9878,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930941087409084e-06,
+      "loss": 0.9961,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93082649410545e-06,
+      "loss": 0.9551,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930711806467335e-06,
+      "loss": 0.9834,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930597024496933e-06,
+      "loss": 1.04,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930482148196439e-06,
+      "loss": 0.3424,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930367177568052e-06,
+      "loss": 1.0225,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930252112613975e-06,
+      "loss": 0.9658,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930136953336402e-06,
+      "loss": 0.9741,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930021699737542e-06,
+      "loss": 0.9448,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929906351819598e-06,
+      "loss": 0.9561,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929790909584777e-06,
+      "loss": 0.9507,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929675373035288e-06,
+      "loss": 0.978,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92955974217334e-06,
+      "loss": 0.9258,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929444017001146e-06,
+      "loss": 0.981,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92932819752092e-06,
+      "loss": 0.98,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92921228373488e-06,
+      "loss": 1.0337,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92909627564524e-06,
+      "loss": 0.9277,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928980173254221e-06,
+      "loss": 0.9531,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928863976564042e-06,
+      "loss": 1.0264,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928747685576931e-06,
+      "loss": 1.0229,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928631300295111e-06,
+      "loss": 1.001,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928514820720805e-06,
+      "loss": 0.9385,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928398246856246e-06,
+      "loss": 0.9541,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928281578703661e-06,
+      "loss": 1.0254,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928164816265283e-06,
+      "loss": 0.9429,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928047959543348e-06,
+      "loss": 0.9819,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927931008540088e-06,
+      "loss": 0.9292,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927813963257744e-06,
+      "loss": 0.9702,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927696823698552e-06,
+      "loss": 0.9785,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927579589864756e-06,
+      "loss": 0.9561,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927462261758597e-06,
+      "loss": 1.0215,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927344839382321e-06,
+      "loss": 0.9077,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927227322738173e-06,
+      "loss": 0.9414,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927109711828403e-06,
+      "loss": 0.8643,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926992006655258e-06,
+      "loss": 0.9131,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926874207220993e-06,
+      "loss": 1.0186,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92675631352786e-06,
+      "loss": 0.937,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926638325578118e-06,
+      "loss": 1.0063,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926520243374018e-06,
+      "loss": 0.9575,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926402066917825e-06,
+      "loss": 0.9575,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926283796211796e-06,
+      "loss": 0.9795,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926165431258195e-06,
+      "loss": 0.9604,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926046972059286e-06,
+      "loss": 0.9697,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925928418617338e-06,
+      "loss": 0.9517,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925809770934617e-06,
+      "loss": 1.0117,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925691029013392e-06,
+      "loss": 1.0059,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925572192855935e-06,
+      "loss": 0.9683,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925453262464519e-06,
+      "loss": 0.9663,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925334237841423e-06,
+      "loss": 0.9619,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92521511898892e-06,
+      "loss": 0.9492,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925095905909291e-06,
+      "loss": 0.9614,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924976598604818e-06,
+      "loss": 1.001,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924857197077781e-06,
+      "loss": 0.9814,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924737701330466e-06,
+      "loss": 0.9033,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924618111365157e-06,
+      "loss": 0.9634,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924498427184144e-06,
+      "loss": 0.9443,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924378648789715e-06,
+      "loss": 0.9756,
+      "step": 1962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924258776184163e-06,
+      "loss": 0.9761,
+      "step": 1963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924138809369782e-06,
+      "loss": 0.9355,
+      "step": 1964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924018748348865e-06,
+      "loss": 0.9946,
+      "step": 1965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923898593123711e-06,
+      "loss": 0.9536,
+      "step": 1966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923778343696616e-06,
+      "loss": 0.9409,
+      "step": 1967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923658000069885e-06,
+      "loss": 0.9185,
+      "step": 1968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923537562245815e-06,
+      "loss": 0.9341,
+      "step": 1969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923417030226715e-06,
+      "loss": 0.9717,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923296404014888e-06,
+      "loss": 0.9565,
+      "step": 1971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923175683612643e-06,
+      "loss": 0.9844,
+      "step": 1972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923054869022288e-06,
+      "loss": 0.9731,
+      "step": 1973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922933960246137e-06,
+      "loss": 0.9878,
+      "step": 1974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922812957286502e-06,
+      "loss": 0.9243,
+      "step": 1975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922691860145696e-06,
+      "loss": 1.0015,
+      "step": 1976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92257066882604e-06,
+      "loss": 0.9922,
+      "step": 1977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922449383329848e-06,
+      "loss": 0.9697,
+      "step": 1978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922328003659443e-06,
+      "loss": 0.9658,
+      "step": 1979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922206529817146e-06,
+      "loss": 0.9629,
+      "step": 1980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922084961805282e-06,
+      "loss": 1.02,
+      "step": 1981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921963299626178e-06,
+      "loss": 0.9854,
+      "step": 1982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92184154328216e-06,
+      "loss": 0.9028,
+      "step": 1983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921719692775556e-06,
+      "loss": 0.9927,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921597748108698e-06,
+      "loss": 0.9355,
+      "step": 1985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921475709283921e-06,
+      "loss": 1.0195,
+      "step": 1986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921353576303558e-06,
+      "loss": 0.3987,
+      "step": 1987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921231349169946e-06,
+      "loss": 0.9263,
+      "step": 1988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921109027885423e-06,
+      "loss": 0.9546,
+      "step": 1989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92098661245233e-06,
+      "loss": 0.9116,
+      "step": 1990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920864102873009e-06,
+      "loss": 0.9639,
+      "step": 1991
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920741499149802e-06,
+      "loss": 0.9946,
+      "step": 1992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920618801285057e-06,
+      "loss": 0.9893,
+      "step": 1993
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920496009281121e-06,
+      "loss": 1.0083,
+      "step": 1994
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920373123140342e-06,
+      "loss": 0.9795,
+      "step": 1995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920250142865072e-06,
+      "loss": 1.0063,
+      "step": 1996
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920127068457663e-06,
+      "loss": 0.9536,
+      "step": 1997
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92000389992047e-06,
+      "loss": 0.9819,
+      "step": 1998
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91988063725585e-06,
+      "loss": 0.9707,
+      "step": 1999
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91975728046616e-06,
+      "loss": 0.9829,
+      "step": 2000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91963382955376e-06,
+      "loss": 0.9468,
+      "step": 2001
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919510284521013e-06,
+      "loss": 0.9272,
+      "step": 2002
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919386645370283e-06,
+      "loss": 0.8682,
+      "step": 2003
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919262912103934e-06,
+      "loss": 0.9624,
+      "step": 2004
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919139084724331e-06,
+      "loss": 0.9409,
+      "step": 2005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919015163233849e-06,
+      "loss": 1.0166,
+      "step": 2006
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918891147634856e-06,
+      "loss": 0.9185,
+      "step": 2007
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918767037929722e-06,
+      "loss": 0.9907,
+      "step": 2008
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918642834120825e-06,
+      "loss": 1.0103,
+      "step": 2009
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918518536210539e-06,
+      "loss": 0.9971,
+      "step": 2010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918394144201242e-06,
+      "loss": 0.8647,
+      "step": 2011
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918269658095316e-06,
+      "loss": 0.9863,
+      "step": 2012
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918145077895141e-06,
+      "loss": 0.9761,
+      "step": 2013
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918020403603099e-06,
+      "loss": 0.939,
+      "step": 2014
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91789563522158e-06,
+      "loss": 1.0073,
+      "step": 2015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917770772752965e-06,
+      "loss": 0.915,
+      "step": 2016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917645816199646e-06,
+      "loss": 0.9478,
+      "step": 2017
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917520765564015e-06,
+      "loss": 0.9639,
+      "step": 2018
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917395620848463e-06,
+      "loss": 0.9331,
+      "step": 2019
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917270382055382e-06,
+      "loss": 0.9297,
+      "step": 2020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91714504918717e-06,
+      "loss": 1.0137,
+      "step": 2021
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917019622246224e-06,
+      "loss": 0.9443,
+      "step": 2022
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916894101234947e-06,
+      "loss": 0.9893,
+      "step": 2023
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916768486155736e-06,
+      "loss": 0.9722,
+      "step": 2024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916642777010996e-06,
+      "loss": 0.918,
+      "step": 2025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916516973803133e-06,
+      "loss": 0.9116,
+      "step": 2026
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916391076534552e-06,
+      "loss": 0.9644,
+      "step": 2027
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916265085207663e-06,
+      "loss": 0.9434,
+      "step": 2028
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916138999824876e-06,
+      "loss": 0.9258,
+      "step": 2029
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916012820388603e-06,
+      "loss": 1.0298,
+      "step": 2030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915886546901257e-06,
+      "loss": 0.9321,
+      "step": 2031
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915760179365256e-06,
+      "loss": 1.0454,
+      "step": 2032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915633717783015e-06,
+      "loss": 0.9609,
+      "step": 2033
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915507162156956e-06,
+      "loss": 1.0205,
+      "step": 2034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915380512489499e-06,
+      "loss": 0.9307,
+      "step": 2035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915253768783069e-06,
+      "loss": 0.9868,
+      "step": 2036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915126931040086e-06,
+      "loss": 0.9834,
+      "step": 2037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91499999926298e-06,
+      "loss": 0.9453,
+      "step": 2038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91487297345418e-06,
+      "loss": 0.9375,
+      "step": 2039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914745853616114e-06,
+      "loss": 0.9736,
+      "step": 2040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914618639751216e-06,
+      "loss": 0.9717,
+      "step": 2041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91449133186192e-06,
+      "loss": 0.9824,
+      "step": 2042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91436392995066e-06,
+      "loss": 0.9712,
+      "step": 2043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914236434019873e-06,
+      "loss": 0.3856,
+      "step": 2044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914108844071999e-06,
+      "loss": 0.9785,
+      "step": 2045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913981160109481e-06,
+      "loss": 0.957,
+      "step": 2046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913853382134757e-06,
+      "loss": 0.9668,
+      "step": 2047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913725510150276e-06,
+      "loss": 0.9473,
+      "step": 2048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913597544158485e-06,
+      "loss": 1.0049,
+      "step": 2049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913469484161826e-06,
+      "loss": 0.9526,
+      "step": 2050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913341330162756e-06,
+      "loss": 0.9771,
+      "step": 2051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913213082163723e-06,
+      "loss": 0.9546,
+      "step": 2052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91308474016718e-06,
+      "loss": 0.9463,
+      "step": 2053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912956304175587e-06,
+      "loss": 0.9912,
+      "step": 2054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912827774191395e-06,
+      "loss": 0.9336,
+      "step": 2055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91269915021707e-06,
+      "loss": 0.9365,
+      "step": 2056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912570432255065e-06,
+      "loss": 0.978,
+      "step": 2057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912441620307849e-06,
+      "loss": 0.9849,
+      "step": 2058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91231271437788e-06,
+      "loss": 0.8989,
+      "step": 2059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912183714467632e-06,
+      "loss": 0.9438,
+      "step": 2060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912054620579568e-06,
+      "loss": 0.9648,
+      "step": 2061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911925432716157e-06,
+      "loss": 0.9595,
+      "step": 2062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911796150879873e-06,
+      "loss": 1.0073,
+      "step": 2063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91166677507319e-06,
+      "loss": 0.9634,
+      "step": 2064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91153730529858e-06,
+      "loss": 0.9858,
+      "step": 2065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911407741558523e-06,
+      "loss": 0.9629,
+      "step": 2066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911278083855495e-06,
+      "loss": 0.9585,
+      "step": 2067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911148332191979e-06,
+      "loss": 0.9644,
+      "step": 2068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911018486570456e-06,
+      "loss": 0.915,
+      "step": 2069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910888546993412e-06,
+      "loss": 0.9717,
+      "step": 2070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91075851346333e-06,
+      "loss": 0.9966,
+      "step": 2071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910628385982698e-06,
+      "loss": 0.9155,
+      "step": 2072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91049816455401e-06,
+      "loss": 0.9473,
+      "step": 2073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910367849179751e-06,
+      "loss": 0.9858,
+      "step": 2074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910237439862419e-06,
+      "loss": 1.002,
+      "step": 2075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910106936604506e-06,
+      "loss": 0.9375,
+      "step": 2076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909976339408512e-06,
+      "loss": 0.9907,
+      "step": 2077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909845648276931e-06,
+      "loss": 0.9883,
+      "step": 2078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909714863212269e-06,
+      "loss": 0.9648,
+      "step": 2079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909583984217022e-06,
+      "loss": 0.2609,
+      "step": 2080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909453011293699e-06,
+      "loss": 1.0151,
+      "step": 2081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909321944444802e-06,
+      "loss": 0.9429,
+      "step": 2082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90919078367284e-06,
+      "loss": 0.9507,
+      "step": 2083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909059528980323e-06,
+      "loss": 0.9536,
+      "step": 2084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908928180369761e-06,
+      "loss": 0.9619,
+      "step": 2085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90879673784367e-06,
+      "loss": 0.9299,
+      "step": 2086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90866520140456e-06,
+      "loss": 0.9346,
+      "step": 2087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908533571054948e-06,
+      "loss": 0.9287,
+      "step": 2088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908401846797354e-06,
+      "loss": 0.9751,
+      "step": 2089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908270028634302e-06,
+      "loss": 1.0278,
+      "step": 2090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908138116568306e-06,
+      "loss": 0.9912,
+      "step": 2091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908006110601895e-06,
+      "loss": 0.9517,
+      "step": 2092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907874010737592e-06,
+      "loss": 0.9644,
+      "step": 2093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907741816977927e-06,
+      "loss": 0.9722,
+      "step": 2094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907609529325427e-06,
+      "loss": 1.0044,
+      "step": 2095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907477147782623e-06,
+      "loss": 0.9565,
+      "step": 2096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907344672352047e-06,
+      "loss": 0.9141,
+      "step": 2097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907212103036235e-06,
+      "loss": 0.9692,
+      "step": 2098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907079439837724e-06,
+      "loss": 0.9756,
+      "step": 2099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90694668275905e-06,
+      "loss": 0.9468,
+      "step": 2100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906813831802753e-06,
+      "loss": 1.0142,
+      "step": 2101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906680886971376e-06,
+      "loss": 0.978,
+      "step": 2102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90654784826746e-06,
+      "loss": 0.9585,
+      "step": 2103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906414715693554e-06,
+      "loss": 0.9653,
+      "step": 2104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9062814892522e-06,
+      "loss": 0.9265,
+      "step": 2105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906148168945952e-06,
+      "loss": 0.9878,
+      "step": 2106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906014754777357e-06,
+      "loss": 0.9146,
+      "step": 2107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905881246748972e-06,
+      "loss": 0.9146,
+      "step": 2108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905747644863346e-06,
+      "loss": 1.0083,
+      "step": 2109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905613949123036e-06,
+      "loss": 0.9561,
+      "step": 2110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905480159530603e-06,
+      "loss": 1.0063,
+      "step": 2111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905346276088603e-06,
+      "loss": 0.939,
+      "step": 2112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905212298799597e-06,
+      "loss": 0.9961,
+      "step": 2113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905078227666153e-06,
+      "loss": 1.0049,
+      "step": 2114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904944062690834e-06,
+      "loss": 0.9531,
+      "step": 2115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904809803876204e-06,
+      "loss": 0.9702,
+      "step": 2116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904675451224834e-06,
+      "loss": 0.999,
+      "step": 2117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904541004739292e-06,
+      "loss": 0.9922,
+      "step": 2118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904406464422154e-06,
+      "loss": 0.98,
+      "step": 2119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904271830275991e-06,
+      "loss": 0.9429,
+      "step": 2120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90413710230338e-06,
+      "loss": 0.9458,
+      "step": 2121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9040022805069e-06,
+      "loss": 0.9399,
+      "step": 2122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903867364889125e-06,
+      "loss": 0.9702,
+      "step": 2123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903732355452642e-06,
+      "loss": 0.8945,
+      "step": 2124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903597252200033e-06,
+      "loss": 0.9634,
+      "step": 2125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90346205513388e-06,
+      "loss": 0.9409,
+      "step": 2126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903326764256771e-06,
+      "loss": 0.9517,
+      "step": 2127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903191379571294e-06,
+      "loss": 0.9551,
+      "step": 2128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903055901080038e-06,
+      "loss": 0.9741,
+      "step": 2129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902920328785599e-06,
+      "loss": 0.9824,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902784662690567e-06,
+      "loss": 0.959,
+      "step": 2131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90264890279754e-06,
+      "loss": 0.9307,
+      "step": 2132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902513049109111e-06,
+      "loss": 0.9688,
+      "step": 2133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902377101627884e-06,
+      "loss": 0.9683,
+      "step": 2134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902241060356458e-06,
+      "loss": 0.9487,
+      "step": 2135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902104925297435e-06,
+      "loss": 0.3428,
+      "step": 2136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901968696453419e-06,
+      "loss": 0.9487,
+      "step": 2137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901832373827019e-06,
+      "loss": 0.939,
+      "step": 2138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90169595742084e-06,
+      "loss": 0.9131,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901559447237494e-06,
+      "loss": 0.9048,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901422843279592e-06,
+      "loss": 0.9224,
+      "step": 2141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901286145549747e-06,
+      "loss": 0.3169,
+      "step": 2142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901149354050575e-06,
+      "loss": 0.9292,
+      "step": 2143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901012468784693e-06,
+      "loss": 0.9375,
+      "step": 2144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900875489754718e-06,
+      "loss": 0.9873,
+      "step": 2145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900738416963273e-06,
+      "loss": 0.9297,
+      "step": 2146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900601250412977e-06,
+      "loss": 1.0049,
+      "step": 2147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90046399010646e-06,
+      "loss": 1.0215,
+      "step": 2148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900326636046344e-06,
+      "loss": 0.9775,
+      "step": 2149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900189188235256e-06,
+      "loss": 0.9282,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90005164667583e-06,
+      "loss": 0.9834,
+      "step": 2151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899914011370692e-06,
+      "loss": 0.9312,
+      "step": 2152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899776282322479e-06,
+      "loss": 0.9561,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899638459533824e-06,
+      "loss": 0.959,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899500543007363e-06,
+      "loss": 1.0181,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899362532745738e-06,
+      "loss": 0.9951,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899224428751585e-06,
+      "loss": 0.9863,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899086231027552e-06,
+      "loss": 0.8711,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898947939576279e-06,
+      "loss": 0.9858,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898809554400408e-06,
+      "loss": 0.9683,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898671075502594e-06,
+      "loss": 0.9736,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898532502885483e-06,
+      "loss": 0.9116,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898393836551725e-06,
+      "loss": 0.9814,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898255076503973e-06,
+      "loss": 0.9399,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898116222744884e-06,
+      "loss": 1.0034,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897977275277113e-06,
+      "loss": 0.9219,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897838234103321e-06,
+      "loss": 0.9834,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897699099226162e-06,
+      "loss": 0.8965,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897559870648303e-06,
+      "loss": 0.9341,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897420548372406e-06,
+      "loss": 0.9399,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897281132401135e-06,
+      "loss": 0.9946,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89714162273716e-06,
+      "loss": 1.0073,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89700201938315e-06,
+      "loss": 0.9668,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896862322341774e-06,
+      "loss": 0.9673,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896722531615704e-06,
+      "loss": 0.9556,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896582647207618e-06,
+      "loss": 0.9917,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896442669120188e-06,
+      "loss": 0.9487,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896302597356094e-06,
+      "loss": 0.9556,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896162431918015e-06,
+      "loss": 0.9917,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896022172808635e-06,
+      "loss": 0.9893,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895881820030636e-06,
+      "loss": 0.9429,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8957413735867e-06,
+      "loss": 0.9165,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89560083347952e-06,
+      "loss": 0.9619,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89546019971178e-06,
+      "loss": 0.9424,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895319472286171e-06,
+      "loss": 0.9536,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895178651205389e-06,
+      "loss": 0.8989,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895037736472124e-06,
+      "loss": 0.9185,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894896728089074e-06,
+      "loss": 0.915,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894755626058935e-06,
+      "loss": 0.9448,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894614430384408e-06,
+      "loss": 1.0054,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894473141068195e-06,
+      "loss": 0.9541,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894331758112998e-06,
+      "loss": 0.9336,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894190281521522e-06,
+      "loss": 0.9536,
+      "step": 2193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894048711296473e-06,
+      "loss": 1.0098,
+      "step": 2194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893907047440561e-06,
+      "loss": 0.8263,
+      "step": 2195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893765289956494e-06,
+      "loss": 0.9111,
+      "step": 2196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893623438846986e-06,
+      "loss": 0.9375,
+      "step": 2197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893481494114751e-06,
+      "loss": 0.959,
+      "step": 2198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893339455762504e-06,
+      "loss": 0.9478,
+      "step": 2199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89319732379296e-06,
+      "loss": 0.9922,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89305509820884e-06,
+      "loss": 0.9233,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892912779012869e-06,
+      "loss": 0.9077,
+      "step": 2202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892770366207765e-06,
+      "loss": 0.9717,
+      "step": 2203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892627859796253e-06,
+      "loss": 0.9473,
+      "step": 2204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892485259781062e-06,
+      "loss": 0.9688,
+      "step": 2205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892342566164916e-06,
+      "loss": 0.9292,
+      "step": 2206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892199778950549e-06,
+      "loss": 0.978,
+      "step": 2207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89205689814069e-06,
+      "loss": 0.9585,
+      "step": 2208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891913923738074e-06,
+      "loss": 0.9961,
+      "step": 2209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891770855745437e-06,
+      "loss": 0.9888,
+      "step": 2210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891627694165514e-06,
+      "loss": 0.9487,
+      "step": 2211
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891484439001045e-06,
+      "loss": 0.9946,
+      "step": 2212
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89134109025477e-06,
+      "loss": 0.9751,
+      "step": 2213
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891197647929433e-06,
+      "loss": 1.0005,
+      "step": 2214
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891054112027778e-06,
+      "loss": 0.98,
+      "step": 2215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890910482552549e-06,
+      "loss": 1.001,
+      "step": 2216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890766759506496e-06,
+      "loss": 0.9814,
+      "step": 2217
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890622942892369e-06,
+      "loss": 0.9023,
+      "step": 2218
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890479032712918e-06,
+      "loss": 0.9175,
+      "step": 2219
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890335028970895e-06,
+      "loss": 0.9854,
+      "step": 2220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890190931669059e-06,
+      "loss": 0.9692,
+      "step": 2221
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890046740810163e-06,
+      "loss": 0.958,
+      "step": 2222
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889902456396967e-06,
+      "loss": 0.9102,
+      "step": 2223
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889758078432232e-06,
+      "loss": 0.9766,
+      "step": 2224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88961360691872e-06,
+      "loss": 0.8975,
+      "step": 2225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889469041859194e-06,
+      "loss": 0.9888,
+      "step": 2226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889324383256421e-06,
+      "loss": 0.9751,
+      "step": 2227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889179631113167e-06,
+      "loss": 0.9243,
+      "step": 2228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889034785432203e-06,
+      "loss": 0.8906,
+      "step": 2229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8888898462163e-06,
+      "loss": 0.958,
+      "step": 2230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88874481346823e-06,
+      "loss": 0.9907,
+      "step": 2231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888599687190767e-06,
+      "loss": 1.0098,
+      "step": 2232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88845446738669e-06,
+      "loss": 0.8853,
+      "step": 2233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888309154058775e-06,
+      "loss": 0.9009,
+      "step": 2234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888163747209803e-06,
+      "loss": 0.9741,
+      "step": 2235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888018246842555e-06,
+      "loss": 0.9697,
+      "step": 2236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887872652959816e-06,
+      "loss": 0.957,
+      "step": 2237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887726965564371e-06,
+      "loss": 0.9751,
+      "step": 2238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887581184659005e-06,
+      "loss": 1.021,
+      "step": 2239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887435310246512e-06,
+      "loss": 0.9624,
+      "step": 2240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887289342329678e-06,
+      "loss": 0.3147,
+      "step": 2241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887143280911299e-06,
+      "loss": 0.9819,
+      "step": 2242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886997125994166e-06,
+      "loss": 0.9644,
+      "step": 2243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886850877581079e-06,
+      "loss": 0.9873,
+      "step": 2244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886704535674834e-06,
+      "loss": 0.9106,
+      "step": 2245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886558100278228e-06,
+      "loss": 0.9653,
+      "step": 2246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886411571394068e-06,
+      "loss": 0.938,
+      "step": 2247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886264949025152e-06,
+      "loss": 0.9058,
+      "step": 2248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886118233174292e-06,
+      "loss": 0.9697,
+      "step": 2249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885971423844288e-06,
+      "loss": 0.9102,
+      "step": 2250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88582452103795e-06,
+      "loss": 0.9482,
+      "step": 2251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885677524758091e-06,
+      "loss": 1.0332,
+      "step": 2252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885530435007522e-06,
+      "loss": 0.9355,
+      "step": 2253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885383251789058e-06,
+      "loss": 0.9927,
+      "step": 2254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885235975105514e-06,
+      "loss": 0.9448,
+      "step": 2255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885088604959708e-06,
+      "loss": 0.9766,
+      "step": 2256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884941141354455e-06,
+      "loss": 0.9741,
+      "step": 2257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884793584292585e-06,
+      "loss": 0.9067,
+      "step": 2258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884645933776915e-06,
+      "loss": 0.9082,
+      "step": 2259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88449818981027e-06,
+      "loss": 1.001,
+      "step": 2260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884350352395477e-06,
+      "loss": 0.9429,
+      "step": 2261
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884202421535367e-06,
+      "loss": 0.9243,
+      "step": 2262
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884054397232767e-06,
+      "loss": 0.9312,
+      "step": 2263
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883906279490512e-06,
+      "loss": 0.9424,
+      "step": 2264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88375806831143e-06,
+      "loss": 0.9854,
+      "step": 2265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883609763698364e-06,
+      "loss": 1.0234,
+      "step": 2266
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883461365654146e-06,
+      "loss": 0.9023,
+      "step": 2267
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883312874181617e-06,
+      "loss": 0.957,
+      "step": 2268
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883164289283617e-06,
+      "loss": 0.8594,
+      "step": 2269
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883015610962989e-06,
+      "loss": 1.0215,
+      "step": 2270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882866839222579e-06,
+      "loss": 0.9956,
+      "step": 2271
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88271797406523e-06,
+      "loss": 0.9844,
+      "step": 2272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882569015493794e-06,
+      "loss": 1.0029,
+      "step": 2273
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882419963511118e-06,
+      "loss": 0.9663,
+      "step": 2274
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882270818120055e-06,
+      "loss": 0.9458,
+      "step": 2275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882121579323457e-06,
+      "loss": 1.0127,
+      "step": 2276
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88197224712418e-06,
+      "loss": 0.9341,
+      "step": 2277
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881822821525081e-06,
+      "loss": 0.9766,
+      "step": 2278
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881673302529018e-06,
+      "loss": 0.9062,
+      "step": 2279
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881523690138852e-06,
+      "loss": 0.936,
+      "step": 2280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881373984357447e-06,
+      "loss": 0.9507,
+      "step": 2281
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881224185187666e-06,
+      "loss": 1.0166,
+      "step": 2282
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881074292632372e-06,
+      "loss": 0.9785,
+      "step": 2283
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880924306694438e-06,
+      "loss": 0.8588,
+      "step": 2284
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880774227376727e-06,
+      "loss": 0.8975,
+      "step": 2285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880624054682117e-06,
+      "loss": 0.9546,
+      "step": 2286
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880473788613475e-06,
+      "loss": 0.915,
+      "step": 2287
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880323429173682e-06,
+      "loss": 0.9336,
+      "step": 2288
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88017297636561e-06,
+      "loss": 0.916,
+      "step": 2289
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880022430192137e-06,
+      "loss": 0.9536,
+      "step": 2290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879871790656148e-06,
+      "loss": 0.9902,
+      "step": 2291
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87972105776052e-06,
+      "loss": 0.9058,
+      "step": 2292
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87957023150814e-06,
+      "loss": 0.9585,
+      "step": 2293
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879419311901891e-06,
+      "loss": 0.9297,
+      "step": 2294
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879268298944662e-06,
+      "loss": 0.8721,
+      "step": 2295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879117192639343e-06,
+      "loss": 0.8945,
+      "step": 2296
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87896599298882e-06,
+      "loss": 0.959,
+      "step": 2297
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878814699995994e-06,
+      "loss": 0.9336,
+      "step": 2298
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87866331366375e-06,
+      "loss": 0.9277,
+      "step": 2299
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878511833994993e-06,
+      "loss": 0.9507,
+      "step": 2300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878360260992614e-06,
+      "loss": 0.9453,
+      "step": 2301
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878208594659518e-06,
+      "loss": 0.9082,
+      "step": 2302
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878056834998604e-06,
+      "loss": 0.854,
+      "step": 2303
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877904982012775e-06,
+      "loss": 0.9717,
+      "step": 2304
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877753035704938e-06,
+      "loss": 0.9404,
+      "step": 2305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877600996077998e-06,
+      "loss": 0.939,
+      "step": 2306
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877448863134867e-06,
+      "loss": 0.9351,
+      "step": 2307
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87729663687845e-06,
+      "loss": 0.9976,
+      "step": 2308
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877144317311664e-06,
+      "loss": 0.9243,
+      "step": 2309
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876991904437423e-06,
+      "loss": 0.9102,
+      "step": 2310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87683939825864e-06,
+      "loss": 0.9399,
+      "step": 2311
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876686798778237e-06,
+      "loss": 0.9121,
+      "step": 2312
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876534105999128e-06,
+      "loss": 0.9395,
+      "step": 2313
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876381319924238e-06,
+      "loss": 0.959,
+      "step": 2314
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87622844055649e-06,
+      "loss": 0.9839,
+      "step": 2315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876075467898806e-06,
+      "loss": 0.9731,
+      "step": 2316
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875922401954117e-06,
+      "loss": 0.9756,
+      "step": 2317
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875769242725348e-06,
+      "loss": 0.9604,
+      "step": 2318
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87561599021543e-06,
+      "loss": 0.98,
+      "step": 2319
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875462644427294e-06,
+      "loss": 1.0146,
+      "step": 2320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875309205363876e-06,
+      "loss": 0.9771,
+      "step": 2321
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875155673028113e-06,
+      "loss": 0.937,
+      "step": 2322
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875002047422935e-06,
+      "loss": 0.8882,
+      "step": 2323
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874848328551287e-06,
+      "loss": 0.9536,
+      "step": 2324
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87469451641611e-06,
+      "loss": 0.8989,
+      "step": 2325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874540611020345e-06,
+      "loss": 0.8979,
+      "step": 2326
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874386612366936e-06,
+      "loss": 0.9478,
+      "step": 2327
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87423252045883e-06,
+      "loss": 0.9121,
+      "step": 2328
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874078335298977e-06,
+      "loss": 0.9736,
+      "step": 2329
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873924056890324e-06,
+      "loss": 0.9624,
+      "step": 2330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873769685235821e-06,
+      "loss": 0.9883,
+      "step": 2331
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873615220338426e-06,
+      "loss": 0.9375,
+      "step": 2332
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873460662201093e-06,
+      "loss": 0.9189,
+      "step": 2333
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873306010826776e-06,
+      "loss": 0.9487,
+      "step": 2334
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873151266218435e-06,
+      "loss": 0.959,
+      "step": 2335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872996428379033e-06,
+      "loss": 1.0107,
+      "step": 2336
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872841497311531e-06,
+      "loss": 0.8833,
+      "step": 2337
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872686473018891e-06,
+      "loss": 0.9419,
+      "step": 2338
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87253135550408e-06,
+      "loss": 0.9175,
+      "step": 2339
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872376144770067e-06,
+      "loss": 0.9419,
+      "step": 2340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87222084081982e-06,
+      "loss": 0.9814,
+      "step": 2341
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87206544365631e-06,
+      "loss": 0.9204,
+      "step": 2342
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871909953282513e-06,
+      "loss": 0.9531,
+      "step": 2343
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8717543697014e-06,
+      "loss": 1.0034,
+      "step": 2344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871598692915948e-06,
+      "loss": 0.9844,
+      "step": 2345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871442922929137e-06,
+      "loss": 0.9902,
+      "step": 2346
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871287059743947e-06,
+      "loss": 0.9893,
+      "step": 2347
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871131103363359e-06,
+      "loss": 0.9653,
+      "step": 2348
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870975053790357e-06,
+      "loss": 0.9756,
+      "step": 2349
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870818911027926e-06,
+      "loss": 0.915,
+      "step": 2350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870662675079055e-06,
+      "loss": 0.9502,
+      "step": 2351
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87050634594673e-06,
+      "loss": 0.9316,
+      "step": 2352
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870349923633947e-06,
+      "loss": 0.9316,
+      "step": 2353
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870193408143691e-06,
+      "loss": 0.9766,
+      "step": 2354
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870036799478963e-06,
+      "loss": 0.9526,
+      "step": 2355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869880097642757e-06,
+      "loss": 0.9507,
+      "step": 2356
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86972330263807e-06,
+      "loss": 0.9858,
+      "step": 2357
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869566414467902e-06,
+      "loss": 0.9233,
+      "step": 2358
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869409433135255e-06,
+      "loss": 0.9741,
+      "step": 2359
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869252358643133e-06,
+      "loss": 1.0337,
+      "step": 2360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869095190994542e-06,
+      "loss": 0.9146,
+      "step": 2361
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868937930192485e-06,
+      "loss": 0.9058,
+      "step": 2362
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868780576239973e-06,
+      "loss": 0.9492,
+      "step": 2363
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868623129140017e-06,
+      "loss": 0.9546,
+      "step": 2364
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868465588895628e-06,
+      "loss": 0.9521,
+      "step": 2365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868307955509822e-06,
+      "loss": 0.9272,
+      "step": 2366
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868150228985613e-06,
+      "loss": 0.9819,
+      "step": 2367
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867992409326016e-06,
+      "loss": 1.0063,
+      "step": 2368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867834496534056e-06,
+      "loss": 1.0283,
+      "step": 2369
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867676490612752e-06,
+      "loss": 0.9141,
+      "step": 2370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867518391565125e-06,
+      "loss": 0.9263,
+      "step": 2371
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8673601993942e-06,
+      "loss": 0.9199,
+      "step": 2372
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867201914103005e-06,
+      "loss": 0.9302,
+      "step": 2373
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867043535694568e-06,
+      "loss": 0.9209,
+      "step": 2374
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866885064171918e-06,
+      "loss": 0.9663,
+      "step": 2375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866726499538088e-06,
+      "loss": 0.9272,
+      "step": 2376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866567841796111e-06,
+      "loss": 0.9702,
+      "step": 2377
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866409090949023e-06,
+      "loss": 0.979,
+      "step": 2378
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86625024699986e-06,
+      "loss": 0.9478,
+      "step": 2379
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866091309951662e-06,
+      "loss": 0.9971,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865932279807469e-06,
+      "loss": 0.9165,
+      "step": 2381
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865773156570323e-06,
+      "loss": 0.9951,
+      "step": 2382
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865613940243268e-06,
+      "loss": 0.9771,
+      "step": 2383
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865454630829354e-06,
+      "loss": 0.9653,
+      "step": 2384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865295228331624e-06,
+      "loss": 0.9321,
+      "step": 2385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865135732753131e-06,
+      "loss": 0.9336,
+      "step": 2386
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864976144096922e-06,
+      "loss": 0.9541,
+      "step": 2387
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864816462366056e-06,
+      "loss": 0.9629,
+      "step": 2388
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864656687563584e-06,
+      "loss": 0.9751,
+      "step": 2389
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864496819692564e-06,
+      "loss": 0.937,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864336858756054e-06,
+      "loss": 0.9502,
+      "step": 2391
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864176804757115e-06,
+      "loss": 0.957,
+      "step": 2392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86401665769881e-06,
+      "loss": 0.9814,
+      "step": 2393
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863856417584199e-06,
+      "loss": 0.9302,
+      "step": 2394
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863696084416351e-06,
+      "loss": 0.9678,
+      "step": 2395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863535658198333e-06,
+      "loss": 0.9624,
+      "step": 2396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863375138933215e-06,
+      "loss": 0.9565,
+      "step": 2397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863214526624065e-06,
+      "loss": 0.9692,
+      "step": 2398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86305382127396e-06,
+      "loss": 0.9341,
+      "step": 2399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86289302288597e-06,
+      "loss": 0.9609,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862732131463175e-06,
+      "loss": 0.9805,
+      "step": 2401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86257114700865e-06,
+      "loss": 0.9946,
+      "step": 2402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862410069525479e-06,
+      "loss": 0.9478,
+      "step": 2403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862248899016738e-06,
+      "loss": 0.8696,
+      "step": 2404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862087635485516e-06,
+      "loss": 0.9541,
+      "step": 2405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861926278934894e-06,
+      "loss": 0.9282,
+      "step": 2406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861764829367963e-06,
+      "loss": 1.0093,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86160328678781e-06,
+      "loss": 0.9053,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861441651197521e-06,
+      "loss": 0.9702,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861279922600195e-06,
+      "loss": 0.9746,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861118100998925e-06,
+      "loss": 1.002,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860956186396804e-06,
+      "loss": 0.9712,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860794178796931e-06,
+      "loss": 0.5087,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860632078202406e-06,
+      "loss": 0.9185,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860469884616331e-06,
+      "loss": 0.938,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860307598041807e-06,
+      "loss": 0.9194,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86014521848194e-06,
+      "loss": 0.9346,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859982745939837e-06,
+      "loss": 0.9785,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859820180418604e-06,
+      "loss": 0.9673,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859657521921355e-06,
+      "loss": 0.3691,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859494770451197e-06,
+      "loss": 0.9688,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859331926011248e-06,
+      "loss": 1.0127,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859168988604624e-06,
+      "loss": 0.9727,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859005958234438e-06,
+      "loss": 0.9233,
+      "step": 2424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858842834903811e-06,
+      "loss": 0.9458,
+      "step": 2425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858679618615863e-06,
+      "loss": 0.896,
+      "step": 2426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85851630937372e-06,
+      "loss": 0.9951,
+      "step": 2427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858352907180502e-06,
+      "loss": 0.9683,
+      "step": 2428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858189412039338e-06,
+      "loss": 0.9873,
+      "step": 2429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858025823953354e-06,
+      "loss": 0.9453,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857862142925682e-06,
+      "loss": 0.9824,
+      "step": 2431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857698368959453e-06,
+      "loss": 0.938,
+      "step": 2432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857534502057797e-06,
+      "loss": 0.96,
+      "step": 2433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857370542223851e-06,
+      "loss": 0.9863,
+      "step": 2434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857206489460756e-06,
+      "loss": 0.957,
+      "step": 2435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857042343771643e-06,
+      "loss": 0.9565,
+      "step": 2436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856878105159659e-06,
+      "loss": 0.9893,
+      "step": 2437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856713773627941e-06,
+      "loss": 0.9297,
+      "step": 2438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856549349179638e-06,
+      "loss": 0.9453,
+      "step": 2439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856384831817892e-06,
+      "loss": 1.0005,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856220221545849e-06,
+      "loss": 0.9766,
+      "step": 2441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856055518366663e-06,
+      "loss": 0.9268,
+      "step": 2442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855890722283482e-06,
+      "loss": 0.9805,
+      "step": 2443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855725833299461e-06,
+      "loss": 0.895,
+      "step": 2444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855560851417752e-06,
+      "loss": 0.9536,
+      "step": 2445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855395776641513e-06,
+      "loss": 1.022,
+      "step": 2446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855230608973901e-06,
+      "loss": 0.9849,
+      "step": 2447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855065348418076e-06,
+      "loss": 0.9619,
+      "step": 2448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8548999949772e-06,
+      "loss": 0.9326,
+      "step": 2449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854734548654438e-06,
+      "loss": 0.9263,
+      "step": 2450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854569009452954e-06,
+      "loss": 0.8564,
+      "step": 2451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854403377375916e-06,
+      "loss": 0.9668,
+      "step": 2452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85423765242649e-06,
+      "loss": 0.3099,
+      "step": 2453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854071834607848e-06,
+      "loss": 0.874,
+      "step": 2454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853905923923164e-06,
+      "loss": 0.9844,
+      "step": 2455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853739920375611e-06,
+      "loss": 0.9751,
+      "step": 2456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853573823968363e-06,
+      "loss": 0.8452,
+      "step": 2457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853407634704602e-06,
+      "loss": 0.9961,
+      "step": 2458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853241352587503e-06,
+      "loss": 0.9575,
+      "step": 2459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85307497762025e-06,
+      "loss": 0.9087,
+      "step": 2460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852908509806024e-06,
+      "loss": 0.8755,
+      "step": 2461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852741949148012e-06,
+      "loss": 0.9268,
+      "step": 2462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852575295649401e-06,
+      "loss": 0.9512,
+      "step": 2463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852408549313376e-06,
+      "loss": 0.9629,
+      "step": 2464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852241710143128e-06,
+      "loss": 0.9507,
+      "step": 2465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852074778141854e-06,
+      "loss": 0.9316,
+      "step": 2466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85190775331274e-06,
+      "loss": 0.958,
+      "step": 2467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851740635658986e-06,
+      "loss": 0.9575,
+      "step": 2468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851573425183789e-06,
+      "loss": 0.9624,
+      "step": 2469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851406121890345e-06,
+      "loss": 0.9893,
+      "step": 2470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851238725781858e-06,
+      "loss": 0.8728,
+      "step": 2471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85107123686153e-06,
+      "loss": 1.0098,
+      "step": 2472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850903655132567e-06,
+      "loss": 0.9814,
+      "step": 2473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850735980598169e-06,
+      "loss": 0.9463,
+      "step": 2474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85056821326155e-06,
+      "loss": 0.9551,
+      "step": 2475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850400353125918e-06,
+      "loss": 0.9277,
+      "step": 2476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850232400194484e-06,
+      "loss": 0.8779,
+      "step": 2477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85006435447046e-06,
+      "loss": 0.3264,
+      "step": 2478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849896215957062e-06,
+      "loss": 0.9297,
+      "step": 2479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849727984657507e-06,
+      "loss": 0.9917,
+      "step": 2480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849559660575015e-06,
+      "loss": 0.938,
+      "step": 2481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849391243712803e-06,
+      "loss": 0.9463,
+      "step": 2482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849222734074096e-06,
+      "loss": 0.9062,
+      "step": 2483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849054131662116e-06,
+      "loss": 0.9478,
+      "step": 2484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84888543648009e-06,
+      "loss": 0.9209,
+      "step": 2485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848716648531245e-06,
+      "loss": 0.9697,
+      "step": 2486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848547767818808e-06,
+      "loss": 0.9766,
+      "step": 2487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848378794346012e-06,
+      "loss": 0.9219,
+      "step": 2488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848209728116092e-06,
+      "loss": 0.8867,
+      "step": 2489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848040569132278e-06,
+      "loss": 0.9707,
+      "step": 2490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84787131739781e-06,
+      "loss": 0.9512,
+      "step": 2491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847701972915921e-06,
+      "loss": 0.9233,
+      "step": 2492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847532535689858e-06,
+      "loss": 0.9458,
+      "step": 2493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847363005722857e-06,
+      "loss": 0.9741,
+      "step": 2494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847193383018164e-06,
+      "loss": 0.9502,
+      "step": 2495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847023667579023e-06,
+      "loss": 1.0142,
+      "step": 2496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84685385940868e-06,
+      "loss": 0.9443,
+      "step": 2497
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846683958510386e-06,
+      "loss": 0.9595,
+      "step": 2498
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846513964887392e-06,
+      "loss": 0.981,
+      "step": 2499
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846343878542947e-06,
+      "loss": 0.9692,
+      "step": 2500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846173699480308e-06,
+      "loss": 0.9565,
+      "step": 2501
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846003427702729e-06,
+      "loss": 0.9316,
+      "step": 2502
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845833063213466e-06,
+      "loss": 0.9897,
+      "step": 2503
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845662606015782e-06,
+      "loss": 0.9463,
+      "step": 2504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845492056112938e-06,
+      "loss": 0.9707,
+      "step": 2505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845321413508193e-06,
+      "loss": 0.9722,
+      "step": 2506
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845150678204815e-06,
+      "loss": 0.917,
+      "step": 2507
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844979850206069e-06,
+      "loss": 0.9121,
+      "step": 2508
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844808929515225e-06,
+      "loss": 0.9297,
+      "step": 2509
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84463791613555e-06,
+      "loss": 0.9292,
+      "step": 2510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844466810070319e-06,
+      "loss": 0.9043,
+      "step": 2511
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844295611322804e-06,
+      "loss": 0.9502,
+      "step": 2512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84412431989628e-06,
+      "loss": 1.0044,
+      "step": 2513
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843952935794024e-06,
+      "loss": 0.9121,
+      "step": 2514
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843781459019314e-06,
+      "loss": 0.9561,
+      "step": 2515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843609889575434e-06,
+      "loss": 0.491,
+      "step": 2516
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843438227465665e-06,
+      "loss": 0.9595,
+      "step": 2517
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843266472693289e-06,
+      "loss": 0.937,
+      "step": 2518
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843094625261593e-06,
+      "loss": 0.8198,
+      "step": 2519
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842922685173864e-06,
+      "loss": 0.9741,
+      "step": 2520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842750652433396e-06,
+      "loss": 0.9927,
+      "step": 2521
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842578527043472e-06,
+      "loss": 0.9458,
+      "step": 2522
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842406309007393e-06,
+      "loss": 0.9961,
+      "step": 2523
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84223399832845e-06,
+      "loss": 0.9814,
+      "step": 2524
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842061595009939e-06,
+      "loss": 0.9385,
+      "step": 2525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841889099055161e-06,
+      "loss": 0.9429,
+      "step": 2526
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841716510467416e-06,
+      "loss": 0.9487,
+      "step": 2527
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841543829250002e-06,
+      "loss": 1.022,
+      "step": 2528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841371055406227e-06,
+      "loss": 0.9072,
+      "step": 2529
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841198188939394e-06,
+      "loss": 0.9443,
+      "step": 2530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84102522985281e-06,
+      "loss": 0.9419,
+      "step": 2531
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840852178149787e-06,
+      "loss": 0.9424,
+      "step": 2532
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840679033833631e-06,
+      "loss": 0.8975,
+      "step": 2533
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840505796907657e-06,
+      "loss": 0.9868,
+      "step": 2534
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840332467375183e-06,
+      "loss": 0.9434,
+      "step": 2535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840159045239519e-06,
+      "loss": 0.9556,
+      "step": 2536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839985530503985e-06,
+      "loss": 0.9058,
+      "step": 2537
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8398119231719e-06,
+      "loss": 0.9077,
+      "step": 2538
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839638223246589e-06,
+      "loss": 0.9556,
+      "step": 2539
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839464430731371e-06,
+      "loss": 0.3513,
+      "step": 2540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839290545629573e-06,
+      "loss": 1.0356,
+      "step": 2541
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839116567944521e-06,
+      "loss": 0.9644,
+      "step": 2542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838942497679543e-06,
+      "loss": 0.9365,
+      "step": 2543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83876833483797e-06,
+      "loss": 0.9766,
+      "step": 2544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838594079423133e-06,
+      "loss": 0.2918,
+      "step": 2545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838419731438369e-06,
+      "loss": 0.9165,
+      "step": 2546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83824529088701e-06,
+      "loss": 0.9233,
+      "step": 2547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838070757772393e-06,
+      "loss": 0.9932,
+      "step": 2548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83789613209786e-06,
+      "loss": 0.9473,
+      "step": 2549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83772141386675e-06,
+      "loss": 0.958,
+      "step": 2550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837546603082406e-06,
+      "loss": 0.9849,
+      "step": 2551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837371699748172e-06,
+      "loss": 0.9512,
+      "step": 2552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837196703867395e-06,
+      "loss": 0.9697,
+      "step": 2553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837021615443423e-06,
+      "loss": 0.9834,
+      "step": 2554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836846434479605e-06,
+      "loss": 0.9087,
+      "step": 2555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836671160979294e-06,
+      "loss": 0.9585,
+      "step": 2556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83649579494584e-06,
+      "loss": 0.9067,
+      "step": 2557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836320336382602e-06,
+      "loss": 0.9614,
+      "step": 2558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836144785292932e-06,
+      "loss": 0.9893,
+      "step": 2559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835969141680195e-06,
+      "loss": 0.9233,
+      "step": 2560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835793405547744e-06,
+      "loss": 0.9341,
+      "step": 2561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835617576898947e-06,
+      "loss": 0.9365,
+      "step": 2562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835441655737168e-06,
+      "loss": 0.9404,
+      "step": 2563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835265642065766e-06,
+      "loss": 0.9487,
+      "step": 2564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835089535888116e-06,
+      "loss": 0.9565,
+      "step": 2565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834913337207584e-06,
+      "loss": 0.8843,
+      "step": 2566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834737046027539e-06,
+      "loss": 0.8843,
+      "step": 2567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834560662351355e-06,
+      "loss": 0.9658,
+      "step": 2568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83438418618241e-06,
+      "loss": 0.9653,
+      "step": 2569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834207617524077e-06,
+      "loss": 0.96,
+      "step": 2570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834030956379734e-06,
+      "loss": 0.9404,
+      "step": 2571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83385420275276e-06,
+      "loss": 0.9712,
+      "step": 2572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83367735664654e-06,
+      "loss": 0.9492,
+      "step": 2573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833500418064455e-06,
+      "loss": 0.4791,
+      "step": 2574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833323387009888e-06,
+      "loss": 0.9482,
+      "step": 2575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83314626348623e-06,
+      "loss": 0.9683,
+      "step": 2576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832969047496868e-06,
+      "loss": 0.9307,
+      "step": 2577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832791739045193e-06,
+      "loss": 0.9106,
+      "step": 2578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832614338134595e-06,
+      "loss": 0.9829,
+      "step": 2579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83243684476847e-06,
+      "loss": 0.9517,
+      "step": 2580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832259258950214e-06,
+      "loss": 0.915,
+      "step": 2581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832081580683222e-06,
+      "loss": 0.9614,
+      "step": 2582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831903809970897e-06,
+      "loss": 0.9697,
+      "step": 2583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831725946816636e-06,
+      "loss": 0.9238,
+      "step": 2584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831547991223844e-06,
+      "loss": 0.8608,
+      "step": 2585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831369943195928e-06,
+      "loss": 0.9771,
+      "step": 2586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831191802736289e-06,
+      "loss": 0.9668,
+      "step": 2587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831013569848337e-06,
+      "loss": 1.0088,
+      "step": 2588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830835244535486e-06,
+      "loss": 0.9814,
+      "step": 2589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830656826801142e-06,
+      "loss": 0.937,
+      "step": 2590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830478316648722e-06,
+      "loss": 0.9263,
+      "step": 2591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830299714081638e-06,
+      "loss": 0.402,
+      "step": 2592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83012101910331e-06,
+      "loss": 0.916,
+      "step": 2593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829942231717156e-06,
+      "loss": 0.9678,
+      "step": 2594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829763351926595e-06,
+      "loss": 0.9434,
+      "step": 2595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829584379735051e-06,
+      "loss": 0.9341,
+      "step": 2596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829405315145948e-06,
+      "loss": 0.938,
+      "step": 2597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82922615816271e-06,
+      "loss": 1.0034,
+      "step": 2598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829046908788767e-06,
+      "loss": 0.959,
+      "step": 2599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828867567027544e-06,
+      "loss": 1.0386,
+      "step": 2600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828688132882478e-06,
+      "loss": 0.9297,
+      "step": 2601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828508606356998e-06,
+      "loss": 0.3824,
+      "step": 2602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82832898745454e-06,
+      "loss": 0.979,
+      "step": 2603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82814927617854e-06,
+      "loss": 0.9375,
+      "step": 2604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827969472532436e-06,
+      "loss": 0.9175,
+      "step": 2605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827789576519668e-06,
+      "loss": 0.8804,
+      "step": 2606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827609588143676e-06,
+      "loss": 0.9536,
+      "step": 2607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827429507407909e-06,
+      "loss": 0.877,
+      "step": 2608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827249334315805e-06,
+      "loss": 1.0122,
+      "step": 2609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827069068870815e-06,
+      "loss": 0.9639,
+      "step": 2610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82688871107639e-06,
+      "loss": 0.958,
+      "step": 2611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826708260935976e-06,
+      "loss": 0.9556,
+      "step": 2612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826527718453024e-06,
+      "loss": 0.9819,
+      "step": 2613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826347083630996e-06,
+      "loss": 0.9536,
+      "step": 2614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82616635647334e-06,
+      "loss": 0.9595,
+      "step": 2615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825985536983518e-06,
+      "loss": 1.0879,
+      "step": 2616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825804625164987e-06,
+      "loss": 0.8862,
+      "step": 2617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825623621021207e-06,
+      "loss": 0.9229,
+      "step": 2618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825442524555646e-06,
+      "loss": 0.9688,
+      "step": 2619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825261335771765e-06,
+      "loss": 0.9634,
+      "step": 2620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825080054673031e-06,
+      "loss": 0.8982,
+      "step": 2621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824898681262912e-06,
+      "loss": 0.9229,
+      "step": 2622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824717215544876e-06,
+      "loss": 0.9785,
+      "step": 2623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8245356575224e-06,
+      "loss": 0.9131,
+      "step": 2624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824354007198953e-06,
+      "loss": 0.9995,
+      "step": 2625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824172264578012e-06,
+      "loss": 0.9404,
+      "step": 2626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823990429663052e-06,
+      "loss": 0.8818,
+      "step": 2627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823808502457555e-06,
+      "loss": 0.9717,
+      "step": 2628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823626482964998e-06,
+      "loss": 0.9399,
+      "step": 2629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823444371188869e-06,
+      "loss": 0.9561,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823262167132645e-06,
+      "loss": 0.8872,
+      "step": 2631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823079870799816e-06,
+      "loss": 0.9126,
+      "step": 2632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82289748219387e-06,
+      "loss": 0.3027,
+      "step": 2633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822715001318294e-06,
+      "loss": 0.9238,
+      "step": 2634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822532428176582e-06,
+      "loss": 0.9438,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822349762772224e-06,
+      "loss": 0.9482,
+      "step": 2636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822167005108715e-06,
+      "loss": 0.9434,
+      "step": 2637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821984155189554e-06,
+      "loss": 1.0483,
+      "step": 2638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821801213018238e-06,
+      "loss": 0.9756,
+      "step": 2639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821618178598266e-06,
+      "loss": 1.0337,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82143505193314e-06,
+      "loss": 0.9106,
+      "step": 2641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821251833026364e-06,
+      "loss": 0.9219,
+      "step": 2642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821068521881443e-06,
+      "loss": 0.9692,
+      "step": 2643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820885118501885e-06,
+      "loss": 1.0063,
+      "step": 2644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820701622891196e-06,
+      "loss": 0.9336,
+      "step": 2645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82051803505289e-06,
+      "loss": 0.9653,
+      "step": 2646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820334354990478e-06,
+      "loss": 0.9937,
+      "step": 2647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820150582707473e-06,
+      "loss": 0.9521,
+      "step": 2648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81996671820739e-06,
+      "loss": 0.9888,
+      "step": 2649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81978276149375e-06,
+      "loss": 0.9951,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819598712570071e-06,
+      "loss": 0.9648,
+      "step": 2651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819414571439874e-06,
+      "loss": 0.9395,
+      "step": 2652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819230338106682e-06,
+      "loss": 0.9751,
+      "step": 2653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81904601257402e-06,
+      "loss": 0.9497,
+      "step": 2654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818861594845413e-06,
+      "loss": 0.9482,
+      "step": 2655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81867708492439e-06,
+      "loss": 0.9507,
+      "step": 2656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818492482814481e-06,
+      "loss": 0.9106,
+      "step": 2657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81830778851922e-06,
+      "loss": 1.0151,
+      "step": 2658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818123002042135e-06,
+      "loss": 0.9399,
+      "step": 2659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817938123386766e-06,
+      "loss": 1.0054,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817753152556649e-06,
+      "loss": 0.9873,
+      "step": 2661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817568089555321e-06,
+      "loss": 0.9399,
+      "step": 2662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817382934386324e-06,
+      "loss": 0.3378,
+      "step": 2663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8171976870532e-06,
+      "loss": 0.9595,
+      "step": 2664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817012347559496e-06,
+      "loss": 0.9468,
+      "step": 2665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816826915908754e-06,
+      "loss": 0.9658,
+      "step": 2666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816641392104522e-06,
+      "loss": 0.9038,
+      "step": 2667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81645577615035e-06,
+      "loss": 0.877,
+      "step": 2668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81627006804979e-06,
+      "loss": 0.3695,
+      "step": 2669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816084267806392e-06,
+      "loss": 0.9043,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815898375423715e-06,
+      "loss": 0.8691,
+      "step": 2671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815712390905312e-06,
+      "loss": 0.9287,
+      "step": 2672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815526314254743e-06,
+      "loss": 0.9062,
+      "step": 2673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815340145475568e-06,
+      "loss": 0.9604,
+      "step": 2674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815153884571347e-06,
+      "loss": 0.9404,
+      "step": 2675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814967531545643e-06,
+      "loss": 0.9526,
+      "step": 2676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814781086402025e-06,
+      "loss": 0.9766,
+      "step": 2677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814594549144056e-06,
+      "loss": 0.958,
+      "step": 2678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814407919775308e-06,
+      "loss": 0.981,
+      "step": 2679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81422119829935e-06,
+      "loss": 0.936,
+      "step": 2680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814034384719753e-06,
+      "loss": 0.9697,
+      "step": 2681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813847479040093e-06,
+      "loss": 0.9272,
+      "step": 2682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813660481263944e-06,
+      "loss": 0.9248,
+      "step": 2683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813473391394885e-06,
+      "loss": 0.9497,
+      "step": 2684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813286209436494e-06,
+      "loss": 0.9736,
+      "step": 2685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813098935392355e-06,
+      "loss": 0.9517,
+      "step": 2686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812911569266048e-06,
+      "loss": 0.9438,
+      "step": 2687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812724111061158e-06,
+      "loss": 0.9351,
+      "step": 2688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812536560781275e-06,
+      "loss": 0.9106,
+      "step": 2689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812348918429978e-06,
+      "loss": 0.9746,
+      "step": 2690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812161184010868e-06,
+      "loss": 0.9375,
+      "step": 2691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81197335752753e-06,
+      "loss": 1.0283,
+      "step": 2692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811785438983558e-06,
+      "loss": 0.9458,
+      "step": 2693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811597428382549e-06,
+      "loss": 0.9531,
+      "step": 2694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811409325728099e-06,
+      "loss": 1.0273,
+      "step": 2695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811221131023807e-06,
+      "loss": 0.9409,
+      "step": 2696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811032844273272e-06,
+      "loss": 1.0405,
+      "step": 2697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810844465480098e-06,
+      "loss": 0.896,
+      "step": 2698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810655994647889e-06,
+      "loss": 0.9033,
+      "step": 2699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.81046743178025e-06,
+      "loss": 0.9673,
+      "step": 2700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810278776880789e-06,
+      "loss": 0.9458,
+      "step": 2701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810090029953115e-06,
+      "loss": 0.9395,
+      "step": 2702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809901191000839e-06,
+      "loss": 0.9683,
+      "step": 2703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809712260027573e-06,
+      "loss": 0.959,
+      "step": 2704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809523237036932e-06,
+      "loss": 0.9692,
+      "step": 2705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809334122032534e-06,
+      "loss": 0.9893,
+      "step": 2706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809144915017995e-06,
+      "loss": 0.9326,
+      "step": 2707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808955615996934e-06,
+      "loss": 0.9834,
+      "step": 2708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808766224972976e-06,
+      "loss": 0.9287,
+      "step": 2709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808576741949742e-06,
+      "loss": 0.9282,
+      "step": 2710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808387166930856e-06,
+      "loss": 0.9287,
+      "step": 2711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808197499919949e-06,
+      "loss": 0.9023,
+      "step": 2712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808007740920647e-06,
+      "loss": 0.9292,
+      "step": 2713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807817889936578e-06,
+      "loss": 0.9307,
+      "step": 2714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80762794697138e-06,
+      "loss": 1.0498,
+      "step": 2715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807437912028679e-06,
+      "loss": 0.9326,
+      "step": 2716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807247785112118e-06,
+      "loss": 0.9668,
+      "step": 2717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80705756622533e-06,
+      "loss": 0.9067,
+      "step": 2718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806867255371959e-06,
+      "loss": 0.9312,
+      "step": 2719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80667685255564e-06,
+      "loss": 0.9414,
+      "step": 2720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806486357780019e-06,
+      "loss": 0.9795,
+      "step": 2721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806295771048741e-06,
+      "loss": 0.9448,
+      "step": 2722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80610509236545e-06,
+      "loss": 0.9839,
+      "step": 2723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805914321733795e-06,
+      "loss": 0.9521,
+      "step": 2724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805723459157426e-06,
+      "loss": 0.9946,
+      "step": 2725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805532504639994e-06,
+      "loss": 0.9551,
+      "step": 2726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805341458185154e-06,
+      "loss": 0.9404,
+      "step": 2727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805150319796558e-06,
+      "loss": 0.9414,
+      "step": 2728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804959089477865e-06,
+      "loss": 0.9336,
+      "step": 2729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804767767232733e-06,
+      "loss": 0.9634,
+      "step": 2730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804576353064822e-06,
+      "loss": 0.9272,
+      "step": 2731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804384846977795e-06,
+      "loss": 0.9072,
+      "step": 2732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804193248975314e-06,
+      "loss": 0.9463,
+      "step": 2733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804001559061046e-06,
+      "loss": 0.9209,
+      "step": 2734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80380977723866e-06,
+      "loss": 0.9785,
+      "step": 2735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80361790351182e-06,
+      "loss": 0.9941,
+      "step": 2736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803425937884202e-06,
+      "loss": 0.8838,
+      "step": 2737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803233880359475e-06,
+      "loss": 0.9653,
+      "step": 2738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803041730941317e-06,
+      "loss": 0.9854,
+      "step": 2739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802849489633402e-06,
+      "loss": 0.9556,
+      "step": 2740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802657156439405e-06,
+      "loss": 0.9473,
+      "step": 2741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80246473136301e-06,
+      "loss": 0.9692,
+      "step": 2742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802272214407898e-06,
+      "loss": 0.937,
+      "step": 2743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80207960557775e-06,
+      "loss": 0.998,
+      "step": 2744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801886904876251e-06,
+      "loss": 0.8755,
+      "step": 2745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801694112307092e-06,
+      "loss": 0.9297,
+      "step": 2746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801501227873955e-06,
+      "loss": 0.9741,
+      "step": 2747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801308251580536e-06,
+      "loss": 0.9204,
+      "step": 2748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801115183430524e-06,
+      "loss": 0.3063,
+      "step": 2749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80092202342761e-06,
+      "loss": 0.9541,
+      "step": 2750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800728771575494e-06,
+      "loss": 0.9634,
+      "step": 2751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800535427877871e-06,
+      "loss": 0.9551,
+      "step": 2752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800341992338443e-06,
+      "loss": 0.8989,
+      "step": 2753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800148464960906e-06,
+      "loss": 0.9756,
+      "step": 2754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799954845748964e-06,
+      "loss": 0.9688,
+      "step": 2755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799761134706323e-06,
+      "loss": 0.9302,
+      "step": 2756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799567331836688e-06,
+      "loss": 1.0234,
+      "step": 2757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799373437143768e-06,
+      "loss": 0.98,
+      "step": 2758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799179450631269e-06,
+      "loss": 0.9634,
+      "step": 2759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798985372302908e-06,
+      "loss": 0.9204,
+      "step": 2760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79879120216239e-06,
+      "loss": 0.3009,
+      "step": 2761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798596940213438e-06,
+      "loss": 0.9951,
+      "step": 2762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798402586459763e-06,
+      "loss": 0.9033,
+      "step": 2763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798208140905087e-06,
+      "loss": 0.9048,
+      "step": 2764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798013603553127e-06,
+      "loss": 0.8896,
+      "step": 2765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797818974407606e-06,
+      "loss": 0.9629,
+      "step": 2766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797624253472246e-06,
+      "loss": 0.9556,
+      "step": 2767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797429440750777e-06,
+      "loss": 0.9482,
+      "step": 2768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797234536246921e-06,
+      "loss": 0.9634,
+      "step": 2769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79703953996441e-06,
+      "loss": 0.9771,
+      "step": 2770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796844451906973e-06,
+      "loss": 0.9268,
+      "step": 2771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796649272078342e-06,
+      "loss": 0.9556,
+      "step": 2772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796454000482252e-06,
+      "loss": 0.9482,
+      "step": 2773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79625863712244e-06,
+      "loss": 1.0024,
+      "step": 2774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796063182002642e-06,
+      "loss": 0.9619,
+      "step": 2775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795867635126595e-06,
+      "loss": 0.9473,
+      "step": 2776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795671996498046e-06,
+      "loss": 0.9546,
+      "step": 2777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795476266120734e-06,
+      "loss": 0.3466,
+      "step": 2778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795280443998402e-06,
+      "loss": 0.9414,
+      "step": 2779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795084530134801e-06,
+      "loss": 1.0186,
+      "step": 2780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794888524533676e-06,
+      "loss": 0.9229,
+      "step": 2781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79469242719878e-06,
+      "loss": 0.9536,
+      "step": 2782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79449623813386e-06,
+      "loss": 1.0142,
+      "step": 2783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794299957342672e-06,
+      "loss": 0.9629,
+      "step": 2784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794103584828972e-06,
+      "loss": 0.9448,
+      "step": 2785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793907120596513e-06,
+      "loss": 0.9927,
+      "step": 2786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793710564649059e-06,
+      "loss": 0.9795,
+      "step": 2787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793513916990367e-06,
+      "loss": 0.9697,
+      "step": 2788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793317177624201e-06,
+      "loss": 0.8916,
+      "step": 2789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793120346554324e-06,
+      "loss": 0.8789,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792923423784502e-06,
+      "loss": 0.9595,
+      "step": 2791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7927264093185e-06,
+      "loss": 0.9336,
+      "step": 2792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792529303160093e-06,
+      "loss": 0.9292,
+      "step": 2793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792332105313045e-06,
+      "loss": 0.9009,
+      "step": 2794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792134815781133e-06,
+      "loss": 0.9639,
+      "step": 2795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791937434568132e-06,
+      "loss": 0.9937,
+      "step": 2796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791739961677815e-06,
+      "loss": 0.9531,
+      "step": 2797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791542397113963e-06,
+      "loss": 0.8994,
+      "step": 2798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791344740880353e-06,
+      "loss": 1.0317,
+      "step": 2799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791146992980769e-06,
+      "loss": 0.9604,
+      "step": 2800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790949153418992e-06,
+      "loss": 0.978,
+      "step": 2801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79075122219881e-06,
+      "loss": 0.9497,
+      "step": 2802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790553199324007e-06,
+      "loss": 0.9341,
+      "step": 2803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790355084798371e-06,
+      "loss": 0.9302,
+      "step": 2804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790156878625693e-06,
+      "loss": 0.9966,
+      "step": 2805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789958580809768e-06,
+      "loss": 0.9551,
+      "step": 2806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789760191354387e-06,
+      "loss": 0.9204,
+      "step": 2807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789561710263346e-06,
+      "loss": 1.0112,
+      "step": 2808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789363137540441e-06,
+      "loss": 0.9595,
+      "step": 2809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789164473189472e-06,
+      "loss": 0.9131,
+      "step": 2810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78896571721424e-06,
+      "loss": 0.9458,
+      "step": 2811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788766869618548e-06,
+      "loss": 0.9497,
+      "step": 2812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788567930406198e-06,
+      "loss": 0.9082,
+      "step": 2813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788368899581e-06,
+      "loss": 0.9941,
+      "step": 2814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788169777146758e-06,
+      "loss": 0.9941,
+      "step": 2815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787970563107283e-06,
+      "loss": 0.9214,
+      "step": 2816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787771257466388e-06,
+      "loss": 0.9473,
+      "step": 2817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787571860227882e-06,
+      "loss": 0.8813,
+      "step": 2818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787372371395583e-06,
+      "loss": 0.9849,
+      "step": 2819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787172790973307e-06,
+      "loss": 0.9458,
+      "step": 2820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78697311896487e-06,
+      "loss": 0.3655,
+      "step": 2821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786773355374095e-06,
+      "loss": 0.9688,
+      "step": 2822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786573500204804e-06,
+      "loss": 0.8696,
+      "step": 2823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786373553460817e-06,
+      "loss": 0.8857,
+      "step": 2824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786173515145963e-06,
+      "loss": 0.981,
+      "step": 2825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785973385264066e-06,
+      "loss": 0.9199,
+      "step": 2826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785773163818959e-06,
+      "loss": 0.9092,
+      "step": 2827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785572850814468e-06,
+      "loss": 0.9316,
+      "step": 2828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785372446254427e-06,
+      "loss": 0.9614,
+      "step": 2829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785171950142671e-06,
+      "loss": 0.9473,
+      "step": 2830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784971362483036e-06,
+      "loss": 0.96,
+      "step": 2831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784770683279357e-06,
+      "loss": 0.9351,
+      "step": 2832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784569912535475e-06,
+      "loss": 0.9282,
+      "step": 2833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784369050255233e-06,
+      "loss": 0.9536,
+      "step": 2834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78416809644247e-06,
+      "loss": 0.9077,
+      "step": 2835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783967051101033e-06,
+      "loss": 0.9351,
+      "step": 2836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783765914234768e-06,
+      "loss": 0.8853,
+      "step": 2837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783564685847521e-06,
+      "loss": 0.9849,
+      "step": 2838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783363365943145e-06,
+      "loss": 0.9429,
+      "step": 2839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783161954525492e-06,
+      "loss": 0.9131,
+      "step": 2840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782960451598411e-06,
+      "loss": 0.9902,
+      "step": 2841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782758857165761e-06,
+      "loss": 0.9229,
+      "step": 2842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782557171231397e-06,
+      "loss": 0.8945,
+      "step": 2843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782355393799176e-06,
+      "loss": 0.9688,
+      "step": 2844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782153524872962e-06,
+      "loss": 0.9414,
+      "step": 2845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781951564456615e-06,
+      "loss": 0.9644,
+      "step": 2846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781749512554e-06,
+      "loss": 0.9658,
+      "step": 2847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781547369168983e-06,
+      "loss": 0.9829,
+      "step": 2848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781345134305426e-06,
+      "loss": 0.9248,
+      "step": 2849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781142807967205e-06,
+      "loss": 0.9722,
+      "step": 2850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780940390158186e-06,
+      "loss": 0.9248,
+      "step": 2851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780737880882245e-06,
+      "loss": 0.9009,
+      "step": 2852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780535280143255e-06,
+      "loss": 0.9995,
+      "step": 2853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780332587945092e-06,
+      "loss": 1.0239,
+      "step": 2854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780129804291633e-06,
+      "loss": 0.9795,
+      "step": 2855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779926929186758e-06,
+      "loss": 0.8989,
+      "step": 2856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779723962634349e-06,
+      "loss": 0.9966,
+      "step": 2857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779520904638289e-06,
+      "loss": 0.9424,
+      "step": 2858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779317755202463e-06,
+      "loss": 0.9263,
+      "step": 2859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779114514330756e-06,
+      "loss": 0.9658,
+      "step": 2860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778911182027057e-06,
+      "loss": 0.9409,
+      "step": 2861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778707758295258e-06,
+      "loss": 0.9888,
+      "step": 2862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778504243139248e-06,
+      "loss": 0.319,
+      "step": 2863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778300636562923e-06,
+      "loss": 0.9697,
+      "step": 2864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778096938570176e-06,
+      "loss": 0.8921,
+      "step": 2865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777893149164905e-06,
+      "loss": 0.9541,
+      "step": 2866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77768926835101e-06,
+      "loss": 0.9683,
+      "step": 2867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777485296132391e-06,
+      "loss": 0.9736,
+      "step": 2868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77728123251295e-06,
+      "loss": 0.9302,
+      "step": 2869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777077077496588e-06,
+      "loss": 1.019,
+      "step": 2870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776872831087216e-06,
+      "loss": 0.9585,
+      "step": 2871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77666849328874e-06,
+      "loss": 0.9233,
+      "step": 2872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776464064105068e-06,
+      "loss": 0.9092,
+      "step": 2873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776259543540112e-06,
+      "loss": 0.9092,
+      "step": 2874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776054931597785e-06,
+      "loss": 0.9883,
+      "step": 2875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775850228281999e-06,
+      "loss": 0.9541,
+      "step": 2876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775645433596674e-06,
+      "loss": 0.9707,
+      "step": 2877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775440547545725e-06,
+      "loss": 0.8887,
+      "step": 2878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775235570133074e-06,
+      "loss": 0.9224,
+      "step": 2879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775030501362641e-06,
+      "loss": 0.958,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774825341238353e-06,
+      "loss": 0.9478,
+      "step": 2881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774620089764128e-06,
+      "loss": 0.9263,
+      "step": 2882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7744147469439e-06,
+      "loss": 0.9214,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774209312781591e-06,
+      "loss": 0.9785,
+      "step": 2884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774003787281137e-06,
+      "loss": 0.9585,
+      "step": 2885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773798170446467e-06,
+      "loss": 0.9062,
+      "step": 2886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773592462281516e-06,
+      "loss": 0.9683,
+      "step": 2887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773386662790219e-06,
+      "loss": 0.9277,
+      "step": 2888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773180771976511e-06,
+      "loss": 0.9316,
+      "step": 2889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772974789844335e-06,
+      "loss": 0.8809,
+      "step": 2890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772768716397632e-06,
+      "loss": 1.0103,
+      "step": 2891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772562551640339e-06,
+      "loss": 0.3024,
+      "step": 2892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772356295576407e-06,
+      "loss": 0.9126,
+      "step": 2893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772149948209774e-06,
+      "loss": 0.9038,
+      "step": 2894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771943509544397e-06,
+      "loss": 0.3109,
+      "step": 2895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77173697958422e-06,
+      "loss": 1.001,
+      "step": 2896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771530358333194e-06,
+      "loss": 0.9663,
+      "step": 2897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771323645795275e-06,
+      "loss": 0.958,
+      "step": 2898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771116841974415e-06,
+      "loss": 0.9614,
+      "step": 2899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770909946874572e-06,
+      "loss": 0.9736,
+      "step": 2900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770702960499702e-06,
+      "loss": 0.8799,
+      "step": 2901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770495882853768e-06,
+      "loss": 0.958,
+      "step": 2902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77028871394073e-06,
+      "loss": 0.9448,
+      "step": 2903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770081453764553e-06,
+      "loss": 0.9731,
+      "step": 2904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7698741023292e-06,
+      "loss": 0.9614,
+      "step": 2905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769666659638638e-06,
+      "loss": 0.9355,
+      "step": 2906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769459125696838e-06,
+      "loss": 0.8857,
+      "step": 2907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76925150050777e-06,
+      "loss": 0.9629,
+      "step": 2908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769043784075404e-06,
+      "loss": 0.9219,
+      "step": 2909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768835976403716e-06,
+      "loss": 0.9043,
+      "step": 2910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768628077496682e-06,
+      "loss": 0.8794,
+      "step": 2911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768420087358276e-06,
+      "loss": 0.8921,
+      "step": 2912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768212005992482e-06,
+      "loss": 0.9268,
+      "step": 2913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768003833403278e-06,
+      "loss": 0.918,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767795569594647e-06,
+      "loss": 0.9043,
+      "step": 2915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767587214570573e-06,
+      "loss": 0.9365,
+      "step": 2916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767378768335042e-06,
+      "loss": 1.0005,
+      "step": 2917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767170230892045e-06,
+      "loss": 0.9517,
+      "step": 2918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766961602245568e-06,
+      "loss": 0.9409,
+      "step": 2919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766752882399603e-06,
+      "loss": 0.9517,
+      "step": 2920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766544071358146e-06,
+      "loss": 0.9204,
+      "step": 2921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766335169125189e-06,
+      "loss": 0.9507,
+      "step": 2922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766126175704729e-06,
+      "loss": 0.8901,
+      "step": 2923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765917091100766e-06,
+      "loss": 0.9492,
+      "step": 2924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765707915317297e-06,
+      "loss": 0.853,
+      "step": 2925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765498648358326e-06,
+      "loss": 0.999,
+      "step": 2926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765289290227858e-06,
+      "loss": 0.9663,
+      "step": 2927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.765079840929894e-06,
+      "loss": 0.9834,
+      "step": 2928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764870300468446e-06,
+      "loss": 1.0171,
+      "step": 2929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76466066884752e-06,
+      "loss": 0.9121,
+      "step": 2930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764450946071127e-06,
+      "loss": 0.9614,
+      "step": 2931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76424113214328e-06,
+      "loss": 0.8662,
+      "step": 2932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764031227067993e-06,
+      "loss": 0.9609,
+      "step": 2933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763821230849282e-06,
+      "loss": 0.9302,
+      "step": 2934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763611143491163e-06,
+      "loss": 0.9092,
+      "step": 2935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763400964997656e-06,
+      "loss": 0.9409,
+      "step": 2936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763190695372782e-06,
+      "loss": 0.8701,
+      "step": 2937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762980334620564e-06,
+      "loss": 0.9478,
+      "step": 2938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76276988274503e-06,
+      "loss": 0.9722,
+      "step": 2939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762559339750199e-06,
+      "loss": 0.9429,
+      "step": 2940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762348705640103e-06,
+      "loss": 0.9487,
+      "step": 2941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762137980418775e-06,
+      "loss": 0.9561,
+      "step": 2942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76192716409024e-06,
+      "loss": 0.957,
+      "step": 2943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761716256658536e-06,
+      "loss": 1.0068,
+      "step": 2944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761505258127697e-06,
+      "loss": 0.9458,
+      "step": 2945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761294168501758e-06,
+      "loss": 0.3306,
+      "step": 2946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76108298778476e-06,
+      "loss": 0.939,
+      "step": 2947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76087171598074e-06,
+      "loss": 0.8687,
+      "step": 2948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760660353093743e-06,
+      "loss": 0.957,
+      "step": 2949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760448899127812e-06,
+      "loss": 0.9351,
+      "step": 2950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760237354086992e-06,
+      "loss": 0.9556,
+      "step": 2951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76002571797533e-06,
+      "loss": 0.9404,
+      "step": 2952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759813990796874e-06,
+      "loss": 0.9209,
+      "step": 2953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759602172555678e-06,
+      "loss": 0.9473,
+      "step": 2954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75939026325579e-06,
+      "loss": 0.9233,
+      "step": 2955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759178262901267e-06,
+      "loss": 0.9053,
+      "step": 2956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758966171496165e-06,
+      "loss": 0.9058,
+      "step": 2957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75875398904454e-06,
+      "loss": 0.9873,
+      "step": 2958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758541715550454e-06,
+      "loss": 0.936,
+      "step": 2959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758329351017965e-06,
+      "loss": 0.9502,
+      "step": 2960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758116895451137e-06,
+      "loss": 0.9287,
+      "step": 2961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757904348854037e-06,
+      "loss": 0.9087,
+      "step": 2962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757691711230728e-06,
+      "loss": 0.9756,
+      "step": 2963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757478982585279e-06,
+      "loss": 0.9214,
+      "step": 2964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757266162921762e-06,
+      "loss": 0.894,
+      "step": 2965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757053252244245e-06,
+      "loss": 0.9116,
+      "step": 2966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756840250556803e-06,
+      "loss": 0.9512,
+      "step": 2967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756627157863512e-06,
+      "loss": 0.8823,
+      "step": 2968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756413974168446e-06,
+      "loss": 0.9097,
+      "step": 2969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756200699475688e-06,
+      "loss": 0.9224,
+      "step": 2970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755987333789314e-06,
+      "loss": 0.9404,
+      "step": 2971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75577387711341e-06,
+      "loss": 0.8999,
+      "step": 2972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755560329452055e-06,
+      "loss": 0.9517,
+      "step": 2973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755346690809338e-06,
+      "loss": 0.9092,
+      "step": 2974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755132961189346e-06,
+      "loss": 0.9565,
+      "step": 2975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754919140596165e-06,
+      "loss": 0.9644,
+      "step": 2976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754705229033893e-06,
+      "loss": 0.9473,
+      "step": 2977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754491226506613e-06,
+      "loss": 0.9775,
+      "step": 2978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754277133018425e-06,
+      "loss": 0.96,
+      "step": 2979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754062948573424e-06,
+      "loss": 0.9736,
+      "step": 2980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753848673175707e-06,
+      "loss": 0.9834,
+      "step": 2981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753634306829373e-06,
+      "loss": 0.9434,
+      "step": 2982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753419849538526e-06,
+      "loss": 0.9175,
+      "step": 2983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753205301307265e-06,
+      "loss": 0.9355,
+      "step": 2984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752990662139696e-06,
+      "loss": 0.9707,
+      "step": 2985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752775932039928e-06,
+      "loss": 0.9688,
+      "step": 2986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752561111012066e-06,
+      "loss": 0.9268,
+      "step": 2987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752346199060219e-06,
+      "loss": 0.9307,
+      "step": 2988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752131196188503e-06,
+      "loss": 0.9062,
+      "step": 2989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751916102401028e-06,
+      "loss": 0.9341,
+      "step": 2990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751700917701909e-06,
+      "loss": 0.9253,
+      "step": 2991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751485642095266e-06,
+      "loss": 0.9326,
+      "step": 2992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751270275585213e-06,
+      "loss": 0.917,
+      "step": 2993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751054818175873e-06,
+      "loss": 1.0049,
+      "step": 2994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750839269871368e-06,
+      "loss": 0.9497,
+      "step": 2995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750623630675822e-06,
+      "loss": 0.9189,
+      "step": 2996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750407900593358e-06,
+      "loss": 0.9932,
+      "step": 2997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750192079628107e-06,
+      "loss": 0.9658,
+      "step": 2998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749976167784194e-06,
+      "loss": 0.9771,
+      "step": 2999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749760165065754e-06,
+      "loss": 0.9966,
+      "step": 3000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749544071476917e-06,
+      "loss": 0.9004,
+      "step": 3001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749327887021817e-06,
+      "loss": 0.9561,
+      "step": 3002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749111611704592e-06,
+      "loss": 0.9048,
+      "step": 3003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748895245529375e-06,
+      "loss": 0.9492,
+      "step": 3004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748678788500311e-06,
+      "loss": 0.8926,
+      "step": 3005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74846224062154e-06,
+      "loss": 0.9712,
+      "step": 3006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7482456018972e-06,
+      "loss": 0.9526,
+      "step": 3007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748028872331442e-06,
+      "loss": 0.9492,
+      "step": 3008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74781205192841e-06,
+      "loss": 0.9521,
+      "step": 3009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74759514069225e-06,
+      "loss": 0.8979,
+      "step": 3010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747378138627115e-06,
+      "loss": 0.9795,
+      "step": 3011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747161045737154e-06,
+      "loss": 0.9238,
+      "step": 3012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746943862026522e-06,
+      "loss": 0.959,
+      "step": 3013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746726587499374e-06,
+      "loss": 0.9424,
+      "step": 3014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746509222159868e-06,
+      "loss": 1.0039,
+      "step": 3015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74629176601216e-06,
+      "loss": 0.9321,
+      "step": 3016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74607421906041e-06,
+      "loss": 0.8794,
+      "step": 3017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745856581308782e-06,
+      "loss": 0.9126,
+      "step": 3018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745638852761439e-06,
+      "loss": 0.9316,
+      "step": 3019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745421033422547e-06,
+      "loss": 0.8931,
+      "step": 3020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745203123296272e-06,
+      "loss": 0.9263,
+      "step": 3021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744985122386784e-06,
+      "loss": 0.9907,
+      "step": 3022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744767030698254e-06,
+      "loss": 0.9263,
+      "step": 3023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744548848234854e-06,
+      "loss": 0.9155,
+      "step": 3024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744330575000757e-06,
+      "loss": 0.8931,
+      "step": 3025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74411221100014e-06,
+      "loss": 0.9492,
+      "step": 3026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743893756237181e-06,
+      "loss": 0.3562,
+      "step": 3027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74367521071606e-06,
+      "loss": 0.9565,
+      "step": 3028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743456574440957e-06,
+      "loss": 0.9482,
+      "step": 3029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743237847416053e-06,
+      "loss": 0.9458,
+      "step": 3030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743019029645537e-06,
+      "loss": 0.9717,
+      "step": 3031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742800121133592e-06,
+      "loss": 0.9087,
+      "step": 3032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742581121884408e-06,
+      "loss": 0.9531,
+      "step": 3033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74236203190217e-06,
+      "loss": 0.8823,
+      "step": 3034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742142851191077e-06,
+      "loss": 0.9497,
+      "step": 3035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741923579755316e-06,
+      "loss": 0.9619,
+      "step": 3036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741704217599088e-06,
+      "loss": 0.9697,
+      "step": 3037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741484764726582e-06,
+      "loss": 0.8955,
+      "step": 3038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741265221142005e-06,
+      "loss": 0.9243,
+      "step": 3039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74104558684955e-06,
+      "loss": 0.9272,
+      "step": 3040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740825861853425e-06,
+      "loss": 0.9668,
+      "step": 3041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740606046157828e-06,
+      "loss": 0.9658,
+      "step": 3042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740386139766968e-06,
+      "loss": 0.9697,
+      "step": 3043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74016614268505e-06,
+      "loss": 0.9717,
+      "step": 3044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739946054916286e-06,
+      "loss": 0.8755,
+      "step": 3045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739725876464883e-06,
+      "loss": 0.9624,
+      "step": 3046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739505607335055e-06,
+      "loss": 0.9102,
+      "step": 3047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739285247531019e-06,
+      "loss": 1.0029,
+      "step": 3048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739064797056985e-06,
+      "loss": 0.9434,
+      "step": 3049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738844255917174e-06,
+      "loss": 0.9043,
+      "step": 3050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738623624115807e-06,
+      "loss": 0.9854,
+      "step": 3051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738402901657101e-06,
+      "loss": 0.8574,
+      "step": 3052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738182088545282e-06,
+      "loss": 0.9741,
+      "step": 3053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737961184784571e-06,
+      "loss": 0.959,
+      "step": 3054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737740190379199e-06,
+      "loss": 0.9585,
+      "step": 3055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73751910533339e-06,
+      "loss": 0.9829,
+      "step": 3056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737297929651376e-06,
+      "loss": 1.0205,
+      "step": 3057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737076663337388e-06,
+      "loss": 0.959,
+      "step": 3058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736855306395656e-06,
+      "loss": 0.9985,
+      "step": 3059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73663385883042e-06,
+      "loss": 0.917,
+      "step": 3060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736412320645915e-06,
+      "loss": 0.9536,
+      "step": 3061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736190691846377e-06,
+      "loss": 0.8984,
+      "step": 3062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735968972436047e-06,
+      "loss": 0.9219,
+      "step": 3063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735747162419168e-06,
+      "loss": 0.9702,
+      "step": 3064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735525261799985e-06,
+      "loss": 0.981,
+      "step": 3065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73530327058274e-06,
+      "loss": 1.0142,
+      "step": 3066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735081188771683e-06,
+      "loss": 0.9521,
+      "step": 3067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734859016371058e-06,
+      "loss": 0.9419,
+      "step": 3068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734636753385122e-06,
+      "loss": 1.0049,
+      "step": 3069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734414399818123e-06,
+      "loss": 0.9385,
+      "step": 3070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734191955674316e-06,
+      "loss": 0.9287,
+      "step": 3071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733969420957955e-06,
+      "loss": 0.9219,
+      "step": 3072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733746795673302e-06,
+      "loss": 0.9492,
+      "step": 3073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733524079824612e-06,
+      "loss": 0.9268,
+      "step": 3074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733301273416148e-06,
+      "loss": 0.8911,
+      "step": 3075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733078376452172e-06,
+      "loss": 0.939,
+      "step": 3076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732855388936946e-06,
+      "loss": 0.3834,
+      "step": 3077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73263231087474e-06,
+      "loss": 0.9053,
+      "step": 3078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73240914226982e-06,
+      "loss": 0.8911,
+      "step": 3079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732185883126456e-06,
+      "loss": 0.9565,
+      "step": 3080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73196253344892e-06,
+      "loss": 0.958,
+      "step": 3081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731739093241484e-06,
+      "loss": 0.9482,
+      "step": 3082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731515562508422e-06,
+      "loss": 0.9604,
+      "step": 3083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73129194125401e-06,
+      "loss": 0.936,
+      "step": 3084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73106822948253e-06,
+      "loss": 0.9746,
+      "step": 3085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730844427198259e-06,
+      "loss": 1.0054,
+      "step": 3086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730620534405478e-06,
+      "loss": 0.936,
+      "step": 3087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730396551108472e-06,
+      "loss": 0.9648,
+      "step": 3088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730172477311526e-06,
+      "loss": 0.9512,
+      "step": 3089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729948313018926e-06,
+      "loss": 0.4202,
+      "step": 3090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729724058234964e-06,
+      "loss": 0.9468,
+      "step": 3091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729499712963923e-06,
+      "loss": 0.309,
+      "step": 3092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729275277210104e-06,
+      "loss": 0.9775,
+      "step": 3093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729050750977795e-06,
+      "loss": 0.9399,
+      "step": 3094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728826134271292e-06,
+      "loss": 0.9004,
+      "step": 3095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728601427094894e-06,
+      "loss": 0.9604,
+      "step": 3096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7283766294529e-06,
+      "loss": 0.9565,
+      "step": 3097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72815174134961e-06,
+      "loss": 0.9614,
+      "step": 3098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727926762789327e-06,
+      "loss": 0.9126,
+      "step": 3099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727701693776355e-06,
+      "loss": 0.9395,
+      "step": 3100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727476534314997e-06,
+      "loss": 0.9053,
+      "step": 3101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727251284409566e-06,
+      "loss": 0.9287,
+      "step": 3102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727025944064368e-06,
+      "loss": 0.9551,
+      "step": 3103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726800513283714e-06,
+      "loss": 0.8984,
+      "step": 3104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726574992071918e-06,
+      "loss": 0.9897,
+      "step": 3105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726349380433296e-06,
+      "loss": 0.96,
+      "step": 3106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72612367837216e-06,
+      "loss": 0.9229,
+      "step": 3107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725897885892831e-06,
+      "loss": 0.9863,
+      "step": 3108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725672002999629e-06,
+      "loss": 0.8999,
+      "step": 3109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725446029696876e-06,
+      "loss": 0.9175,
+      "step": 3110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725219965988892e-06,
+      "loss": 0.9106,
+      "step": 3111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724993811880006e-06,
+      "loss": 0.8245,
+      "step": 3112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72476756737454e-06,
+      "loss": 0.9551,
+      "step": 3113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724541232476827e-06,
+      "loss": 0.9648,
+      "step": 3114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724314807191197e-06,
+      "loss": 0.96,
+      "step": 3115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724088291521979e-06,
+      "loss": 0.9731,
+      "step": 3116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723861685473505e-06,
+      "loss": 0.9341,
+      "step": 3117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723634989050116e-06,
+      "loss": 0.9331,
+      "step": 3118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723408202256144e-06,
+      "loss": 0.9517,
+      "step": 3119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723181325095931e-06,
+      "loss": 0.9497,
+      "step": 3120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722954357573818e-06,
+      "loss": 0.9072,
+      "step": 3121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722727299694143e-06,
+      "loss": 0.8999,
+      "step": 3122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722500151461253e-06,
+      "loss": 0.9717,
+      "step": 3123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722272912879496e-06,
+      "loss": 0.8823,
+      "step": 3124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722045583953215e-06,
+      "loss": 0.96,
+      "step": 3125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72181816468676e-06,
+      "loss": 0.9487,
+      "step": 3126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721590655084485e-06,
+      "loss": 0.9502,
+      "step": 3127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721363055150739e-06,
+      "loss": 0.9219,
+      "step": 3128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721135364889878e-06,
+      "loss": 0.9355,
+      "step": 3129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720907584306259e-06,
+      "loss": 0.9307,
+      "step": 3130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720679713404238e-06,
+      "loss": 0.9053,
+      "step": 3131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720451752188173e-06,
+      "loss": 0.9624,
+      "step": 3132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72022370066243e-06,
+      "loss": 0.9258,
+      "step": 3133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719995558831369e-06,
+      "loss": 0.8726,
+      "step": 3134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719767326699355e-06,
+      "loss": 0.8901,
+      "step": 3135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719539004270754e-06,
+      "loss": 0.9614,
+      "step": 3136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719310591549937e-06,
+      "loss": 0.9248,
+      "step": 3137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71908208854127e-06,
+      "loss": 0.9746,
+      "step": 3138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718853495249127e-06,
+      "loss": 0.8794,
+      "step": 3139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71862481167788e-06,
+      "loss": 0.9502,
+      "step": 3140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718396037831904e-06,
+      "loss": 0.9722,
+      "step": 3141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718167173715577e-06,
+      "loss": 0.8926,
+      "step": 3142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717938219333279e-06,
+      "loss": 0.9956,
+      "step": 3143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717709174689385e-06,
+      "loss": 0.3762,
+      "step": 3144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717480039788283e-06,
+      "loss": 0.9492,
+      "step": 3145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717250814634353e-06,
+      "loss": 0.9561,
+      "step": 3146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71702149923198e-06,
+      "loss": 0.9907,
+      "step": 3147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716792093585555e-06,
+      "loss": 0.9087,
+      "step": 3148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716562597699464e-06,
+      "loss": 0.9541,
+      "step": 3149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716333011578097e-06,
+      "loss": 1.0078,
+      "step": 3150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716103335225849e-06,
+      "loss": 0.9053,
+      "step": 3151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715873568647111e-06,
+      "loss": 0.9824,
+      "step": 3152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71564371184628e-06,
+      "loss": 0.9375,
+      "step": 3153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715413764827757e-06,
+      "loss": 0.7905,
+      "step": 3154
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715183727595937e-06,
+      "loss": 0.9751,
+      "step": 3155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71495360015522e-06,
+      "loss": 0.9443,
+      "step": 3156
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714723382510011e-06,
+      "loss": 0.8931,
+      "step": 3157
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714493074664717e-06,
+      "loss": 0.9429,
+      "step": 3158
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714262676623737e-06,
+      "loss": 0.9478,
+      "step": 3159
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714032188391488e-06,
+      "loss": 0.9766,
+      "step": 3160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71380160997237e-06,
+      "loss": 1.0146,
+      "step": 3161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713570941370804e-06,
+      "loss": 0.8979,
+      "step": 3162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713340182591196e-06,
+      "loss": 0.9302,
+      "step": 3163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713109333637962e-06,
+      "loss": 0.9209,
+      "step": 3164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712878394515522e-06,
+      "loss": 0.9536,
+      "step": 3165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712647365228289e-06,
+      "loss": 0.9014,
+      "step": 3166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712416245780686e-06,
+      "loss": 0.9746,
+      "step": 3167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712185036177137e-06,
+      "loss": 0.9258,
+      "step": 3168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71195373642206e-06,
+      "loss": 0.9561,
+      "step": 3169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711722346519885e-06,
+      "loss": 0.8818,
+      "step": 3170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711490866475035e-06,
+      "loss": 0.9502,
+      "step": 3171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711259296291941e-06,
+      "loss": 0.9282,
+      "step": 3172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711027635975032e-06,
+      "loss": 0.8989,
+      "step": 3173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710795885528741e-06,
+      "loss": 0.9951,
+      "step": 3174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710564044957503e-06,
+      "loss": 0.9272,
+      "step": 3175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710332114265751e-06,
+      "loss": 0.9194,
+      "step": 3176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710100093457923e-06,
+      "loss": 0.9541,
+      "step": 3177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709867982538457e-06,
+      "loss": 0.915,
+      "step": 3178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709635781511796e-06,
+      "loss": 0.916,
+      "step": 3179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709403490382381e-06,
+      "loss": 0.8999,
+      "step": 3180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709171109154657e-06,
+      "loss": 0.9688,
+      "step": 3181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708938637833065e-06,
+      "loss": 0.9355,
+      "step": 3182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708706076422059e-06,
+      "loss": 0.9619,
+      "step": 3183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708473424926087e-06,
+      "loss": 0.978,
+      "step": 3184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708240683349598e-06,
+      "loss": 0.9097,
+      "step": 3185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708007851697044e-06,
+      "loss": 0.9292,
+      "step": 3186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707774929972881e-06,
+      "loss": 0.9482,
+      "step": 3187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707541918181567e-06,
+      "loss": 0.9639,
+      "step": 3188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707308816327557e-06,
+      "loss": 0.8794,
+      "step": 3189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707075624415312e-06,
+      "loss": 0.9072,
+      "step": 3190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706842342449291e-06,
+      "loss": 0.4014,
+      "step": 3191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706608970433962e-06,
+      "loss": 0.9761,
+      "step": 3192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706375508373785e-06,
+      "loss": 0.9077,
+      "step": 3193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706141956273227e-06,
+      "loss": 0.9521,
+      "step": 3194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70590831413676e-06,
+      "loss": 0.9126,
+      "step": 3195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70567458196885e-06,
+      "loss": 0.9204,
+      "step": 3196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70544075977397e-06,
+      "loss": 0.9082,
+      "step": 3197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705206847556593e-06,
+      "loss": 0.9023,
+      "step": 3198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704972845321195e-06,
+      "loss": 0.9326,
+      "step": 3199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704738753072253e-06,
+      "loss": 0.9038,
+      "step": 3200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704504570814244e-06,
+      "loss": 0.9629,
+      "step": 3201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704270298551649e-06,
+      "loss": 0.9229,
+      "step": 3202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70403593628895e-06,
+      "loss": 0.9692,
+      "step": 3203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703801484030633e-06,
+      "loss": 0.9688,
+      "step": 3204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70356694178118e-06,
+      "loss": 0.957,
+      "step": 3205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703332309545079e-06,
+      "loss": 0.96,
+      "step": 3206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70309758732682e-06,
+      "loss": 0.936,
+      "step": 3207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702862775130893e-06,
+      "loss": 0.9507,
+      "step": 3208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70262787296179e-06,
+      "loss": 0.957,
+      "step": 3209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702392880824004e-06,
+      "loss": 0.9146,
+      "step": 3210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702157798722034e-06,
+      "loss": 0.9414,
+      "step": 3211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701922626660375e-06,
+      "loss": 0.9175,
+      "step": 3212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701687364643527e-06,
+      "loss": 0.9312,
+      "step": 3213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701452012675991e-06,
+      "loss": 0.9429,
+      "step": 3214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701216570762269e-06,
+      "loss": 0.9624,
+      "step": 3215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700981038906866e-06,
+      "loss": 0.9673,
+      "step": 3216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700745417114289e-06,
+      "loss": 0.9658,
+      "step": 3217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700509705389043e-06,
+      "loss": 0.874,
+      "step": 3218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70027390373564e-06,
+      "loss": 0.8696,
+      "step": 3219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70003801215859e-06,
+      "loss": 0.9194,
+      "step": 3220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699802030662407e-06,
+      "loss": 0.9629,
+      "step": 3221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699565959251604e-06,
+      "loss": 0.9619,
+      "step": 3222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699329797930701e-06,
+      "loss": 0.9609,
+      "step": 3223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699093546704213e-06,
+      "loss": 0.8901,
+      "step": 3224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69885720557666e-06,
+      "loss": 0.9932,
+      "step": 3225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698620774552564e-06,
+      "loss": 0.9653,
+      "step": 3226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698384253636448e-06,
+      "loss": 1.0049,
+      "step": 3227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698147642832838e-06,
+      "loss": 0.9297,
+      "step": 3228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69791094214626e-06,
+      "loss": 0.8994,
+      "step": 3229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697674151581244e-06,
+      "loss": 0.8872,
+      "step": 3230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697437271142318e-06,
+      "loss": 0.9199,
+      "step": 3231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697200300834015e-06,
+      "loss": 0.8882,
+      "step": 3232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696963240660869e-06,
+      "loss": 0.8835,
+      "step": 3233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696726090627414e-06,
+      "loss": 0.9526,
+      "step": 3234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696488850738186e-06,
+      "loss": 0.9912,
+      "step": 3235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696251520997728e-06,
+      "loss": 1.0195,
+      "step": 3236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696014101410576e-06,
+      "loss": 0.9404,
+      "step": 3237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695776591981277e-06,
+      "loss": 0.9263,
+      "step": 3238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695538992714369e-06,
+      "loss": 0.9414,
+      "step": 3239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695301303614402e-06,
+      "loss": 0.979,
+      "step": 3240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695063524685923e-06,
+      "loss": 0.896,
+      "step": 3241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694825655933479e-06,
+      "loss": 0.9121,
+      "step": 3242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694587697361624e-06,
+      "loss": 0.9521,
+      "step": 3243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694349648974908e-06,
+      "loss": 0.9448,
+      "step": 3244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694111510777885e-06,
+      "loss": 0.9814,
+      "step": 3245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693873282775112e-06,
+      "loss": 0.9067,
+      "step": 3246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693634964971146e-06,
+      "loss": 0.8967,
+      "step": 3247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69339655737055e-06,
+      "loss": 0.918,
+      "step": 3248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693158059977879e-06,
+      "loss": 0.9746,
+      "step": 3249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692919472797698e-06,
+      "loss": 0.9883,
+      "step": 3250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692680795834574e-06,
+      "loss": 0.9526,
+      "step": 3251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692442029093072e-06,
+      "loss": 0.9873,
+      "step": 3252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692203172577759e-06,
+      "loss": 0.9458,
+      "step": 3253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691964226293205e-06,
+      "loss": 0.8989,
+      "step": 3254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691725190243982e-06,
+      "loss": 0.3405,
+      "step": 3255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691486064434662e-06,
+      "loss": 0.9634,
+      "step": 3256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691246848869822e-06,
+      "loss": 0.9482,
+      "step": 3257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691007543554036e-06,
+      "loss": 0.9585,
+      "step": 3258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690768148491885e-06,
+      "loss": 0.9268,
+      "step": 3259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690528663687948e-06,
+      "loss": 0.9146,
+      "step": 3260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690289089146805e-06,
+      "loss": 0.3163,
+      "step": 3261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69004942487304e-06,
+      "loss": 0.9561,
+      "step": 3262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68980967087124e-06,
+      "loss": 0.9346,
+      "step": 3263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68956982714599e-06,
+      "loss": 0.9595,
+      "step": 3264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689329893701882e-06,
+      "loss": 0.9829,
+      "step": 3265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689089870543502e-06,
+      "loss": 0.9468,
+      "step": 3266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688849757675442e-06,
+      "loss": 0.9194,
+      "step": 3267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6886095551023e-06,
+      "loss": 0.9722,
+      "step": 3268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688369262828668e-06,
+      "loss": 0.9478,
+      "step": 3269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688128880859145e-06,
+      "loss": 0.9214,
+      "step": 3270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687888409198327e-06,
+      "loss": 0.9536,
+      "step": 3271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687647847850816e-06,
+      "loss": 0.9331,
+      "step": 3272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687407196821216e-06,
+      "loss": 0.9351,
+      "step": 3273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687166456114131e-06,
+      "loss": 0.8789,
+      "step": 3274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686925625734165e-06,
+      "loss": 0.9561,
+      "step": 3275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686684705685925e-06,
+      "loss": 0.8843,
+      "step": 3276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686443695974023e-06,
+      "loss": 0.9409,
+      "step": 3277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686202596603066e-06,
+      "loss": 0.9258,
+      "step": 3278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68596140757767e-06,
+      "loss": 0.9414,
+      "step": 3279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685720128902449e-06,
+      "loss": 0.9219,
+      "step": 3280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685478760582017e-06,
+      "loss": 0.9033,
+      "step": 3281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685237302620994e-06,
+      "loss": 0.9214,
+      "step": 3282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684995755023998e-06,
+      "loss": 0.9375,
+      "step": 3283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684754117795649e-06,
+      "loss": 0.9399,
+      "step": 3284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684512390940574e-06,
+      "loss": 0.9287,
+      "step": 3285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684270574463392e-06,
+      "loss": 0.938,
+      "step": 3286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684028668368734e-06,
+      "loss": 0.9238,
+      "step": 3287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683786672661229e-06,
+      "loss": 0.9478,
+      "step": 3288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683544587345502e-06,
+      "loss": 0.9414,
+      "step": 3289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683302412426185e-06,
+      "loss": 0.8867,
+      "step": 3290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683060147907915e-06,
+      "loss": 0.9624,
+      "step": 3291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682817793795325e-06,
+      "loss": 0.9775,
+      "step": 3292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68257535009305e-06,
+      "loss": 0.9219,
+      "step": 3293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68233281680573e-06,
+      "loss": 0.9326,
+      "step": 3294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682090193938005e-06,
+      "loss": 0.9868,
+      "step": 3295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681847481494518e-06,
+      "loss": 0.9478,
+      "step": 3296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681604679479908e-06,
+      "loss": 0.9292,
+      "step": 3297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681361787898825e-06,
+      "loss": 0.8872,
+      "step": 3298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681118806755914e-06,
+      "loss": 0.8438,
+      "step": 3299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680875736055823e-06,
+      "loss": 0.894,
+      "step": 3300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680632575803205e-06,
+      "loss": 0.957,
+      "step": 3301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680389326002708e-06,
+      "loss": 1.0049,
+      "step": 3302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680145986658989e-06,
+      "loss": 0.9521,
+      "step": 3303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679902557776702e-06,
+      "loss": 0.9375,
+      "step": 3304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679659039360506e-06,
+      "loss": 0.9829,
+      "step": 3305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679415431415056e-06,
+      "loss": 0.9243,
+      "step": 3306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679171733945014e-06,
+      "loss": 0.9443,
+      "step": 3307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678927946955046e-06,
+      "loss": 0.917,
+      "step": 3308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678684070449811e-06,
+      "loss": 0.9707,
+      "step": 3309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67844010443398e-06,
+      "loss": 0.9438,
+      "step": 3310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678196048912215e-06,
+      "loss": 0.8882,
+      "step": 3311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677951903889189e-06,
+      "loss": 0.9121,
+      "step": 3312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67770766936957e-06,
+      "loss": 0.917,
+      "step": 3313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677463345358034e-06,
+      "loss": 0.8652,
+      "step": 3314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677218931859252e-06,
+      "loss": 0.9204,
+      "step": 3315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6769744288779e-06,
+      "loss": 0.9932,
+      "step": 3316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676729836418662e-06,
+      "loss": 0.9346,
+      "step": 3317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676485154486208e-06,
+      "loss": 0.8965,
+      "step": 3318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676240383085226e-06,
+      "loss": 0.917,
+      "step": 3319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675995522220396e-06,
+      "loss": 0.9336,
+      "step": 3320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675750571896403e-06,
+      "loss": 0.9697,
+      "step": 3321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675505532117935e-06,
+      "loss": 0.959,
+      "step": 3322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675260402889677e-06,
+      "loss": 0.9668,
+      "step": 3323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675015184216321e-06,
+      "loss": 0.8975,
+      "step": 3324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674769876102557e-06,
+      "loss": 0.9927,
+      "step": 3325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674524478553077e-06,
+      "loss": 0.9536,
+      "step": 3326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67427899157258e-06,
+      "loss": 0.9473,
+      "step": 3327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67403341516576e-06,
+      "loss": 0.979,
+      "step": 3328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673787749337316e-06,
+      "loss": 1.0039,
+      "step": 3329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673541994091947e-06,
+      "loss": 0.958,
+      "step": 3330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673296149434354e-06,
+      "loss": 0.9141,
+      "step": 3331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673050215369242e-06,
+      "loss": 0.9312,
+      "step": 3332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672804191901315e-06,
+      "loss": 0.9619,
+      "step": 3333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672558079035282e-06,
+      "loss": 0.957,
+      "step": 3334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67231187677585e-06,
+      "loss": 0.3528,
+      "step": 3335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672065585127725e-06,
+      "loss": 0.9102,
+      "step": 3336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671819204095628e-06,
+      "loss": 0.9619,
+      "step": 3337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671572733684265e-06,
+      "loss": 0.9751,
+      "step": 3338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671326173898354e-06,
+      "loss": 0.3813,
+      "step": 3339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671079524742615e-06,
+      "loss": 0.9551,
+      "step": 3340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670832786221762e-06,
+      "loss": 0.9541,
+      "step": 3341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670585958340518e-06,
+      "loss": 0.9766,
+      "step": 3342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670339041103603e-06,
+      "loss": 0.9033,
+      "step": 3343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670092034515744e-06,
+      "loss": 0.9771,
+      "step": 3344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669844938581665e-06,
+      "loss": 0.9814,
+      "step": 3345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669597753306095e-06,
+      "loss": 0.8682,
+      "step": 3346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66935047869376e-06,
+      "loss": 0.8975,
+      "step": 3347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669103114749392e-06,
+      "loss": 1.0122,
+      "step": 3348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668855661477726e-06,
+      "loss": 0.9702,
+      "step": 3349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668608118883491e-06,
+      "loss": 0.957,
+      "step": 3350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668360486971428e-06,
+      "loss": 0.3065,
+      "step": 3351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668112765746271e-06,
+      "loss": 0.9341,
+      "step": 3352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667864955212762e-06,
+      "loss": 0.937,
+      "step": 3353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66761705537564e-06,
+      "loss": 0.9697,
+      "step": 3354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667369066239649e-06,
+      "loss": 0.9194,
+      "step": 3355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667120987809532e-06,
+      "loss": 0.9785,
+      "step": 3356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666872820090038e-06,
+      "loss": 0.9458,
+      "step": 3357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666624563085911e-06,
+      "loss": 0.937,
+      "step": 3358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666376216801903e-06,
+      "loss": 0.9404,
+      "step": 3359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666127781242764e-06,
+      "loss": 0.9521,
+      "step": 3360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665879256413249e-06,
+      "loss": 0.896,
+      "step": 3361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665630642318111e-06,
+      "loss": 0.9478,
+      "step": 3362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665381938962106e-06,
+      "loss": 0.9814,
+      "step": 3363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665133146349994e-06,
+      "loss": 0.9321,
+      "step": 3364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664884264486533e-06,
+      "loss": 0.897,
+      "step": 3365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664635293376485e-06,
+      "loss": 0.8921,
+      "step": 3366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664386233024614e-06,
+      "loss": 0.9341,
+      "step": 3367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664137083435682e-06,
+      "loss": 0.9097,
+      "step": 3368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663887844614458e-06,
+      "loss": 0.9224,
+      "step": 3369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663638516565712e-06,
+      "loss": 0.9131,
+      "step": 3370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663389099294211e-06,
+      "loss": 0.4298,
+      "step": 3371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663139592804728e-06,
+      "loss": 0.9561,
+      "step": 3372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662889997102039e-06,
+      "loss": 0.9644,
+      "step": 3373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662640312190914e-06,
+      "loss": 0.9194,
+      "step": 3374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662390538076132e-06,
+      "loss": 0.8735,
+      "step": 3375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662140674762473e-06,
+      "loss": 0.9238,
+      "step": 3376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661890722254717e-06,
+      "loss": 0.9575,
+      "step": 3377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661640680557644e-06,
+      "loss": 0.9722,
+      "step": 3378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661390549676039e-06,
+      "loss": 0.9297,
+      "step": 3379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661140329614688e-06,
+      "loss": 0.9492,
+      "step": 3380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660890020378378e-06,
+      "loss": 0.9585,
+      "step": 3381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660639621971894e-06,
+      "loss": 0.9351,
+      "step": 3382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660389134400034e-06,
+      "loss": 0.9487,
+      "step": 3383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660138557667583e-06,
+      "loss": 0.8608,
+      "step": 3384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65988789177934e-06,
+      "loss": 0.9067,
+      "step": 3385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659637136740096e-06,
+      "loss": 0.9688,
+      "step": 3386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659386292554654e-06,
+      "loss": 0.9609,
+      "step": 3387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659135359227807e-06,
+      "loss": 0.8892,
+      "step": 3388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658884336764362e-06,
+      "loss": 0.9023,
+      "step": 3389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658633225169116e-06,
+      "loss": 0.8872,
+      "step": 3390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658382024446875e-06,
+      "loss": 0.9189,
+      "step": 3391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658130734602446e-06,
+      "loss": 0.9751,
+      "step": 3392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657879355640636e-06,
+      "loss": 0.9443,
+      "step": 3393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657627887566253e-06,
+      "loss": 0.937,
+      "step": 3394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65737633038411e-06,
+      "loss": 0.9385,
+      "step": 3395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.657124684099018e-06,
+      "loss": 0.9619,
+      "step": 3396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656872948715792e-06,
+      "loss": 0.937,
+      "step": 3397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65662112423925e-06,
+      "loss": 0.9268,
+      "step": 3398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656369210674206e-06,
+      "loss": 0.8901,
+      "step": 3399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656117208025482e-06,
+      "loss": 0.9312,
+      "step": 3400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655865116297898e-06,
+      "loss": 0.9375,
+      "step": 3401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655612935496278e-06,
+      "loss": 0.9346,
+      "step": 3402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655360665625446e-06,
+      "loss": 0.9253,
+      "step": 3403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655108306690229e-06,
+      "loss": 0.9746,
+      "step": 3404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654855858695456e-06,
+      "loss": 0.8511,
+      "step": 3405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654603321645953e-06,
+      "loss": 0.8853,
+      "step": 3406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654350695546555e-06,
+      "loss": 0.9985,
+      "step": 3407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654097980402093e-06,
+      "loss": 0.897,
+      "step": 3408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653845176217402e-06,
+      "loss": 0.9189,
+      "step": 3409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653592282997321e-06,
+      "loss": 0.9487,
+      "step": 3410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653339300746686e-06,
+      "loss": 0.2859,
+      "step": 3411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653086229470337e-06,
+      "loss": 0.9199,
+      "step": 3412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652833069173115e-06,
+      "loss": 0.9521,
+      "step": 3413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652579819859867e-06,
+      "loss": 0.9009,
+      "step": 3414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652326481535434e-06,
+      "loss": 0.8882,
+      "step": 3415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652073054204665e-06,
+      "loss": 0.8633,
+      "step": 3416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651819537872406e-06,
+      "loss": 0.9404,
+      "step": 3417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65156593254351e-06,
+      "loss": 0.979,
+      "step": 3418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651312238222827e-06,
+      "loss": 0.8833,
+      "step": 3419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651058454915214e-06,
+      "loss": 0.896,
+      "step": 3420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650804582625521e-06,
+      "loss": 0.9507,
+      "step": 3421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650550621358608e-06,
+      "loss": 0.9741,
+      "step": 3422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650296571119334e-06,
+      "loss": 0.9072,
+      "step": 3423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650042431912558e-06,
+      "loss": 0.936,
+      "step": 3424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649788203743143e-06,
+      "loss": 0.9551,
+      "step": 3425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649533886615952e-06,
+      "loss": 0.96,
+      "step": 3426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649279480535852e-06,
+      "loss": 0.9487,
+      "step": 3427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649024985507709e-06,
+      "loss": 0.9844,
+      "step": 3428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648770401536392e-06,
+      "loss": 0.9014,
+      "step": 3429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64851572862677e-06,
+      "loss": 0.9644,
+      "step": 3430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648260966783719e-06,
+      "loss": 0.9604,
+      "step": 3431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64800611601211e-06,
+      "loss": 0.915,
+      "step": 3432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64775117631682e-06,
+      "loss": 0.8809,
+      "step": 3433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647496147702727e-06,
+      "loss": 0.8423,
+      "step": 3434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647241030174707e-06,
+      "loss": 0.9292,
+      "step": 3435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646985823737644e-06,
+      "loss": 0.9629,
+      "step": 3436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646730528396419e-06,
+      "loss": 0.8833,
+      "step": 3437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646475144155917e-06,
+      "loss": 0.8804,
+      "step": 3438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646219671021025e-06,
+      "loss": 0.9238,
+      "step": 3439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645964108996629e-06,
+      "loss": 0.9341,
+      "step": 3440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645708458087615e-06,
+      "loss": 0.9883,
+      "step": 3441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64545271829888e-06,
+      "loss": 0.9321,
+      "step": 3442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645196889635316e-06,
+      "loss": 0.9688,
+      "step": 3443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644940972101813e-06,
+      "loss": 0.9482,
+      "step": 3444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64468496570327e-06,
+      "loss": 0.8975,
+      "step": 3445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644428870444584e-06,
+      "loss": 0.9463,
+      "step": 3446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644172686330656e-06,
+      "loss": 0.9771,
+      "step": 3447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643916413366386e-06,
+      "loss": 0.9663,
+      "step": 3448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643660051556677e-06,
+      "loss": 0.9062,
+      "step": 3449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643403600906433e-06,
+      "loss": 0.9453,
+      "step": 3450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643147061420562e-06,
+      "loss": 0.8838,
+      "step": 3451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642890433103972e-06,
+      "loss": 0.9316,
+      "step": 3452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64263371596157e-06,
+      "loss": 0.9863,
+      "step": 3453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64237690999827e-06,
+      "loss": 0.9482,
+      "step": 3454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642120015218983e-06,
+      "loss": 1.0015,
+      "step": 3455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641863031628627e-06,
+      "loss": 0.9365,
+      "step": 3456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641605959232115e-06,
+      "loss": 0.8994,
+      "step": 3457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641348798034366e-06,
+      "loss": 0.9087,
+      "step": 3458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641091548040304e-06,
+      "loss": 0.9204,
+      "step": 3459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640834209254845e-06,
+      "loss": 0.9634,
+      "step": 3460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640576781682914e-06,
+      "loss": 0.9565,
+      "step": 3461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640319265329437e-06,
+      "loss": 0.937,
+      "step": 3462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64006166019934e-06,
+      "loss": 0.9395,
+      "step": 3463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639803966297551e-06,
+      "loss": 0.9502,
+      "step": 3464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639546183629003e-06,
+      "loss": 0.8633,
+      "step": 3465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639288312198624e-06,
+      "loss": 0.938,
+      "step": 3466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639030352011349e-06,
+      "loss": 0.9209,
+      "step": 3467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638772303072115e-06,
+      "loss": 1.0122,
+      "step": 3468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638514165385855e-06,
+      "loss": 0.9517,
+      "step": 3469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638255938957509e-06,
+      "loss": 0.9312,
+      "step": 3470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637997623792018e-06,
+      "loss": 0.396,
+      "step": 3471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637739219894326e-06,
+      "loss": 0.9243,
+      "step": 3472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637480727269374e-06,
+      "loss": 0.9595,
+      "step": 3473
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637222145922106e-06,
+      "loss": 0.9741,
+      "step": 3474
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636963475857473e-06,
+      "loss": 0.9399,
+      "step": 3475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63670471708042e-06,
+      "loss": 0.9077,
+      "step": 3476
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6364458695959e-06,
+      "loss": 0.9287,
+      "step": 3477
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636186933408867e-06,
+      "loss": 0.9458,
+      "step": 3478
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635927908524269e-06,
+      "loss": 0.9043,
+      "step": 3479
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635668794947066e-06,
+      "loss": 0.9146,
+      "step": 3480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635409592682213e-06,
+      "loss": 0.9678,
+      "step": 3481
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635150301734671e-06,
+      "loss": 0.9365,
+      "step": 3482
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634890922109398e-06,
+      "loss": 0.8652,
+      "step": 3483
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63463145381136e-06,
+      "loss": 0.9126,
+      "step": 3484
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63437189684552e-06,
+      "loss": 0.8916,
+      "step": 3485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63411225121684e-06,
+      "loss": 0.9233,
+      "step": 3486
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63385251693029e-06,
+      "loss": 0.9546,
+      "step": 3487
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633592693990843e-06,
+      "loss": 0.917,
+      "step": 3488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633332782403463e-06,
+      "loss": 0.9482,
+      "step": 3489
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633072782173127e-06,
+      "loss": 0.9458,
+      "step": 3490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632812693304807e-06,
+      "loss": 0.9482,
+      "step": 3491
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63255251580348e-06,
+      "loss": 0.8633,
+      "step": 3492
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632292249674122e-06,
+      "loss": 0.9438,
+      "step": 3493
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632031894921717e-06,
+      "loss": 0.9019,
+      "step": 3494
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63177145155124e-06,
+      "loss": 0.8755,
+      "step": 3495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631510919567677e-06,
+      "loss": 0.9424,
+      "step": 3496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631250298976012e-06,
+      "loss": 0.936,
+      "step": 3497
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63098958978123e-06,
+      "loss": 0.8936,
+      "step": 3498
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63072879198832e-06,
+      "loss": 0.938,
+      "step": 3499
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630467905602272e-06,
+      "loss": 0.9126,
+      "step": 3500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630206930628077e-06,
+      "loss": 0.9336,
+      "step": 3501
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629945867070725e-06,
+      "loss": 0.9707,
+      "step": 3502
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629684714935213e-06,
+      "loss": 0.9253,
+      "step": 3503
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629423474226538e-06,
+      "loss": 0.9443,
+      "step": 3504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629162144949696e-06,
+      "loss": 1.0137,
+      "step": 3505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628900727109687e-06,
+      "loss": 0.9219,
+      "step": 3506
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628639220711514e-06,
+      "loss": 0.9448,
+      "step": 3507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628377625760179e-06,
+      "loss": 0.8848,
+      "step": 3508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628115942260685e-06,
+      "loss": 0.9253,
+      "step": 3509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627854170218043e-06,
+      "loss": 0.9849,
+      "step": 3510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627592309637255e-06,
+      "loss": 0.9443,
+      "step": 3511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627330360523335e-06,
+      "loss": 1.0127,
+      "step": 3512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627068322881294e-06,
+      "loss": 0.9331,
+      "step": 3513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626806196716145e-06,
+      "loss": 0.9312,
+      "step": 3514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6265439820329e-06,
+      "loss": 0.9502,
+      "step": 3515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626281678836581e-06,
+      "loss": 0.9136,
+      "step": 3516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626019287132202e-06,
+      "loss": 0.9111,
+      "step": 3517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625756806924783e-06,
+      "loss": 0.9126,
+      "step": 3518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625494238219351e-06,
+      "loss": 0.9478,
+      "step": 3519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625231581020922e-06,
+      "loss": 0.9692,
+      "step": 3520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624968835334525e-06,
+      "loss": 0.9272,
+      "step": 3521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624706001165188e-06,
+      "loss": 0.9229,
+      "step": 3522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624443078517935e-06,
+      "loss": 0.9058,
+      "step": 3523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624180067397799e-06,
+      "loss": 0.9229,
+      "step": 3524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623916967809813e-06,
+      "loss": 0.96,
+      "step": 3525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623653779759008e-06,
+      "loss": 0.9248,
+      "step": 3526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62339050325042e-06,
+      "loss": 0.9614,
+      "step": 3527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623127138289087e-06,
+      "loss": 0.9434,
+      "step": 3528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622863684880046e-06,
+      "loss": 0.9653,
+      "step": 3529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622600143028339e-06,
+      "loss": 0.8716,
+      "step": 3530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622336512739005e-06,
+      "loss": 0.9268,
+      "step": 3531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622072794017091e-06,
+      "loss": 0.9624,
+      "step": 3532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621808986867643e-06,
+      "loss": 0.9038,
+      "step": 3533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621545091295704e-06,
+      "loss": 0.9282,
+      "step": 3534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621281107306323e-06,
+      "loss": 0.915,
+      "step": 3535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621017034904554e-06,
+      "loss": 0.5055,
+      "step": 3536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620752874095446e-06,
+      "loss": 0.9268,
+      "step": 3537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620488624884057e-06,
+      "loss": 0.959,
+      "step": 3538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620224287275438e-06,
+      "loss": 0.9258,
+      "step": 3539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619959861274646e-06,
+      "loss": 0.9541,
+      "step": 3540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619695346886743e-06,
+      "loss": 0.9312,
+      "step": 3541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61943074411679e-06,
+      "loss": 0.9893,
+      "step": 3542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619166052969845e-06,
+      "loss": 0.9629,
+      "step": 3543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618901273450977e-06,
+      "loss": 0.9077,
+      "step": 3544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618636405565247e-06,
+      "loss": 0.9027,
+      "step": 3545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618371449317726e-06,
+      "loss": 0.9214,
+      "step": 3546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61810640471348e-06,
+      "loss": 0.9268,
+      "step": 3547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617841271757582e-06,
+      "loss": 0.9204,
+      "step": 3548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617576050455103e-06,
+      "loss": 0.9463,
+      "step": 3549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61731074081112e-06,
+      "loss": 0.9414,
+      "step": 3550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617045342830705e-06,
+      "loss": 0.9023,
+      "step": 3551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616779856518937e-06,
+      "loss": 0.8739,
+      "step": 3552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616514281880896e-06,
+      "loss": 0.9443,
+      "step": 3553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616248618921661e-06,
+      "loss": 0.9771,
+      "step": 3554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615982867646315e-06,
+      "loss": 0.9126,
+      "step": 3555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615717028059945e-06,
+      "loss": 0.9048,
+      "step": 3556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615451100167633e-06,
+      "loss": 0.9531,
+      "step": 3557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61518508397447e-06,
+      "loss": 0.9355,
+      "step": 3558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614918979485542e-06,
+      "loss": 0.2769,
+      "step": 3559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614652786705943e-06,
+      "loss": 0.8931,
+      "step": 3560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614386505640762e-06,
+      "loss": 0.9038,
+      "step": 3561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614120136295099e-06,
+      "loss": 1.0146,
+      "step": 3562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613853678674045e-06,
+      "loss": 0.8955,
+      "step": 3563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613587132782701e-06,
+      "loss": 0.8604,
+      "step": 3564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613320498626164e-06,
+      "loss": 0.9341,
+      "step": 3565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613053776209535e-06,
+      "loss": 0.9468,
+      "step": 3566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61278696553792e-06,
+      "loss": 0.9551,
+      "step": 3567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61252006661642e-06,
+      "loss": 1.0117,
+      "step": 3568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612253079450145e-06,
+      "loss": 0.9937,
+      "step": 3569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611986004044198e-06,
+      "loss": 0.8765,
+      "step": 3570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611718840403692e-06,
+      "loss": 0.9336,
+      "step": 3571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611451588533737e-06,
+      "loss": 0.9971,
+      "step": 3572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611184248439447e-06,
+      "loss": 0.8535,
+      "step": 3573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610916820125938e-06,
+      "loss": 0.9136,
+      "step": 3574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610649303598321e-06,
+      "loss": 0.8691,
+      "step": 3575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61038169886172e-06,
+      "loss": 0.9302,
+      "step": 3576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61011400592125e-06,
+      "loss": 0.9258,
+      "step": 3577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609846224782036e-06,
+      "loss": 0.9146,
+      "step": 3578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6095783554492e-06,
+      "loss": 0.9463,
+      "step": 3579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609310397927863e-06,
+      "loss": 0.9551,
+      "step": 3580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609042352223156e-06,
+      "loss": 0.9395,
+      "step": 3581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608774218340207e-06,
+      "loss": 0.915,
+      "step": 3582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608505996284144e-06,
+      "loss": 0.9165,
+      "step": 3583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608237686060099e-06,
+      "loss": 0.8906,
+      "step": 3584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607969287673206e-06,
+      "loss": 0.8865,
+      "step": 3585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6077008011286e-06,
+      "loss": 0.9658,
+      "step": 3586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607432226431415e-06,
+      "loss": 0.9448,
+      "step": 3587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607163563586792e-06,
+      "loss": 0.9507,
+      "step": 3588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60689481259987e-06,
+      "loss": 0.9219,
+      "step": 3589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60662597347579e-06,
+      "loss": 0.9277,
+      "step": 3590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606357046219697e-06,
+      "loss": 0.42,
+      "step": 3591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606088030836736e-06,
+      "loss": 0.9219,
+      "step": 3592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60581892733205e-06,
+      "loss": 0.9312,
+      "step": 3593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605549735710792e-06,
+      "loss": 0.9302,
+      "step": 3594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60528045597811e-06,
+      "loss": 0.9438,
+      "step": 3595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605011088139154e-06,
+      "loss": 0.9502,
+      "step": 3596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604741632199081e-06,
+      "loss": 0.9189,
+      "step": 3597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604472088163044e-06,
+      "loss": 0.9658,
+      "step": 3598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6042024560362e-06,
+      "loss": 0.9287,
+      "step": 3599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603932735823709e-06,
+      "loss": 0.8662,
+      "step": 3600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603662927530729e-06,
+      "loss": 0.8599,
+      "step": 3601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603393031162423e-06,
+      "loss": 0.9463,
+      "step": 3602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603123046723953e-06,
+      "loss": 0.9517,
+      "step": 3603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602852974220485e-06,
+      "loss": 0.9229,
+      "step": 3604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602582813657189e-06,
+      "loss": 0.9546,
+      "step": 3605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602312565039229e-06,
+      "loss": 0.9346,
+      "step": 3606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602042228371777e-06,
+      "loss": 1.0205,
+      "step": 3607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601771803660008e-06,
+      "loss": 0.8989,
+      "step": 3608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60150129090909e-06,
+      "loss": 0.9585,
+      "step": 3609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601230690124203e-06,
+      "loss": 0.9326,
+      "step": 3610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600960001310522e-06,
+      "loss": 0.979,
+      "step": 3611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600689224473226e-06,
+      "loss": 0.9102,
+      "step": 3612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600418359617494e-06,
+      "loss": 0.9014,
+      "step": 3613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60014740674851e-06,
+      "loss": 0.9226,
+      "step": 3614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599876365871459e-06,
+      "loss": 0.3313,
+      "step": 3615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599605236991524e-06,
+      "loss": 0.9355,
+      "step": 3616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599334020113891e-06,
+      "loss": 0.9595,
+      "step": 3617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59906271524375e-06,
+      "loss": 0.9331,
+      "step": 3618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598791322386294e-06,
+      "loss": 0.9541,
+      "step": 3619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598519841546713e-06,
+      "loss": 0.9556,
+      "step": 3620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598248272730201e-06,
+      "loss": 0.9146,
+      "step": 3621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597976615941952e-06,
+      "loss": 1.0225,
+      "step": 3622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597704871187168e-06,
+      "loss": 0.9585,
+      "step": 3623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597433038471043e-06,
+      "loss": 0.9468,
+      "step": 3624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59716111779878e-06,
+      "loss": 0.9644,
+      "step": 3625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596889109175577e-06,
+      "loss": 0.9272,
+      "step": 3626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596617012606646e-06,
+      "loss": 0.9736,
+      "step": 3627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596344828097186e-06,
+      "loss": 0.9888,
+      "step": 3628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596072555652407e-06,
+      "loss": 0.894,
+      "step": 3629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595800195277519e-06,
+      "loss": 0.9351,
+      "step": 3630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595527746977728e-06,
+      "loss": 0.8296,
+      "step": 3631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595255210758251e-06,
+      "loss": 0.8662,
+      "step": 3632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594982586624302e-06,
+      "loss": 0.98,
+      "step": 3633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594709874581094e-06,
+      "loss": 0.8862,
+      "step": 3634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594437074633847e-06,
+      "loss": 0.9375,
+      "step": 3635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594164186787777e-06,
+      "loss": 0.9731,
+      "step": 3636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593891211048108e-06,
+      "loss": 0.936,
+      "step": 3637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59361814742006e-06,
+      "loss": 0.8887,
+      "step": 3638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59334499590886e-06,
+      "loss": 0.9272,
+      "step": 3639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59307175651973e-06,
+      "loss": 0.9272,
+      "step": 3640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592798429257899e-06,
+      "loss": 0.9536,
+      "step": 3641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592525014128598e-06,
+      "loss": 0.9136,
+      "step": 3642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592251511137057e-06,
+      "loss": 0.9146,
+      "step": 3643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591977920288508e-06,
+      "loss": 0.8828,
+      "step": 3644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591704241588185e-06,
+      "loss": 0.9551,
+      "step": 3645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591430475041323e-06,
+      "loss": 0.915,
+      "step": 3646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591156620653162e-06,
+      "loss": 0.8862,
+      "step": 3647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590882678428941e-06,
+      "loss": 0.8906,
+      "step": 3648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590608648373898e-06,
+      "loss": 0.9873,
+      "step": 3649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590334530493279e-06,
+      "loss": 0.8857,
+      "step": 3650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590060324792328e-06,
+      "loss": 0.8838,
+      "step": 3651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589786031276288e-06,
+      "loss": 0.3115,
+      "step": 3652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589511649950409e-06,
+      "loss": 0.9399,
+      "step": 3653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58923718081994e-06,
+      "loss": 0.9678,
+      "step": 3654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588962623890132e-06,
+      "loss": 0.9619,
+      "step": 3655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588687979166238e-06,
+      "loss": 0.9272,
+      "step": 3656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588413246653513e-06,
+      "loss": 0.9565,
+      "step": 3657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588138426357212e-06,
+      "loss": 0.9678,
+      "step": 3658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587863518282592e-06,
+      "loss": 0.8916,
+      "step": 3659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587588522434912e-06,
+      "loss": 0.9673,
+      "step": 3660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587313438819436e-06,
+      "loss": 0.8633,
+      "step": 3661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587038267441424e-06,
+      "loss": 0.9243,
+      "step": 3662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586763008306144e-06,
+      "loss": 0.821,
+      "step": 3663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586487661418857e-06,
+      "loss": 0.8853,
+      "step": 3664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586212226784836e-06,
+      "loss": 0.9414,
+      "step": 3665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585936704409346e-06,
+      "loss": 0.9443,
+      "step": 3666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585661094297663e-06,
+      "loss": 0.9219,
+      "step": 3667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585385396455055e-06,
+      "loss": 0.9292,
+      "step": 3668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585109610886801e-06,
+      "loss": 0.9209,
+      "step": 3669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584833737598171e-06,
+      "loss": 0.9175,
+      "step": 3670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58455777659445e-06,
+      "loss": 0.9751,
+      "step": 3671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584281727880913e-06,
+      "loss": 0.9092,
+      "step": 3672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584005591462843e-06,
+      "loss": 0.9771,
+      "step": 3673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583729367345523e-06,
+      "loss": 0.9438,
+      "step": 3674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583453055534238e-06,
+      "loss": 0.9268,
+      "step": 3675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583176656034273e-06,
+      "loss": 0.939,
+      "step": 3676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582900168850913e-06,
+      "loss": 0.3536,
+      "step": 3677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582623593989455e-06,
+      "loss": 0.9355,
+      "step": 3678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582346931455183e-06,
+      "loss": 0.9722,
+      "step": 3679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582070181253396e-06,
+      "loss": 1.0459,
+      "step": 3680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581793343389386e-06,
+      "loss": 0.959,
+      "step": 3681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581516417868448e-06,
+      "loss": 0.9604,
+      "step": 3682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581239404695882e-06,
+      "loss": 0.9458,
+      "step": 3683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580962303876988e-06,
+      "loss": 0.9429,
+      "step": 3684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580685115417065e-06,
+      "loss": 0.8989,
+      "step": 3685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580407839321417e-06,
+      "loss": 0.3254,
+      "step": 3686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580130475595351e-06,
+      "loss": 0.9502,
+      "step": 3687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57985302424417e-06,
+      "loss": 0.9141,
+      "step": 3688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579575485273186e-06,
+      "loss": 0.8892,
+      "step": 3689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579297858687704e-06,
+      "loss": 0.9751,
+      "step": 3690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579020144493039e-06,
+      "loss": 0.9395,
+      "step": 3691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578742342694503e-06,
+      "loss": 0.3678,
+      "step": 3692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57846445329741e-06,
+      "loss": 0.9253,
+      "step": 3693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578186476307078e-06,
+      "loss": 0.9272,
+      "step": 3694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577908411728825e-06,
+      "loss": 0.8691,
+      "step": 3695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577630259567968e-06,
+      "loss": 0.9731,
+      "step": 3696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577352019829834e-06,
+      "loss": 0.8896,
+      "step": 3697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577073692519738e-06,
+      "loss": 1.001,
+      "step": 3698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576795277643013e-06,
+      "loss": 0.8755,
+      "step": 3699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576516775204983e-06,
+      "loss": 0.9395,
+      "step": 3700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576238185210973e-06,
+      "loss": 0.3306,
+      "step": 3701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575959507666318e-06,
+      "loss": 0.9009,
+      "step": 3702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575680742576344e-06,
+      "loss": 0.9053,
+      "step": 3703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575401889946388e-06,
+      "loss": 0.9585,
+      "step": 3704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575122949781785e-06,
+      "loss": 0.915,
+      "step": 3705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574843922087869e-06,
+      "loss": 0.9312,
+      "step": 3706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574564806869982e-06,
+      "loss": 0.8818,
+      "step": 3707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57428560413346e-06,
+      "loss": 0.9888,
+      "step": 3708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574006313883646e-06,
+      "loss": 0.9092,
+      "step": 3709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573726936125882e-06,
+      "loss": 0.958,
+      "step": 3710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573447470865516e-06,
+      "loss": 0.958,
+      "step": 3711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573167918107893e-06,
+      "loss": 0.939,
+      "step": 3712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572888277858362e-06,
+      "loss": 0.9468,
+      "step": 3713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572608550122272e-06,
+      "loss": 0.9858,
+      "step": 3714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572328734904973e-06,
+      "loss": 0.9312,
+      "step": 3715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572048832211823e-06,
+      "loss": 0.939,
+      "step": 3716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571768842048173e-06,
+      "loss": 0.9438,
+      "step": 3717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571488764419381e-06,
+      "loss": 0.9028,
+      "step": 3718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571208599330806e-06,
+      "loss": 0.9268,
+      "step": 3719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570928346787804e-06,
+      "loss": 0.9595,
+      "step": 3720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570648006795744e-06,
+      "loss": 0.9111,
+      "step": 3721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570367579359982e-06,
+      "loss": 0.9629,
+      "step": 3722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570087064485888e-06,
+      "loss": 0.9111,
+      "step": 3723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569806462178826e-06,
+      "loss": 0.9263,
+      "step": 3724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569525772444167e-06,
+      "loss": 0.9556,
+      "step": 3725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569244995287278e-06,
+      "loss": 0.8926,
+      "step": 3726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56896413071353e-06,
+      "loss": 0.9468,
+      "step": 3727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568683178728301e-06,
+      "loss": 0.9448,
+      "step": 3728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568402139336962e-06,
+      "loss": 0.9209,
+      "step": 3729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568121012544892e-06,
+      "loss": 0.9043,
+      "step": 3730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567839798357467e-06,
+      "loss": 0.9429,
+      "step": 3731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567558496780071e-06,
+      "loss": 0.9688,
+      "step": 3732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567277107818083e-06,
+      "loss": 0.9126,
+      "step": 3733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566995631476887e-06,
+      "loss": 0.9204,
+      "step": 3734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566714067761866e-06,
+      "loss": 0.938,
+      "step": 3735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56643241667841e-06,
+      "loss": 0.9756,
+      "step": 3736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566150678231907e-06,
+      "loss": 0.876,
+      "step": 3737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565868852427744e-06,
+      "loss": 0.3558,
+      "step": 3738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565586939271317e-06,
+      "loss": 0.9658,
+      "step": 3739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565304938768015e-06,
+      "loss": 0.9038,
+      "step": 3740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565022850923237e-06,
+      "loss": 0.9678,
+      "step": 3741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564740675742378e-06,
+      "loss": 0.918,
+      "step": 3742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564458413230837e-06,
+      "loss": 0.9565,
+      "step": 3743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564176063394013e-06,
+      "loss": 0.8799,
+      "step": 3744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56389362623731e-06,
+      "loss": 0.9375,
+      "step": 3745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56361110176613e-06,
+      "loss": 0.9116,
+      "step": 3746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563328489985877e-06,
+      "loss": 0.9238,
+      "step": 3747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563045790901959e-06,
+      "loss": 0.8555,
+      "step": 3748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562763004519786e-06,
+      "loss": 0.9219,
+      "step": 3749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562480130844766e-06,
+      "loss": 0.9009,
+      "step": 3750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562197169882312e-06,
+      "loss": 0.9302,
+      "step": 3751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561914121637837e-06,
+      "loss": 0.9214,
+      "step": 3752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561630986116754e-06,
+      "loss": 0.9038,
+      "step": 3753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561347763324484e-06,
+      "loss": 0.8975,
+      "step": 3754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561064453266445e-06,
+      "loss": 0.8716,
+      "step": 3755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560781055948052e-06,
+      "loss": 0.9297,
+      "step": 3756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560497571374734e-06,
+      "loss": 0.9634,
+      "step": 3757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560213999551908e-06,
+      "loss": 0.9243,
+      "step": 3758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559930340485004e-06,
+      "loss": 0.9194,
+      "step": 3759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559646594179447e-06,
+      "loss": 0.9917,
+      "step": 3760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559362760640665e-06,
+      "loss": 0.9136,
+      "step": 3761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55907883987409e-06,
+      "loss": 0.9795,
+      "step": 3762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55879483188515e-06,
+      "loss": 0.9346,
+      "step": 3763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558510736679283e-06,
+      "loss": 0.8896,
+      "step": 3764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558226554261922e-06,
+      "loss": 0.9277,
+      "step": 3765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557942284638504e-06,
+      "loss": 0.9883,
+      "step": 3766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557657927814468e-06,
+      "loss": 0.9707,
+      "step": 3767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557373483795254e-06,
+      "loss": 0.9092,
+      "step": 3768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557088952586303e-06,
+      "loss": 0.8594,
+      "step": 3769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556804334193059e-06,
+      "loss": 0.9946,
+      "step": 3770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556519628620968e-06,
+      "loss": 0.9175,
+      "step": 3771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556234835875477e-06,
+      "loss": 0.8525,
+      "step": 3772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555949955962033e-06,
+      "loss": 0.8804,
+      "step": 3773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555664988886086e-06,
+      "loss": 0.8872,
+      "step": 3774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55537993465309e-06,
+      "loss": 0.8892,
+      "step": 3775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555094793268498e-06,
+      "loss": 0.9175,
+      "step": 3776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554809564737764e-06,
+      "loss": 1.0142,
+      "step": 3777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554524249066345e-06,
+      "loss": 0.9189,
+      "step": 3778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554238846259702e-06,
+      "loss": 0.9351,
+      "step": 3779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553953356323293e-06,
+      "loss": 0.9199,
+      "step": 3780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55366777926258e-06,
+      "loss": 0.936,
+      "step": 3781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553382115083026e-06,
+      "loss": 0.9897,
+      "step": 3782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553096363790096e-06,
+      "loss": 0.8989,
+      "step": 3783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552810525389258e-06,
+      "loss": 0.9404,
+      "step": 3784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552524599885982e-06,
+      "loss": 0.9761,
+      "step": 3785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552238587285736e-06,
+      "loss": 0.9292,
+      "step": 3786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551952487593992e-06,
+      "loss": 0.9385,
+      "step": 3787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551666300816222e-06,
+      "loss": 0.8818,
+      "step": 3788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551380026957906e-06,
+      "loss": 0.9561,
+      "step": 3789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551093666024519e-06,
+      "loss": 0.3032,
+      "step": 3790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550807218021533e-06,
+      "loss": 1.0024,
+      "step": 3791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550520682954438e-06,
+      "loss": 0.9019,
+      "step": 3792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55023406082871e-06,
+      "loss": 0.9224,
+      "step": 3793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549947351649838e-06,
+      "loss": 0.8354,
+      "step": 3794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549660555423298e-06,
+      "loss": 0.9858,
+      "step": 3795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549373672154586e-06,
+      "loss": 0.9448,
+      "step": 3796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549086701849186e-06,
+      "loss": 0.9038,
+      "step": 3797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548799644512587e-06,
+      "loss": 0.9634,
+      "step": 3798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548512500150285e-06,
+      "loss": 0.9512,
+      "step": 3799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54822526876777e-06,
+      "loss": 0.9419,
+      "step": 3800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547937950370538e-06,
+      "loss": 0.9468,
+      "step": 3801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547650544964086e-06,
+      "loss": 0.9336,
+      "step": 3802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547363052553915e-06,
+      "loss": 0.9497,
+      "step": 3803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547075473145522e-06,
+      "loss": 0.9199,
+      "step": 3804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54678780674441e-06,
+      "loss": 0.9668,
+      "step": 3805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546500053356082e-06,
+      "loss": 0.8643,
+      "step": 3806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546212212986045e-06,
+      "loss": 0.8789,
+      "step": 3807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545924285639803e-06,
+      "loss": 0.8931,
+      "step": 3808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545636271322867e-06,
+      "loss": 0.9292,
+      "step": 3809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545348170040744e-06,
+      "loss": 0.9536,
+      "step": 3810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54505998179895e-06,
+      "loss": 0.9521,
+      "step": 3811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544771706602995e-06,
+      "loss": 0.9521,
+      "step": 3812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544483344458398e-06,
+      "loss": 0.8975,
+      "step": 3813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54419489537067e-06,
+      "loss": 0.9946,
+      "step": 3814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543906359345335e-06,
+      "loss": 0.981,
+      "step": 3815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543617736387911e-06,
+      "loss": 0.9507,
+      "step": 3816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54332902650392e-06,
+      "loss": 0.897,
+      "step": 3817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543040229698885e-06,
+      "loss": 0.8721,
+      "step": 3818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542751345978331e-06,
+      "loss": 0.9404,
+      "step": 3819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542462375347785e-06,
+      "loss": 0.959,
+      "step": 3820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542173317812777e-06,
+      "loss": 0.9292,
+      "step": 3821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541884173378834e-06,
+      "loss": 0.9868,
+      "step": 3822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541594942051493e-06,
+      "loss": 0.916,
+      "step": 3823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541305623836281e-06,
+      "loss": 0.9312,
+      "step": 3824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541016218738736e-06,
+      "loss": 0.8916,
+      "step": 3825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540726726764397e-06,
+      "loss": 0.9702,
+      "step": 3826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540437147918799e-06,
+      "loss": 0.915,
+      "step": 3827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540147482207485e-06,
+      "loss": 0.8787,
+      "step": 3828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539857729635994e-06,
+      "loss": 0.9272,
+      "step": 3829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539567890209871e-06,
+      "loss": 0.9336,
+      "step": 3830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53927796393466e-06,
+      "loss": 0.9541,
+      "step": 3831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53898795081591e-06,
+      "loss": 0.9185,
+      "step": 3832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538697850859167e-06,
+      "loss": 0.9683,
+      "step": 3833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538407664069983e-06,
+      "loss": 0.9727,
+      "step": 3834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538117390453909e-06,
+      "loss": 0.9277,
+      "step": 3835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537827030016496e-06,
+      "loss": 1.0303,
+      "step": 3836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537536582763303e-06,
+      "loss": 0.353,
+      "step": 3837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537246048699885e-06,
+      "loss": 0.9316,
+      "step": 3838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536955427831797e-06,
+      "loss": 0.9512,
+      "step": 3839
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536664720164606e-06,
+      "loss": 0.9556,
+      "step": 3840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536373925703868e-06,
+      "loss": 0.9395,
+      "step": 3841
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53608304445515e-06,
+      "loss": 0.9771,
+      "step": 3842
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535792076424016e-06,
+      "loss": 0.9185,
+      "step": 3843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53550102161603e-06,
+      "loss": 0.8799,
+      "step": 3844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535209880036762e-06,
+      "loss": 0.8955,
+      "step": 3845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534918651691784e-06,
+      "loss": 0.8945,
+      "step": 3846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534627336586665e-06,
+      "loss": 0.9517,
+      "step": 3847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534335934726982e-06,
+      "loss": 0.8843,
+      "step": 3848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534044446118305e-06,
+      "loss": 0.9492,
+      "step": 3849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533752870766211e-06,
+      "loss": 0.856,
+      "step": 3850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533461208676283e-06,
+      "loss": 0.9634,
+      "step": 3851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5331694598541e-06,
+      "loss": 0.9229,
+      "step": 3852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532877624305238e-06,
+      "loss": 0.918,
+      "step": 3853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532585702035286e-06,
+      "loss": 0.9214,
+      "step": 3854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532293693049826e-06,
+      "loss": 0.9556,
+      "step": 3855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532001597354447e-06,
+      "loss": 0.8989,
+      "step": 3856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531709414954736e-06,
+      "loss": 0.9443,
+      "step": 3857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531417145856281e-06,
+      "loss": 0.338,
+      "step": 3858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531124790064676e-06,
+      "loss": 0.9355,
+      "step": 3859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530832347585514e-06,
+      "loss": 0.8599,
+      "step": 3860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53053981842439e-06,
+      "loss": 0.9512,
+      "step": 3861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5302472025869e-06,
+      "loss": 0.9395,
+      "step": 3862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.529954500078643e-06,
+      "loss": 0.9014,
+      "step": 3863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529661710905216e-06,
+      "loss": 0.9243,
+      "step": 3864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529368835072223e-06,
+      "loss": 0.3113,
+      "step": 3865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529075872585266e-06,
+      "loss": 0.9214,
+      "step": 3866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528782823449954e-06,
+      "loss": 0.9375,
+      "step": 3867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528489687671888e-06,
+      "loss": 0.8677,
+      "step": 3868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528196465256679e-06,
+      "loss": 0.917,
+      "step": 3869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527903156209934e-06,
+      "loss": 0.9194,
+      "step": 3870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527609760537268e-06,
+      "loss": 0.895,
+      "step": 3871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527316278244292e-06,
+      "loss": 0.9199,
+      "step": 3872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527022709336622e-06,
+      "loss": 0.873,
+      "step": 3873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526729053819875e-06,
+      "loss": 0.9111,
+      "step": 3874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526435311699666e-06,
+      "loss": 0.9492,
+      "step": 3875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526141482981618e-06,
+      "loss": 0.8589,
+      "step": 3876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525847567671352e-06,
+      "loss": 0.9004,
+      "step": 3877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525553565774489e-06,
+      "loss": 0.9487,
+      "step": 3878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525259477296653e-06,
+      "loss": 0.917,
+      "step": 3879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524965302243474e-06,
+      "loss": 0.3521,
+      "step": 3880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524671040620579e-06,
+      "loss": 0.9004,
+      "step": 3881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524376692433597e-06,
+      "loss": 0.8994,
+      "step": 3882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524082257688158e-06,
+      "loss": 0.9009,
+      "step": 3883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523787736389897e-06,
+      "loss": 0.9663,
+      "step": 3884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523493128544448e-06,
+      "loss": 0.9331,
+      "step": 3885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523198434157447e-06,
+      "loss": 0.8921,
+      "step": 3886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522903653234533e-06,
+      "loss": 0.9424,
+      "step": 3887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522608785781345e-06,
+      "loss": 0.9292,
+      "step": 3888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522313831803523e-06,
+      "loss": 0.9556,
+      "step": 3889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522018791306711e-06,
+      "loss": 0.9395,
+      "step": 3890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521723664296557e-06,
+      "loss": 0.9126,
+      "step": 3891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5214284507787e-06,
+      "loss": 0.8501,
+      "step": 3892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521133150758793e-06,
+      "loss": 0.9312,
+      "step": 3893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520837764242485e-06,
+      "loss": 0.9834,
+      "step": 3894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520542291235426e-06,
+      "loss": 1.02,
+      "step": 3895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52024673174327e-06,
+      "loss": 0.8291,
+      "step": 3896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51995108577167e-06,
+      "loss": 0.9663,
+      "step": 3897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519655353326285e-06,
+      "loss": 0.9702,
+      "step": 3898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51935953441277e-06,
+      "loss": 0.9404,
+      "step": 3899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519063629036784e-06,
+      "loss": 0.9849,
+      "step": 3900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518767637203991e-06,
+      "loss": 0.9331,
+      "step": 3901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518471558920053e-06,
+      "loss": 0.9292,
+      "step": 3902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518175394190634e-06,
+      "loss": 0.9316,
+      "step": 3903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5178791430214e-06,
+      "loss": 0.9917,
+      "step": 3904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517582805418018e-06,
+      "loss": 0.9736,
+      "step": 3905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517286381386157e-06,
+      "loss": 0.9355,
+      "step": 3906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516989870931491e-06,
+      "loss": 0.8901,
+      "step": 3907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51669327405969e-06,
+      "loss": 0.9414,
+      "step": 3908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516396590776429e-06,
+      "loss": 0.9619,
+      "step": 3909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516099821087385e-06,
+      "loss": 0.9268,
+      "step": 3910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515802964998235e-06,
+      "loss": 0.9263,
+      "step": 3911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515506022514658e-06,
+      "loss": 0.895,
+      "step": 3912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515208993642335e-06,
+      "loss": 0.8921,
+      "step": 3913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514911878386949e-06,
+      "loss": 0.9019,
+      "step": 3914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514614676754184e-06,
+      "loss": 0.9927,
+      "step": 3915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514317388749725e-06,
+      "loss": 0.9097,
+      "step": 3916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514020014379262e-06,
+      "loss": 0.9761,
+      "step": 3917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513722553648483e-06,
+      "loss": 0.9189,
+      "step": 3918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51342500656308e-06,
+      "loss": 0.8601,
+      "step": 3919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513127373128743e-06,
+      "loss": 0.9033,
+      "step": 3920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512829653351167e-06,
+      "loss": 0.8872,
+      "step": 3921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512531847236048e-06,
+      "loss": 0.8848,
+      "step": 3922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512233954789084e-06,
+      "loss": 0.9307,
+      "step": 3923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511935976015975e-06,
+      "loss": 0.9248,
+      "step": 3924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511637910922422e-06,
+      "loss": 0.8608,
+      "step": 3925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511339759514125e-06,
+      "loss": 0.8989,
+      "step": 3926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511041521796788e-06,
+      "loss": 0.9561,
+      "step": 3927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51074319777612e-06,
+      "loss": 0.8813,
+      "step": 3928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510444787457827e-06,
+      "loss": 0.8862,
+      "step": 3929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510146290847617e-06,
+      "loss": 0.9297,
+      "step": 3930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509847707951201e-06,
+      "loss": 0.9189,
+      "step": 3931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509549038774292e-06,
+      "loss": 0.939,
+      "step": 3932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509250283322603e-06,
+      "loss": 0.9541,
+      "step": 3933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508951441601851e-06,
+      "loss": 0.9634,
+      "step": 3934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508652513617753e-06,
+      "loss": 0.9482,
+      "step": 3935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508353499376028e-06,
+      "loss": 0.959,
+      "step": 3936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508054398882394e-06,
+      "loss": 0.936,
+      "step": 3937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507755212142577e-06,
+      "loss": 0.9131,
+      "step": 3938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5074559391623e-06,
+      "loss": 0.9824,
+      "step": 3939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507156579947288e-06,
+      "loss": 0.8979,
+      "step": 3940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506857134503268e-06,
+      "loss": 0.916,
+      "step": 3941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506557602835968e-06,
+      "loss": 0.8838,
+      "step": 3942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506257984951122e-06,
+      "loss": 0.9087,
+      "step": 3943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505958280854457e-06,
+      "loss": 0.3815,
+      "step": 3944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505658490551711e-06,
+      "loss": 0.9106,
+      "step": 3945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50535861404862e-06,
+      "loss": 0.8931,
+      "step": 3946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505058651350916e-06,
+      "loss": 0.8911,
+      "step": 3947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504758602464341e-06,
+      "loss": 0.9678,
+      "step": 3948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504458467394637e-06,
+      "loss": 0.9302,
+      "step": 3949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504158246147544e-06,
+      "loss": 0.9414,
+      "step": 3950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503857938728804e-06,
+      "loss": 0.9126,
+      "step": 3951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503557545144168e-06,
+      "loss": 0.9062,
+      "step": 3952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503257065399377e-06,
+      "loss": 0.8887,
+      "step": 3953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502956499500182e-06,
+      "loss": 0.9521,
+      "step": 3954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502655847452335e-06,
+      "loss": 0.9448,
+      "step": 3955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502355109261584e-06,
+      "loss": 0.8818,
+      "step": 3956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502054284933688e-06,
+      "loss": 0.8906,
+      "step": 3957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501753374474397e-06,
+      "loss": 0.8833,
+      "step": 3958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501452377889471e-06,
+      "loss": 0.957,
+      "step": 3959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501151295184667e-06,
+      "loss": 0.9395,
+      "step": 3960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500850126365746e-06,
+      "loss": 0.979,
+      "step": 3961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50054887143847e-06,
+      "loss": 0.9785,
+      "step": 3962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500247530408602e-06,
+      "loss": 0.915,
+      "step": 3963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499946103281906e-06,
+      "loss": 0.8667,
+      "step": 3964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499644590064152e-06,
+      "loss": 0.8994,
+      "step": 3965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499342990761105e-06,
+      "loss": 0.9268,
+      "step": 3966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499041305378538e-06,
+      "loss": 0.8438,
+      "step": 3967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498739533922218e-06,
+      "loss": 0.9268,
+      "step": 3968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498437676397924e-06,
+      "loss": 0.9648,
+      "step": 3969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498135732811429e-06,
+      "loss": 0.9146,
+      "step": 3970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497833703168509e-06,
+      "loss": 0.9531,
+      "step": 3971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497531587474942e-06,
+      "loss": 0.9355,
+      "step": 3972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497229385736508e-06,
+      "loss": 0.9165,
+      "step": 3973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49692709795899e-06,
+      "loss": 0.9077,
+      "step": 3974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496624724148169e-06,
+      "loss": 0.9448,
+      "step": 3975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496322264309832e-06,
+      "loss": 0.9531,
+      "step": 3976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496019718449767e-06,
+      "loss": 0.9619,
+      "step": 3977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495717086573756e-06,
+      "loss": 0.9448,
+      "step": 3978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495414368687596e-06,
+      "loss": 0.9116,
+      "step": 3979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495111564797073e-06,
+      "loss": 0.8608,
+      "step": 3980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494808674907985e-06,
+      "loss": 0.9312,
+      "step": 3981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494505699026122e-06,
+      "loss": 0.9795,
+      "step": 3982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494202637157282e-06,
+      "loss": 0.9697,
+      "step": 3983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493899489307265e-06,
+      "loss": 0.9019,
+      "step": 3984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49359625548187e-06,
+      "loss": 0.999,
+      "step": 3985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493292935686896e-06,
+      "loss": 0.9033,
+      "step": 3986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492989529928147e-06,
+      "loss": 0.9194,
+      "step": 3987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49268603821143e-06,
+      "loss": 0.9258,
+      "step": 3988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492382460542549e-06,
+      "loss": 0.8979,
+      "step": 3989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492078796927311e-06,
+      "loss": 0.9282,
+      "step": 3990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491775047371528e-06,
+      "loss": 0.9219,
+      "step": 3991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491471211881012e-06,
+      "loss": 0.9692,
+      "step": 3992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491167290461572e-06,
+      "loss": 0.9609,
+      "step": 3993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490863283119024e-06,
+      "loss": 0.9556,
+      "step": 3994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490559189859186e-06,
+      "loss": 0.9766,
+      "step": 3995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490255010687875e-06,
+      "loss": 0.3073,
+      "step": 3996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489950745610908e-06,
+      "loss": 0.9521,
+      "step": 3997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48964639463411e-06,
+      "loss": 0.8501,
+      "step": 3998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489341957763299e-06,
+      "loss": 0.9922,
+      "step": 3999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489037435004303e-06,
+      "loss": 0.9146,
+      "step": 4000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488732826362948e-06,
+      "loss": 0.936,
+      "step": 4001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48842813184506e-06,
+      "loss": 0.958,
+      "step": 4002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488123351456468e-06,
+      "loss": 0.9971,
+      "step": 4003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487818485203007e-06,
+      "loss": 0.96,
+      "step": 4004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487513533090505e-06,
+      "loss": 0.9824,
+      "step": 4005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487208495124796e-06,
+      "loss": 0.9404,
+      "step": 4006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486903371311718e-06,
+      "loss": 0.9478,
+      "step": 4007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486598161657108e-06,
+      "loss": 0.9756,
+      "step": 4008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486292866166806e-06,
+      "loss": 0.8843,
+      "step": 4009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485987484846653e-06,
+      "loss": 0.9478,
+      "step": 4010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48568201770249e-06,
+      "loss": 0.9644,
+      "step": 4011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48537646474016e-06,
+      "loss": 0.3021,
+      "step": 4012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48507082596551e-06,
+      "loss": 0.8926,
+      "step": 4013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48476510138439e-06,
+      "loss": 0.9019,
+      "step": 4014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484459291002643e-06,
+      "loss": 0.8853,
+      "step": 4015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484153394826126e-06,
+      "loss": 0.9199,
+      "step": 4016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483847412860686e-06,
+      "loss": 0.9385,
+      "step": 4017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483541345112182e-06,
+      "loss": 0.9365,
+      "step": 4018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483235191586466e-06,
+      "loss": 0.9526,
+      "step": 4019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482928952289395e-06,
+      "loss": 0.9668,
+      "step": 4020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48262262722683e-06,
+      "loss": 0.9443,
+      "step": 4021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48231621640463e-06,
+      "loss": 0.9741,
+      "step": 4022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482009719828658e-06,
+      "loss": 0.854,
+      "step": 4023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481703137504777e-06,
+      "loss": 0.8726,
+      "step": 4024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481396469438852e-06,
+      "loss": 0.9238,
+      "step": 4025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481089715636751e-06,
+      "loss": 0.2861,
+      "step": 4026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480782876104343e-06,
+      "loss": 0.9492,
+      "step": 4027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480475950847496e-06,
+      "loss": 0.9722,
+      "step": 4028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480168939872087e-06,
+      "loss": 0.9023,
+      "step": 4029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479861843183984e-06,
+      "loss": 0.9292,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479554660789065e-06,
+      "loss": 0.9653,
+      "step": 4031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479247392693206e-06,
+      "loss": 0.9229,
+      "step": 4032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478940038902286e-06,
+      "loss": 0.9556,
+      "step": 4033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478632599422186e-06,
+      "loss": 0.9087,
+      "step": 4034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478325074258786e-06,
+      "loss": 0.9595,
+      "step": 4035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478017463417971e-06,
+      "loss": 1.0093,
+      "step": 4036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477709766905625e-06,
+      "loss": 0.9365,
+      "step": 4037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477401984727636e-06,
+      "loss": 0.8447,
+      "step": 4038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47709411688989e-06,
+      "loss": 0.9229,
+      "step": 4039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476786163398279e-06,
+      "loss": 0.9409,
+      "step": 4040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476478124258696e-06,
+      "loss": 0.9277,
+      "step": 4041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47616999947703e-06,
+      "loss": 0.96,
+      "step": 4042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47586178905918e-06,
+      "loss": 0.9321,
+      "step": 4043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47555349301104e-06,
+      "loss": 0.9492,
+      "step": 4044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47524511133851e-06,
+      "loss": 0.9893,
+      "step": 4045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474936644047488e-06,
+      "loss": 0.9619,
+      "step": 4046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474628091143876e-06,
+      "loss": 0.3009,
+      "step": 4047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474319452633579e-06,
+      "loss": 0.9399,
+      "step": 4048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474010728522501e-06,
+      "loss": 0.9453,
+      "step": 4049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473701918816545e-06,
+      "loss": 0.9375,
+      "step": 4050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473393023521623e-06,
+      "loss": 0.8887,
+      "step": 4051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47308404264364e-06,
+      "loss": 0.9106,
+      "step": 4052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472774976188515e-06,
+      "loss": 0.9658,
+      "step": 4053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472465824162155e-06,
+      "loss": 0.9487,
+      "step": 4054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472156586570475e-06,
+      "loss": 0.9395,
+      "step": 4055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471847263419393e-06,
+      "loss": 0.9229,
+      "step": 4056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471537854714827e-06,
+      "loss": 0.9072,
+      "step": 4057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471228360462696e-06,
+      "loss": 0.9692,
+      "step": 4058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470918780668919e-06,
+      "loss": 0.8813,
+      "step": 4059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470609115339421e-06,
+      "loss": 0.957,
+      "step": 4060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470299364480126e-06,
+      "loss": 0.9478,
+      "step": 4061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469989528096959e-06,
+      "loss": 0.897,
+      "step": 4062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46967960619585e-06,
+      "loss": 0.8564,
+      "step": 4063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469369598782725e-06,
+      "loss": 0.9741,
+      "step": 4064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469059505863518e-06,
+      "loss": 0.9331,
+      "step": 4065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46874932744416e-06,
+      "loss": 0.9497,
+      "step": 4066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468439063530585e-06,
+      "loss": 0.8984,
+      "step": 4067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468128714128731e-06,
+      "loss": 0.9526,
+      "step": 4068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467818279244532e-06,
+      "loss": 0.9517,
+      "step": 4069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467507758883929e-06,
+      "loss": 0.9297,
+      "step": 4070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467197153052863e-06,
+      "loss": 0.9189,
+      "step": 4071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466886461757276e-06,
+      "loss": 0.3453,
+      "step": 4072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466575685003114e-06,
+      "loss": 0.9473,
+      "step": 4073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466264822796319e-06,
+      "loss": 0.9326,
+      "step": 4074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465953875142839e-06,
+      "loss": 0.8911,
+      "step": 4075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465642842048625e-06,
+      "loss": 0.8955,
+      "step": 4076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465331723519627e-06,
+      "loss": 0.9121,
+      "step": 4077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465020519561795e-06,
+      "loss": 0.9141,
+      "step": 4078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464709230181085e-06,
+      "loss": 0.9238,
+      "step": 4079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464397855383453e-06,
+      "loss": 0.9155,
+      "step": 4080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464086395174854e-06,
+      "loss": 0.9106,
+      "step": 4081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463774849561249e-06,
+      "loss": 0.9287,
+      "step": 4082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463463218548594e-06,
+      "loss": 0.9858,
+      "step": 4083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463151502142858e-06,
+      "loss": 0.8843,
+      "step": 4084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462839700349999e-06,
+      "loss": 0.8408,
+      "step": 4085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462527813175983e-06,
+      "loss": 0.8379,
+      "step": 4086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462215840626778e-06,
+      "loss": 0.8716,
+      "step": 4087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461903782708354e-06,
+      "loss": 0.8428,
+      "step": 4088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46159163942668e-06,
+      "loss": 0.9663,
+      "step": 4089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461279410787725e-06,
+      "loss": 0.8267,
+      "step": 4090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460967096797466e-06,
+      "loss": 0.9526,
+      "step": 4091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460654697461875e-06,
+      "loss": 0.9082,
+      "step": 4092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460342212786933e-06,
+      "loss": 0.939,
+      "step": 4093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460029642778615e-06,
+      "loss": 0.9365,
+      "step": 4094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459716987442901e-06,
+      "loss": 0.8979,
+      "step": 4095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459404246785775e-06,
+      "loss": 0.918,
+      "step": 4096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459091420813218e-06,
+      "loss": 0.9478,
+      "step": 4097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458778509531216e-06,
+      "loss": 0.9414,
+      "step": 4098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458465512945753e-06,
+      "loss": 0.9053,
+      "step": 4099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458152431062821e-06,
+      "loss": 0.9673,
+      "step": 4100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457839263888406e-06,
+      "loss": 0.9468,
+      "step": 4101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457526011428504e-06,
+      "loss": 0.8843,
+      "step": 4102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457212673689102e-06,
+      "loss": 0.9878,
+      "step": 4103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4568992506762e-06,
+      "loss": 0.999,
+      "step": 4104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456585742395792e-06,
+      "loss": 0.9385,
+      "step": 4105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456272148853875e-06,
+      "loss": 0.9771,
+      "step": 4106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45595847005645e-06,
+      "loss": 0.9409,
+      "step": 4107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455644706009519e-06,
+      "loss": 0.9658,
+      "step": 4108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455330856719082e-06,
+      "loss": 0.9409,
+      "step": 4109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455016922191145e-06,
+      "loss": 0.9004,
+      "step": 4110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454702902431716e-06,
+      "loss": 0.3182,
+      "step": 4111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4543887974468e-06,
+      "loss": 0.96,
+      "step": 4112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454074607242404e-06,
+      "loss": 0.9492,
+      "step": 4113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453760331824544e-06,
+      "loss": 0.9253,
+      "step": 4114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45344597119923e-06,
+      "loss": 0.8755,
+      "step": 4115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45313152537248e-06,
+      "loss": 0.9482,
+      "step": 4116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452816994350304e-06,
+      "loss": 0.9575,
+      "step": 4117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452502378138722e-06,
+      "loss": 0.9971,
+      "step": 4118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452187676743753e-06,
+      "loss": 0.9614,
+      "step": 4119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451872890171419e-06,
+      "loss": 0.9663,
+      "step": 4120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451558018427742e-06,
+      "loss": 0.9072,
+      "step": 4121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451243061518744e-06,
+      "loss": 0.9058,
+      "step": 4122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450928019450451e-06,
+      "loss": 0.8789,
+      "step": 4123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450612892228894e-06,
+      "loss": 0.9136,
+      "step": 4124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450297679860097e-06,
+      "loss": 0.8828,
+      "step": 4125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449982382350093e-06,
+      "loss": 0.9263,
+      "step": 4126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449666999704914e-06,
+      "loss": 0.9375,
+      "step": 4127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449351531930595e-06,
+      "loss": 0.9741,
+      "step": 4128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449035979033167e-06,
+      "loss": 0.9229,
+      "step": 4129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448720341018671e-06,
+      "loss": 0.9736,
+      "step": 4130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448404617893146e-06,
+      "loss": 0.8833,
+      "step": 4131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448088809662629e-06,
+      "loss": 0.9131,
+      "step": 4132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447772916333164e-06,
+      "loss": 1.002,
+      "step": 4133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447456937910796e-06,
+      "loss": 0.9404,
+      "step": 4134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447140874401565e-06,
+      "loss": 0.9141,
+      "step": 4135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446824725811522e-06,
+      "loss": 0.9878,
+      "step": 4136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446508492146717e-06,
+      "loss": 0.9106,
+      "step": 4137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446192173413199e-06,
+      "loss": 0.9409,
+      "step": 4138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445875769617015e-06,
+      "loss": 0.9414,
+      "step": 4139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445559280764223e-06,
+      "loss": 0.9849,
+      "step": 4140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445242706860875e-06,
+      "loss": 0.938,
+      "step": 4141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444926047913033e-06,
+      "loss": 0.9268,
+      "step": 4142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444609303926747e-06,
+      "loss": 0.9619,
+      "step": 4143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444292474908084e-06,
+      "loss": 0.9229,
+      "step": 4144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443975560863101e-06,
+      "loss": 0.9404,
+      "step": 4145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443658561797864e-06,
+      "loss": 0.8906,
+      "step": 4146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443341477718436e-06,
+      "loss": 0.8389,
+      "step": 4147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443024308630884e-06,
+      "loss": 0.9658,
+      "step": 4148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442707054541274e-06,
+      "loss": 0.9146,
+      "step": 4149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44238971545568e-06,
+      "loss": 0.9204,
+      "step": 4150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442072291380168e-06,
+      "loss": 0.9141,
+      "step": 4151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441754782320813e-06,
+      "loss": 0.8716,
+      "step": 4152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441437188283691e-06,
+      "loss": 0.9287,
+      "step": 4153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441119509274877e-06,
+      "loss": 0.9321,
+      "step": 4154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440801745300447e-06,
+      "loss": 0.9795,
+      "step": 4155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440483896366483e-06,
+      "loss": 0.9077,
+      "step": 4156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440165962479065e-06,
+      "loss": 0.8926,
+      "step": 4157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439847943644276e-06,
+      "loss": 0.9453,
+      "step": 4158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439529839868198e-06,
+      "loss": 0.9429,
+      "step": 4159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439211651156918e-06,
+      "loss": 0.9312,
+      "step": 4160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438893377516524e-06,
+      "loss": 0.9087,
+      "step": 4161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438575018953106e-06,
+      "loss": 0.9233,
+      "step": 4162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438256575472753e-06,
+      "loss": 0.9727,
+      "step": 4163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437938047081558e-06,
+      "loss": 0.9214,
+      "step": 4164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437619433785614e-06,
+      "loss": 0.9092,
+      "step": 4165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437300735591018e-06,
+      "loss": 0.8989,
+      "step": 4166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436981952503866e-06,
+      "loss": 0.9053,
+      "step": 4167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436663084530257e-06,
+      "loss": 0.9155,
+      "step": 4168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436344131676293e-06,
+      "loss": 0.9443,
+      "step": 4169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436025093948076e-06,
+      "loss": 0.9375,
+      "step": 4170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435705971351707e-06,
+      "loss": 0.9746,
+      "step": 4171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435386763893294e-06,
+      "loss": 0.9941,
+      "step": 4172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435067471578943e-06,
+      "loss": 0.9492,
+      "step": 4173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43474809441476e-06,
+      "loss": 0.8892,
+      "step": 4174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43442863240686e-06,
+      "loss": 0.9492,
+      "step": 4175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434109085561352e-06,
+      "loss": 0.9043,
+      "step": 4176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43378945388435e-06,
+      "loss": 0.9116,
+      "step": 4177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433469737381968e-06,
+      "loss": 0.9678,
+      "step": 4178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433149936060327e-06,
+      "loss": 0.8376,
+      "step": 4179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432830049925537e-06,
+      "loss": 0.9224,
+      "step": 4180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432510078983726e-06,
+      "loss": 0.9609,
+      "step": 4181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432190023241012e-06,
+      "loss": 0.8838,
+      "step": 4182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431869882703518e-06,
+      "loss": 0.4807,
+      "step": 4183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43154965737737e-06,
+      "loss": 0.9058,
+      "step": 4184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431229347268694e-06,
+      "loss": 0.937,
+      "step": 4185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430908952383617e-06,
+      "loss": 0.895,
+      "step": 4186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430588472728271e-06,
+      "loss": 0.9253,
+      "step": 4187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430267908308783e-06,
+      "loss": 0.307,
+      "step": 4188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42994725913129e-06,
+      "loss": 0.9868,
+      "step": 4189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429626525201927e-06,
+      "loss": 0.9385,
+      "step": 4190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429305706526826e-06,
+      "loss": 0.9033,
+      "step": 4191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428984803112126e-06,
+      "loss": 0.9331,
+      "step": 4192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42866381496397e-06,
+      "loss": 0.9399,
+      "step": 4193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428342742088497e-06,
+      "loss": 0.918,
+      "step": 4194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428021584491849e-06,
+      "loss": 0.9038,
+      "step": 4195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427700342180168e-06,
+      "loss": 0.9126,
+      "step": 4196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427379015159604e-06,
+      "loss": 0.9009,
+      "step": 4197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427057603436303e-06,
+      "loss": 0.939,
+      "step": 4198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426736107016413e-06,
+      "loss": 0.9272,
+      "step": 4199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426414525906087e-06,
+      "loss": 0.9087,
+      "step": 4200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426092860111474e-06,
+      "loss": 0.9111,
+      "step": 4201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425771109638731e-06,
+      "loss": 0.9326,
+      "step": 4202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425449274494012e-06,
+      "loss": 0.8779,
+      "step": 4203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425127354683477e-06,
+      "loss": 0.9575,
+      "step": 4204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42480535021328e-06,
+      "loss": 0.9307,
+      "step": 4205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424483261089584e-06,
+      "loss": 0.9604,
+      "step": 4206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424161087318552e-06,
+      "loss": 0.9214,
+      "step": 4207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423838828906346e-06,
+      "loss": 0.9062,
+      "step": 4208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423516485859133e-06,
+      "loss": 0.9126,
+      "step": 4209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42319405818308e-06,
+      "loss": 0.9189,
+      "step": 4210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422871545884352e-06,
+      "loss": 0.9478,
+      "step": 4211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422548948969124e-06,
+      "loss": 0.959,
+      "step": 4212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422226267443564e-06,
+      "loss": 0.9814,
+      "step": 4213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421903501313846e-06,
+      "loss": 0.8989,
+      "step": 4214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421580650586149e-06,
+      "loss": 0.8857,
+      "step": 4215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421257715266645e-06,
+      "loss": 0.3491,
+      "step": 4216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420934695361515e-06,
+      "loss": 0.8955,
+      "step": 4217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420611590876936e-06,
+      "loss": 0.9341,
+      "step": 4218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420288401819092e-06,
+      "loss": 0.9028,
+      "step": 4219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419965128194166e-06,
+      "loss": 0.915,
+      "step": 4220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41964177000834e-06,
+      "loss": 0.9473,
+      "step": 4221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419318327267805e-06,
+      "loss": 0.9292,
+      "step": 4222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418994799978745e-06,
+      "loss": 0.9458,
+      "step": 4223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418671188147352e-06,
+      "loss": 0.873,
+      "step": 4224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418347491779815e-06,
+      "loss": 0.8892,
+      "step": 4225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41802371088233e-06,
+      "loss": 0.9575,
+      "step": 4226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417699845461086e-06,
+      "loss": 0.9341,
+      "step": 4227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417375895522286e-06,
+      "loss": 0.8926,
+      "step": 4228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417051861072121e-06,
+      "loss": 0.9805,
+      "step": 4229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416727742116797e-06,
+      "loss": 0.9424,
+      "step": 4230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416403538662509e-06,
+      "loss": 0.321,
+      "step": 4231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416079250715463e-06,
+      "loss": 0.9497,
+      "step": 4232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415754878281863e-06,
+      "loss": 0.9595,
+      "step": 4233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41543042136791e-06,
+      "loss": 0.9424,
+      "step": 4234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415105879979818e-06,
+      "loss": 0.9087,
+      "step": 4235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414781254123794e-06,
+      "loss": 0.9097,
+      "step": 4236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414456543806046e-06,
+      "loss": 0.9004,
+      "step": 4237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414131749032789e-06,
+      "loss": 0.896,
+      "step": 4238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413806869810235e-06,
+      "loss": 0.9917,
+      "step": 4239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413481906144602e-06,
+      "loss": 0.9043,
+      "step": 4240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413156858042103e-06,
+      "loss": 0.9595,
+      "step": 4241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41283172550896e-06,
+      "loss": 1.0083,
+      "step": 4242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412506508551392e-06,
+      "loss": 0.979,
+      "step": 4243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412181207175622e-06,
+      "loss": 0.9663,
+      "step": 4244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411855821387872e-06,
+      "loss": 0.9062,
+      "step": 4245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41153035119437e-06,
+      "loss": 0.9033,
+      "step": 4246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411204796601339e-06,
+      "loss": 0.958,
+      "step": 4247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410879157615009e-06,
+      "loss": 0.9106,
+      "step": 4248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410553434241609e-06,
+      "loss": 0.8921,
+      "step": 4249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410227626487374e-06,
+      "loss": 0.333,
+      "step": 4250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409901734358533e-06,
+      "loss": 0.9507,
+      "step": 4251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409575757861322e-06,
+      "loss": 0.9111,
+      "step": 4252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409249697001981e-06,
+      "loss": 0.9673,
+      "step": 4253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408923551786742e-06,
+      "loss": 0.9111,
+      "step": 4254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40859732222185e-06,
+      "loss": 0.9614,
+      "step": 4255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408271008313544e-06,
+      "loss": 0.98,
+      "step": 4256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407944610068068e-06,
+      "loss": 0.9204,
+      "step": 4257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407618127491663e-06,
+      "loss": 0.9414,
+      "step": 4258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40729156059058e-06,
+      "loss": 0.8114,
+      "step": 4259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406964909371062e-06,
+      "loss": 1.0068,
+      "step": 4260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406638173839362e-06,
+      "loss": 1.0,
+      "step": 4261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40631135400173e-06,
+      "loss": 0.9873,
+      "step": 4262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405984449864418e-06,
+      "loss": 0.9717,
+      "step": 4263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405657461433681e-06,
+      "loss": 0.8779,
+      "step": 4264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405330388715774e-06,
+      "loss": 0.9951,
+      "step": 4265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405003231716954e-06,
+      "loss": 0.9033,
+      "step": 4266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404675990443482e-06,
+      "loss": 0.8955,
+      "step": 4267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404348664901618e-06,
+      "loss": 0.9609,
+      "step": 4268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40402125509762e-06,
+      "loss": 0.8984,
+      "step": 4269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403693761037758e-06,
+      "loss": 0.9663,
+      "step": 4270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403366182728296e-06,
+      "loss": 0.938,
+      "step": 4271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4030385201755e-06,
+      "loss": 0.9272,
+      "step": 4272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402710773385638e-06,
+      "loss": 0.9326,
+      "step": 4273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40238294236498e-06,
+      "loss": 0.957,
+      "step": 4274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402055027119799e-06,
+      "loss": 0.9517,
+      "step": 4275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401727027656372e-06,
+      "loss": 0.9346,
+      "step": 4276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401398943980966e-06,
+      "loss": 0.9219,
+      "step": 4277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401070776099866e-06,
+      "loss": 0.8896,
+      "step": 4278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400742524019345e-06,
+      "loss": 0.9893,
+      "step": 4279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400414187745685e-06,
+      "loss": 0.936,
+      "step": 4280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400085767285167e-06,
+      "loss": 0.9912,
+      "step": 4281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399757262644077e-06,
+      "loss": 0.9004,
+      "step": 4282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399428673828694e-06,
+      "loss": 0.939,
+      "step": 4283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39910000084531e-06,
+      "loss": 0.9209,
+      "step": 4284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398771243700211e-06,
+      "loss": 0.9585,
+      "step": 4285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398442402399686e-06,
+      "loss": 0.9126,
+      "step": 4286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398113476950025e-06,
+      "loss": 0.9009,
+      "step": 4287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397784467357525e-06,
+      "loss": 0.9048,
+      "step": 4288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397455373628477e-06,
+      "loss": 0.9424,
+      "step": 4289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39712619576918e-06,
+      "loss": 0.8301,
+      "step": 4290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396796933785928e-06,
+      "loss": 0.9053,
+      "step": 4291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396467587685022e-06,
+      "loss": 0.9258,
+      "step": 4292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396138157472765e-06,
+      "loss": 0.8618,
+      "step": 4293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395808643155455e-06,
+      "loss": 0.9795,
+      "step": 4294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.3954790447394e-06,
+      "loss": 0.9121,
+      "step": 4295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395149362230904e-06,
+      "loss": 0.9756,
+      "step": 4296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394819595636274e-06,
+      "loss": 0.8989,
+      "step": 4297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394489744961823e-06,
+      "loss": 0.9517,
+      "step": 4298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394159810213857e-06,
+      "loss": 0.9272,
+      "step": 4299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393829791398687e-06,
+      "loss": 0.9722,
+      "step": 4300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393499688522632e-06,
+      "loss": 0.8955,
+      "step": 4301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393169501592004e-06,
+      "loss": 0.9463,
+      "step": 4302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39283923061312e-06,
+      "loss": 0.9902,
+      "step": 4303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.3925088755923e-06,
+      "loss": 0.9165,
+      "step": 4304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392178436535864e-06,
+      "loss": 0.8892,
+      "step": 4305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391847913450131e-06,
+      "loss": 0.9282,
+      "step": 4306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391517306341428e-06,
+      "loss": 0.9473,
+      "step": 4307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391186615216078e-06,
+      "loss": 0.9014,
+      "step": 4308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390855840080411e-06,
+      "loss": 0.9365,
+      "step": 4309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390524980940752e-06,
+      "loss": 0.3259,
+      "step": 4310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39019403780343e-06,
+      "loss": 0.8794,
+      "step": 4311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389863010674779e-06,
+      "loss": 0.8511,
+      "step": 4312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38953189956113e-06,
+      "loss": 0.9277,
+      "step": 4313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38920070446882e-06,
+      "loss": 0.9209,
+      "step": 4314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38886942540418e-06,
+      "loss": 0.9619,
+      "step": 4315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388538062373557e-06,
+      "loss": 0.9287,
+      "step": 4316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388206615383284e-06,
+      "loss": 0.9272,
+      "step": 4317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387875084439703e-06,
+      "loss": 0.9814,
+      "step": 4318
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387543469549156e-06,
+      "loss": 0.9521,
+      "step": 4319
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38721177071799e-06,
+      "loss": 0.8643,
+      "step": 4320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386879987952549e-06,
+      "loss": 0.9272,
+      "step": 4321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386548121259182e-06,
+      "loss": 0.35,
+      "step": 4322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386216170644235e-06,
+      "loss": 0.8809,
+      "step": 4323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385884136114063e-06,
+      "loss": 0.916,
+      "step": 4324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385552017675015e-06,
+      "loss": 0.9282,
+      "step": 4325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385219815333447e-06,
+      "loss": 0.9521,
+      "step": 4326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384887529095714e-06,
+      "loss": 0.8882,
+      "step": 4327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384555158968172e-06,
+      "loss": 0.9165,
+      "step": 4328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38422270495718e-06,
+      "loss": 0.8784,
+      "step": 4329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383890167069098e-06,
+      "loss": 0.8901,
+      "step": 4330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383557545310291e-06,
+      "loss": 0.9331,
+      "step": 4331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383224839687121e-06,
+      "loss": 0.9399,
+      "step": 4332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382892050205951e-06,
+      "loss": 0.9326,
+      "step": 4333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382559176873151e-06,
+      "loss": 0.9575,
+      "step": 4334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382226219695087e-06,
+      "loss": 0.8838,
+      "step": 4335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381893178678129e-06,
+      "loss": 0.8716,
+      "step": 4336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381560053828651e-06,
+      "loss": 0.8774,
+      "step": 4337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381226845153024e-06,
+      "loss": 0.897,
+      "step": 4338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380893552657623e-06,
+      "loss": 0.8706,
+      "step": 4339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380560176348826e-06,
+      "loss": 0.9243,
+      "step": 4340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38022671623301e-06,
+      "loss": 0.9111,
+      "step": 4341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379893172316555e-06,
+      "loss": 0.3716,
+      "step": 4342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37955954460584e-06,
+      "loss": 0.9448,
+      "step": 4343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379225833107251e-06,
+      "loss": 0.9746,
+      "step": 4344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378892037827169e-06,
+      "loss": 0.9639,
+      "step": 4345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378558158771984e-06,
+      "loss": 0.8779,
+      "step": 4346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37822419594808e-06,
+      "loss": 0.8313,
+      "step": 4347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377890149361851e-06,
+      "loss": 0.9585,
+      "step": 4348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377556019019682e-06,
+      "loss": 0.8677,
+      "step": 4349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377221804927969e-06,
+      "loss": 0.9873,
+      "step": 4350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376887507093104e-06,
+      "loss": 0.9565,
+      "step": 4351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376553125521486e-06,
+      "loss": 0.9238,
+      "step": 4352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37621866021951e-06,
+      "loss": 0.9468,
+      "step": 4353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375884111193573e-06,
+      "loss": 0.9302,
+      "step": 4354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375549478450078e-06,
+      "loss": 0.9277,
+      "step": 4355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375214761995428e-06,
+      "loss": 0.8687,
+      "step": 4356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374879961836023e-06,
+      "loss": 0.9785,
+      "step": 4357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374545077978273e-06,
+      "loss": 0.9131,
+      "step": 4358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374210110428581e-06,
+      "loss": 0.9629,
+      "step": 4359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373875059193356e-06,
+      "loss": 0.9087,
+      "step": 4360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373539924279009e-06,
+      "loss": 0.9478,
+      "step": 4361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373204705691952e-06,
+      "loss": 0.9692,
+      "step": 4362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372869403438598e-06,
+      "loss": 0.9072,
+      "step": 4363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37253401752536e-06,
+      "loss": 0.3403,
+      "step": 4364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372198547958656e-06,
+      "loss": 0.9404,
+      "step": 4365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371862994744904e-06,
+      "loss": 0.916,
+      "step": 4366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371527357890526e-06,
+      "loss": 0.8555,
+      "step": 4367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37119163740194e-06,
+      "loss": 0.8706,
+      "step": 4368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370855833285568e-06,
+      "loss": 0.937,
+      "step": 4369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370519945547837e-06,
+      "loss": 0.9941,
+      "step": 4370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370183974195175e-06,
+      "loss": 0.938,
+      "step": 4371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369847919234003e-06,
+      "loss": 0.9326,
+      "step": 4372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369511780670758e-06,
+      "loss": 0.9595,
+      "step": 4373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369175558511865e-06,
+      "loss": 0.9468,
+      "step": 4374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36883925276376e-06,
+      "loss": 0.9067,
+      "step": 4375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368502863432873e-06,
+      "loss": 1.0171,
+      "step": 4376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368166390525645e-06,
+      "loss": 0.8843,
+      "step": 4377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36782983404851e-06,
+      "loss": 0.9009,
+      "step": 4378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367493194007906e-06,
+      "loss": 0.8765,
+      "step": 4379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367156470410276e-06,
+      "loss": 0.9609,
+      "step": 4380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366819663262062e-06,
+      "loss": 0.894,
+      "step": 4381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366482772569705e-06,
+      "loss": 0.896,
+      "step": 4382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366145798339652e-06,
+      "loss": 0.9429,
+      "step": 4383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365808740578349e-06,
+      "loss": 1.0332,
+      "step": 4384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365471599292247e-06,
+      "loss": 0.8926,
+      "step": 4385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365134374487792e-06,
+      "loss": 0.9233,
+      "step": 4386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364797066171441e-06,
+      "loss": 0.9189,
+      "step": 4387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364459674349642e-06,
+      "loss": 0.9629,
+      "step": 4388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36412219902885e-06,
+      "loss": 0.9258,
+      "step": 4389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363784640215526e-06,
+      "loss": 0.9146,
+      "step": 4390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363446997916126e-06,
+      "loss": 0.9648,
+      "step": 4391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363109272137108e-06,
+      "loss": 0.9487,
+      "step": 4392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362771462884934e-06,
+      "loss": 0.9307,
+      "step": 4393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362433570166067e-06,
+      "loss": 0.8444,
+      "step": 4394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362095593986971e-06,
+      "loss": 0.915,
+      "step": 4395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361757534354112e-06,
+      "loss": 0.8955,
+      "step": 4396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36141939127396e-06,
+      "loss": 0.8945,
+      "step": 4397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361081164752981e-06,
+      "loss": 0.9438,
+      "step": 4398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360742854797646e-06,
+      "loss": 0.9263,
+      "step": 4399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360404461414432e-06,
+      "loss": 0.9023,
+      "step": 4400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360065984609807e-06,
+      "loss": 0.9531,
+      "step": 4401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359727424390249e-06,
+      "loss": 0.9346,
+      "step": 4402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359388780762236e-06,
+      "loss": 0.9429,
+      "step": 4403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359050053732246e-06,
+      "loss": 0.938,
+      "step": 4404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358711243306759e-06,
+      "loss": 0.9209,
+      "step": 4405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358372349492259e-06,
+      "loss": 0.8281,
+      "step": 4406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358033372295225e-06,
+      "loss": 0.9268,
+      "step": 4407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357694311722149e-06,
+      "loss": 0.8604,
+      "step": 4408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357355167779512e-06,
+      "loss": 0.9546,
+      "step": 4409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357015940473803e-06,
+      "loss": 0.9526,
+      "step": 4410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356676629811515e-06,
+      "loss": 0.9365,
+      "step": 4411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356337235799138e-06,
+      "loss": 0.9155,
+      "step": 4412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355997758443164e-06,
+      "loss": 0.8735,
+      "step": 4413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355658197750088e-06,
+      "loss": 0.8623,
+      "step": 4414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35531855372641e-06,
+      "loss": 0.9766,
+      "step": 4415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354978826378622e-06,
+      "loss": 0.9487,
+      "step": 4416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354639015713227e-06,
+      "loss": 0.9019,
+      "step": 4417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354299121736726e-06,
+      "loss": 0.9438,
+      "step": 4418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353959144455621e-06,
+      "loss": 0.2659,
+      "step": 4419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353619083876418e-06,
+      "loss": 0.9468,
+      "step": 4420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353278940005619e-06,
+      "loss": 0.9043,
+      "step": 4421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352938712849735e-06,
+      "loss": 0.9561,
+      "step": 4422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352598402415274e-06,
+      "loss": 0.9678,
+      "step": 4423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352258008708748e-06,
+      "loss": 0.9126,
+      "step": 4424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351917531736667e-06,
+      "loss": 0.9297,
+      "step": 4425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351576971505545e-06,
+      "loss": 0.9067,
+      "step": 4426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3512363280219e-06,
+      "loss": 0.9014,
+      "step": 4427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350895601292247e-06,
+      "loss": 0.9224,
+      "step": 4428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350554791323103e-06,
+      "loss": 0.9409,
+      "step": 4429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350213898120992e-06,
+      "loss": 0.9219,
+      "step": 4430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349872921692434e-06,
+      "loss": 0.917,
+      "step": 4431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349531862043952e-06,
+      "loss": 0.9131,
+      "step": 4432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349190719182073e-06,
+      "loss": 1.0166,
+      "step": 4433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34884949311332e-06,
+      "loss": 0.8735,
+      "step": 4434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348508183844224e-06,
+      "loss": 0.9424,
+      "step": 4435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348166791381314e-06,
+      "loss": 0.9565,
+      "step": 4436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347825315731122e-06,
+      "loss": 0.8999,
+      "step": 4437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34748375690018e-06,
+      "loss": 0.9004,
+      "step": 4438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347142114895022e-06,
+      "loss": 0.96,
+      "step": 4439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346800389722187e-06,
+      "loss": 0.9292,
+      "step": 4440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346458581388209e-06,
+      "loss": 0.8511,
+      "step": 4441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346116689899629e-06,
+      "loss": 0.9019,
+      "step": 4442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34577471526299e-06,
+      "loss": 0.9634,
+      "step": 4443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345432657484832e-06,
+      "loss": 0.9111,
+      "step": 4444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345090516571699e-06,
+      "loss": 0.8989,
+      "step": 4445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344748292530137e-06,
+      "loss": 0.9277,
+      "step": 4446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344405985366696e-06,
+      "loss": 0.9229,
+      "step": 4447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34406359508792e-06,
+      "loss": 0.9116,
+      "step": 4448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343721121700363e-06,
+      "loss": 0.9146,
+      "step": 4449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343378565210575e-06,
+      "loss": 0.9062,
+      "step": 4450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343035925625112e-06,
+      "loss": 0.8892,
+      "step": 4451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342693202950528e-06,
+      "loss": 0.9321,
+      "step": 4452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34235039719338e-06,
+      "loss": 0.9102,
+      "step": 4453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342007508360227e-06,
+      "loss": 0.9004,
+      "step": 4454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341664536457626e-06,
+      "loss": 1.0078,
+      "step": 4455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341321481492142e-06,
+      "loss": 0.938,
+      "step": 4456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340978343470338e-06,
+      "loss": 0.937,
+      "step": 4457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340635122398776e-06,
+      "loss": 0.9268,
+      "step": 4458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340291818284024e-06,
+      "loss": 0.9033,
+      "step": 4459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33994843113265e-06,
+      "loss": 0.9038,
+      "step": 4460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339604960951225e-06,
+      "loss": 0.9614,
+      "step": 4461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339261407746318e-06,
+      "loss": 0.8916,
+      "step": 4462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338917771524503e-06,
+      "loss": 0.9644,
+      "step": 4463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338574052292353e-06,
+      "loss": 0.981,
+      "step": 4464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338230250056445e-06,
+      "loss": 0.9473,
+      "step": 4465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337886364823356e-06,
+      "loss": 0.8984,
+      "step": 4466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337542396599665e-06,
+      "loss": 0.9507,
+      "step": 4467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337198345391954e-06,
+      "loss": 0.9619,
+      "step": 4468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336854211206803e-06,
+      "loss": 0.8408,
+      "step": 4469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336509994050798e-06,
+      "loss": 0.8159,
+      "step": 4470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33616569393052e-06,
+      "loss": 0.9868,
+      "step": 4471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335821310852562e-06,
+      "loss": 0.8823,
+      "step": 4472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33547684482351e-06,
+      "loss": 0.8809,
+      "step": 4473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335132295849954e-06,
+      "loss": 0.9302,
+      "step": 4474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334787663938485e-06,
+      "loss": 0.8638,
+      "step": 4475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3344429490957e-06,
+      "loss": 0.9531,
+      "step": 4476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334098151328188e-06,
+      "loss": 0.9077,
+      "step": 4477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33375327064255e-06,
+      "loss": 0.9434,
+      "step": 4478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333408307045382e-06,
+      "loss": 0.9082,
+      "step": 4479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333063260543285e-06,
+      "loss": 0.9141,
+      "step": 4480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332718131142859e-06,
+      "loss": 0.9297,
+      "step": 4481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332372918850708e-06,
+      "loss": 0.9224,
+      "step": 4482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332027623673436e-06,
+      "loss": 0.96,
+      "step": 4483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33168224561765e-06,
+      "loss": 0.9551,
+      "step": 4484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331336784689957e-06,
+      "loss": 0.9473,
+      "step": 4485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330991240896964e-06,
+      "loss": 0.9121,
+      "step": 4486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330645614245286e-06,
+      "loss": 0.9336,
+      "step": 4487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330299904741532e-06,
+      "loss": 0.957,
+      "step": 4488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329954112392317e-06,
+      "loss": 0.9355,
+      "step": 4489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329608237204257e-06,
+      "loss": 0.9106,
+      "step": 4490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329262279183969e-06,
+      "loss": 0.917,
+      "step": 4491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328916238338072e-06,
+      "loss": 0.9014,
+      "step": 4492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328570114673184e-06,
+      "loss": 0.9312,
+      "step": 4493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32822390819593e-06,
+      "loss": 0.8926,
+      "step": 4494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327877618912932e-06,
+      "loss": 0.8774,
+      "step": 4495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327531246830817e-06,
+      "loss": 0.9341,
+      "step": 4496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327184791956208e-06,
+      "loss": 0.9023,
+      "step": 4497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326838254295734e-06,
+      "loss": 0.9561,
+      "step": 4498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326491633856027e-06,
+      "loss": 1.0322,
+      "step": 4499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32614493064372e-06,
+      "loss": 0.9614,
+      "step": 4500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32579814466544e-06,
+      "loss": 0.9399,
+      "step": 4501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325451275927827e-06,
+      "loss": 0.9087,
+      "step": 4502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325104324437513e-06,
+      "loss": 0.917,
+      "step": 4503
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324757290201138e-06,
+      "loss": 0.8999,
+      "step": 4504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324410173225342e-06,
+      "loss": 0.8979,
+      "step": 4505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324062973516764e-06,
+      "loss": 0.9507,
+      "step": 4506
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32371569108205e-06,
+      "loss": 0.8687,
+      "step": 4507
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323368325927838e-06,
+      "loss": 0.8936,
+      "step": 4508
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323020878060779e-06,
+      "loss": 0.3696,
+      "step": 4509
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322673347487519e-06,
+      "loss": 0.917,
+      "step": 4510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322325734214705e-06,
+      "loss": 0.9775,
+      "step": 4511
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321978038248987e-06,
+      "loss": 0.8379,
+      "step": 4512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32163025959702e-06,
+      "loss": 0.9287,
+      "step": 4513
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321282398265458e-06,
+      "loss": 0.9126,
+      "step": 4514
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320934454260953e-06,
+      "loss": 0.9297,
+      "step": 4515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320586427590162e-06,
+      "loss": 0.9312,
+      "step": 4516
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320238318259745e-06,
+      "loss": 0.936,
+      "step": 4517
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319890126276361e-06,
+      "loss": 0.3069,
+      "step": 4518
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319541851646671e-06,
+      "loss": 0.9067,
+      "step": 4519
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31919349437734e-06,
+      "loss": 0.9526,
+      "step": 4520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318845054475031e-06,
+      "loss": 0.9272,
+      "step": 4521
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318496531946411e-06,
+      "loss": 0.9438,
+      "step": 4522
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318147926798145e-06,
+      "loss": 0.9023,
+      "step": 4523
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317799239036905e-06,
+      "loss": 0.3676,
+      "step": 4524
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317450468669362e-06,
+      "loss": 0.9229,
+      "step": 4525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317101615702188e-06,
+      "loss": 0.9429,
+      "step": 4526
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31675268014206e-06,
+      "loss": 0.9482,
+      "step": 4527
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316403661995648e-06,
+      "loss": 0.9019,
+      "step": 4528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316054561269631e-06,
+      "loss": 0.9243,
+      "step": 4529
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315705377970692e-06,
+      "loss": 0.9507,
+      "step": 4530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315356112105507e-06,
+      "loss": 0.8896,
+      "step": 4531
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31500676368076e-06,
+      "loss": 0.9556,
+      "step": 4532
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314657332703132e-06,
+      "loss": 0.9287,
+      "step": 4533
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314307819179313e-06,
+      "loss": 0.9502,
+      "step": 4534
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313958223115984e-06,
+      "loss": 0.9019,
+      "step": 4535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313608544519838e-06,
+      "loss": 0.9023,
+      "step": 4536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313258783397564e-06,
+      "loss": 0.9551,
+      "step": 4537
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312908939755852e-06,
+      "loss": 0.8594,
+      "step": 4538
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312559013601395e-06,
+      "loss": 0.9316,
+      "step": 4539
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312209004940887e-06,
+      "loss": 0.8638,
+      "step": 4540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311858913781027e-06,
+      "loss": 0.9209,
+      "step": 4541
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311508740128511e-06,
+      "loss": 0.8809,
+      "step": 4542
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31115848399004e-06,
+      "loss": 0.9365,
+      "step": 4543
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310808145372313e-06,
+      "loss": 0.9053,
+      "step": 4544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310457724282034e-06,
+      "loss": 0.9004,
+      "step": 4545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310107220725905e-06,
+      "loss": 0.9087,
+      "step": 4546
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309756634710632e-06,
+      "loss": 0.8965,
+      "step": 4547
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309405966242926e-06,
+      "loss": 0.8516,
+      "step": 4548
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30905521532949e-06,
+      "loss": 0.981,
+      "step": 4549
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30870438197704e-06,
+      "loss": 0.8608,
+      "step": 4550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308353466192284e-06,
+      "loss": 0.918,
+      "step": 4551
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308002467981938e-06,
+      "loss": 0.9287,
+      "step": 4552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307651387352717e-06,
+      "loss": 1.0015,
+      "step": 4553
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307300224311336e-06,
+      "loss": 0.9229,
+      "step": 4554
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306948978864513e-06,
+      "loss": 0.9092,
+      "step": 4555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306597651018971e-06,
+      "loss": 0.9189,
+      "step": 4556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306246240781428e-06,
+      "loss": 0.9395,
+      "step": 4557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305894748158609e-06,
+      "loss": 0.9341,
+      "step": 4558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305543173157238e-06,
+      "loss": 0.9722,
+      "step": 4559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305191515784043e-06,
+      "loss": 0.9204,
+      "step": 4560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304839776045748e-06,
+      "loss": 0.8789,
+      "step": 4561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304487953949086e-06,
+      "loss": 0.9199,
+      "step": 4562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304136049500786e-06,
+      "loss": 0.325,
+      "step": 4563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303784062707582e-06,
+      "loss": 0.9849,
+      "step": 4564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303431993576206e-06,
+      "loss": 0.9116,
+      "step": 4565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303079842113393e-06,
+      "loss": 0.8755,
+      "step": 4566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302727608325885e-06,
+      "loss": 0.9082,
+      "step": 4567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302375292220416e-06,
+      "loss": 0.9243,
+      "step": 4568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302022893803728e-06,
+      "loss": 0.3632,
+      "step": 4569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301670413082562e-06,
+      "loss": 0.957,
+      "step": 4570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301317850063663e-06,
+      "loss": 0.8718,
+      "step": 4571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300965204753776e-06,
+      "loss": 0.9604,
+      "step": 4572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300612477159648e-06,
+      "loss": 0.9253,
+      "step": 4573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300259667288025e-06,
+      "loss": 0.9248,
+      "step": 4574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.299906775145658e-06,
+      "loss": 0.9414,
+      "step": 4575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2995538007393e-06,
+      "loss": 0.9224,
+      "step": 4576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2992007440757e-06,
+      "loss": 0.8823,
+      "step": 4577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298847605161617e-06,
+      "loss": 0.9336,
+      "step": 4578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298494384003804e-06,
+      "loss": 0.8628,
+      "step": 4579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298141080609023e-06,
+      "loss": 0.4221,
+      "step": 4580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297787694984027e-06,
+      "loss": 0.9155,
+      "step": 4581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29743422713558e-06,
+      "loss": 0.9023,
+      "step": 4582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297080677070445e-06,
+      "loss": 0.9165,
+      "step": 4583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296727044795386e-06,
+      "loss": 0.9429,
+      "step": 4584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296373330317166e-06,
+      "loss": 0.9263,
+      "step": 4585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296019533642555e-06,
+      "loss": 0.8672,
+      "step": 4586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29566565477832e-06,
+      "loss": 0.9077,
+      "step": 4587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295311693731235e-06,
+      "loss": 0.8843,
+      "step": 4588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294957650508065e-06,
+      "loss": 0.9268,
+      "step": 4589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29460352511559e-06,
+      "loss": 0.9575,
+      "step": 4590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294249317560579e-06,
+      "loss": 0.9497,
+      "step": 4591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293895027849814e-06,
+      "loss": 0.8789,
+      "step": 4592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29354065599007e-06,
+      "loss": 0.8662,
+      "step": 4593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293186201988126e-06,
+      "loss": 0.9707,
+      "step": 4594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292831665850769e-06,
+      "loss": 0.9131,
+      "step": 4595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292477047584774e-06,
+      "loss": 0.9058,
+      "step": 4596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292122347196929e-06,
+      "loss": 0.9062,
+      "step": 4597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291767564694021e-06,
+      "loss": 0.9243,
+      "step": 4598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291412700082836e-06,
+      "loss": 0.9316,
+      "step": 4599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291057753370164e-06,
+      "loss": 0.9468,
+      "step": 4600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290702724562794e-06,
+      "loss": 0.9307,
+      "step": 4601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290347613667522e-06,
+      "loss": 0.9692,
+      "step": 4602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289992420691138e-06,
+      "loss": 0.9111,
+      "step": 4603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28963714564044e-06,
+      "loss": 0.9141,
+      "step": 4604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289281788522223e-06,
+      "loss": 0.9126,
+      "step": 4605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288926349343285e-06,
+      "loss": 0.8755,
+      "step": 4606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288570828110428e-06,
+      "loss": 0.9082,
+      "step": 4607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288215224830455e-06,
+      "loss": 0.9565,
+      "step": 4608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287859539510164e-06,
+      "loss": 0.8979,
+      "step": 4609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287503772156365e-06,
+      "loss": 0.9497,
+      "step": 4610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287147922775863e-06,
+      "loss": 0.426,
+      "step": 4611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286791991375465e-06,
+      "loss": 0.937,
+      "step": 4612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286435977961983e-06,
+      "loss": 0.9048,
+      "step": 4613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286079882542223e-06,
+      "loss": 0.9248,
+      "step": 4614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285723705123003e-06,
+      "loss": 0.9365,
+      "step": 4615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285367445711135e-06,
+      "loss": 0.8667,
+      "step": 4616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285011104313434e-06,
+      "loss": 0.296,
+      "step": 4617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284654680936717e-06,
+      "loss": 0.9189,
+      "step": 4618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284298175587807e-06,
+      "loss": 0.9048,
+      "step": 4619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28394158827352e-06,
+      "loss": 0.8994,
+      "step": 4620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28358491900068e-06,
+      "loss": 0.9414,
+      "step": 4621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28322816777611e-06,
+      "loss": 0.9253,
+      "step": 4622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282871334606635e-06,
+      "loss": 0.9629,
+      "step": 4623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282514419499084e-06,
+      "loss": 0.8999,
+      "step": 4624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282157422460284e-06,
+      "loss": 0.9111,
+      "step": 4625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281800343497064e-06,
+      "loss": 0.9648,
+      "step": 4626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281443182616256e-06,
+      "loss": 0.9116,
+      "step": 4627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281085939824694e-06,
+      "loss": 0.9492,
+      "step": 4628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280728615129212e-06,
+      "loss": 0.9165,
+      "step": 4629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280371208536644e-06,
+      "loss": 0.9053,
+      "step": 4630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280013720053831e-06,
+      "loss": 0.937,
+      "step": 4631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27965614968761e-06,
+      "loss": 0.9082,
+      "step": 4632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279298497444824e-06,
+      "loss": 0.9727,
+      "step": 4633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278940763332316e-06,
+      "loss": 0.9658,
+      "step": 4634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278582947356926e-06,
+      "loss": 0.8882,
+      "step": 4635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278225049525502e-06,
+      "loss": 0.854,
+      "step": 4636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277867069844893e-06,
+      "loss": 0.8638,
+      "step": 4637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277509008321943e-06,
+      "loss": 0.9438,
+      "step": 4638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277150864963508e-06,
+      "loss": 0.8579,
+      "step": 4639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276792639776435e-06,
+      "loss": 0.8916,
+      "step": 4640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27643433276758e-06,
+      "loss": 0.8955,
+      "step": 4641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276075943943799e-06,
+      "loss": 0.8643,
+      "step": 4642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275717473311946e-06,
+      "loss": 0.8408,
+      "step": 4643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275358920878878e-06,
+      "loss": 0.9976,
+      "step": 4644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27500028665146e-06,
+      "loss": 0.895,
+      "step": 4645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274641570636549e-06,
+      "loss": 0.916,
+      "step": 4646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274282772841007e-06,
+      "loss": 0.8687,
+      "step": 4647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2739238932717e-06,
+      "loss": 0.936,
+      "step": 4648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273564931935497e-06,
+      "loss": 0.9375,
+      "step": 4649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273205888839261e-06,
+      "loss": 0.2753,
+      "step": 4650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272846763989862e-06,
+      "loss": 0.9399,
+      "step": 4651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272487557394172e-06,
+      "loss": 0.936,
+      "step": 4652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272128269059062e-06,
+      "loss": 0.8872,
+      "step": 4653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271768898991406e-06,
+      "loss": 0.9492,
+      "step": 4654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27140944719808e-06,
+      "loss": 0.9429,
+      "step": 4655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27104991368596e-06,
+      "loss": 0.936,
+      "step": 4656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270690298461925e-06,
+      "loss": 0.8799,
+      "step": 4657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270330601532855e-06,
+      "loss": 0.9019,
+      "step": 4658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269970822905631e-06,
+      "loss": 1.0088,
+      "step": 4659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269610962587136e-06,
+      "loss": 0.9053,
+      "step": 4660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269251020584256e-06,
+      "loss": 0.9629,
+      "step": 4661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268890996903877e-06,
+      "loss": 0.9062,
+      "step": 4662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268530891552885e-06,
+      "loss": 0.9604,
+      "step": 4663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268170704538172e-06,
+      "loss": 0.9023,
+      "step": 4664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267810435866626e-06,
+      "loss": 0.9541,
+      "step": 4665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267450085545143e-06,
+      "loss": 0.9199,
+      "step": 4666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267089653580615e-06,
+      "loss": 0.9531,
+      "step": 4667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266729139979936e-06,
+      "loss": 0.9258,
+      "step": 4668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266368544750006e-06,
+      "loss": 0.9546,
+      "step": 4669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266007867897723e-06,
+      "loss": 0.9004,
+      "step": 4670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265647109429987e-06,
+      "loss": 0.9082,
+      "step": 4671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2652862693537e-06,
+      "loss": 0.8975,
+      "step": 4672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264925347675766e-06,
+      "loss": 0.9468,
+      "step": 4673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264564344403089e-06,
+      "loss": 0.9326,
+      "step": 4674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264203259542576e-06,
+      "loss": 0.4269,
+      "step": 4675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263842093101134e-06,
+      "loss": 0.9673,
+      "step": 4676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263480845085675e-06,
+      "loss": 0.8857,
+      "step": 4677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26311951550311e-06,
+      "loss": 0.9761,
+      "step": 4678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262758104360348e-06,
+      "loss": 0.3955,
+      "step": 4679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262396611664308e-06,
+      "loss": 0.9097,
+      "step": 4680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262035037421903e-06,
+      "loss": 0.9648,
+      "step": 4681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261673381640053e-06,
+      "loss": 0.9775,
+      "step": 4682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261311644325675e-06,
+      "loss": 0.9648,
+      "step": 4683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260949825485692e-06,
+      "loss": 0.856,
+      "step": 4684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260587925127022e-06,
+      "loss": 0.9712,
+      "step": 4685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260225943256591e-06,
+      "loss": 0.8896,
+      "step": 4686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259863879881325e-06,
+      "loss": 0.9214,
+      "step": 4687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259501735008152e-06,
+      "loss": 0.9976,
+      "step": 4688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259139508643997e-06,
+      "loss": 0.9307,
+      "step": 4689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258777200795791e-06,
+      "loss": 0.9316,
+      "step": 4690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258414811470468e-06,
+      "loss": 0.9619,
+      "step": 4691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258052340674958e-06,
+      "loss": 0.8906,
+      "step": 4692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257689788416198e-06,
+      "loss": 0.9282,
+      "step": 4693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25732715470112e-06,
+      "loss": 0.9229,
+      "step": 4694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256964439536668e-06,
+      "loss": 0.9155,
+      "step": 4695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256601642929777e-06,
+      "loss": 0.9517,
+      "step": 4696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256238764887389e-06,
+      "loss": 0.9453,
+      "step": 4697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255875805416447e-06,
+      "loss": 0.9116,
+      "step": 4698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255512764523894e-06,
+      "loss": 0.8628,
+      "step": 4699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255149642216677e-06,
+      "loss": 0.9419,
+      "step": 4700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25478643850174e-06,
+      "loss": 0.9365,
+      "step": 4701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254423153386034e-06,
+      "loss": 0.9302,
+      "step": 4702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254059786876511e-06,
+      "loss": 0.8843,
+      "step": 4703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253696338980119e-06,
+      "loss": 0.8804,
+      "step": 4704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253332809703813e-06,
+      "loss": 0.9077,
+      "step": 4705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25296919905455e-06,
+      "loss": 0.874,
+      "step": 4706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252605507039281e-06,
+      "loss": 1.021,
+      "step": 4707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25224173366497e-06,
+      "loss": 0.9663,
+      "step": 4708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251877878938573e-06,
+      "loss": 0.9102,
+      "step": 4709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251513942867051e-06,
+      "loss": 0.9536,
+      "step": 4710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25114992545737e-06,
+      "loss": 0.8887,
+      "step": 4711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250785826716491e-06,
+      "loss": 0.8921,
+      "step": 4712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250421646651379e-06,
+      "loss": 0.9087,
+      "step": 4713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250057385269005e-06,
+      "loss": 0.9365,
+      "step": 4714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249693042576336e-06,
+      "loss": 0.8579,
+      "step": 4715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249328618580343e-06,
+      "loss": 0.349,
+      "step": 4716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248964113287994e-06,
+      "loss": 0.9702,
+      "step": 4717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24859952670627e-06,
+      "loss": 0.9736,
+      "step": 4718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24823485884214e-06,
+      "loss": 0.9082,
+      "step": 4719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247870109702584e-06,
+      "loss": 0.9067,
+      "step": 4720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247505279294577e-06,
+      "loss": 0.9531,
+      "step": 4721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247140367625102e-06,
+      "loss": 0.9624,
+      "step": 4722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246775374701139e-06,
+      "loss": 0.8931,
+      "step": 4723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24641030052967e-06,
+      "loss": 0.9443,
+      "step": 4724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246045145117683e-06,
+      "loss": 0.9385,
+      "step": 4725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24567990847216e-06,
+      "loss": 0.9199,
+      "step": 4726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24531459060009e-06,
+      "loss": 0.8882,
+      "step": 4727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244949191508462e-06,
+      "loss": 0.9194,
+      "step": 4728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244583711204267e-06,
+      "loss": 0.9277,
+      "step": 4729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244218149694496e-06,
+      "loss": 0.9233,
+      "step": 4730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243852506986145e-06,
+      "loss": 0.9155,
+      "step": 4731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243486783086207e-06,
+      "loss": 0.8813,
+      "step": 4732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24312097800168e-06,
+      "loss": 0.9204,
+      "step": 4733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242755091739564e-06,
+      "loss": 0.8872,
+      "step": 4734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242389124306855e-06,
+      "loss": 0.9229,
+      "step": 4735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242023075710558e-06,
+      "loss": 0.9204,
+      "step": 4736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241656945957676e-06,
+      "loss": 0.9097,
+      "step": 4737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241290735055212e-06,
+      "loss": 0.9019,
+      "step": 4738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24092444301017e-06,
+      "loss": 0.8896,
+      "step": 4739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240558069829565e-06,
+      "loss": 0.9146,
+      "step": 4740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2401916155204e-06,
+      "loss": 0.8638,
+      "step": 4741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239825080089687e-06,
+      "loss": 0.9434,
+      "step": 4742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239458463544439e-06,
+      "loss": 0.9351,
+      "step": 4743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239091765891671e-06,
+      "loss": 0.8843,
+      "step": 4744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238724987138396e-06,
+      "loss": 0.937,
+      "step": 4745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238358127291635e-06,
+      "loss": 0.9604,
+      "step": 4746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237991186358403e-06,
+      "loss": 0.8545,
+      "step": 4747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237624164345722e-06,
+      "loss": 0.9331,
+      "step": 4748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237257061260612e-06,
+      "loss": 0.9072,
+      "step": 4749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236889877110097e-06,
+      "loss": 0.917,
+      "step": 4750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236522611901204e-06,
+      "loss": 0.978,
+      "step": 4751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236155265640955e-06,
+      "loss": 0.8569,
+      "step": 4752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235787838336384e-06,
+      "loss": 0.9556,
+      "step": 4753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235420329994513e-06,
+      "loss": 0.9214,
+      "step": 4754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23505274062238e-06,
+      "loss": 0.8857,
+      "step": 4755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234685070227013e-06,
+      "loss": 0.9341,
+      "step": 4756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234317318815446e-06,
+      "loss": 0.9131,
+      "step": 4757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233949486394718e-06,
+      "loss": 0.8901,
+      "step": 4758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233581572971863e-06,
+      "loss": 0.9375,
+      "step": 4759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233213578553924e-06,
+      "loss": 0.9868,
+      "step": 4760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232845503147937e-06,
+      "loss": 0.8965,
+      "step": 4761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232477346760945e-06,
+      "loss": 0.9097,
+      "step": 4762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232109109399993e-06,
+      "loss": 0.8916,
+      "step": 4763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231740791072124e-06,
+      "loss": 0.8755,
+      "step": 4764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231372391784384e-06,
+      "loss": 0.9268,
+      "step": 4765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231003911543824e-06,
+      "loss": 0.8975,
+      "step": 4766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230635350357493e-06,
+      "loss": 0.8872,
+      "step": 4767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230266708232439e-06,
+      "loss": 0.9111,
+      "step": 4768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229897985175718e-06,
+      "loss": 0.9082,
+      "step": 4769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22952918119438e-06,
+      "loss": 0.9087,
+      "step": 4770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229160296295488e-06,
+      "loss": 0.3306,
+      "step": 4771
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228791330486094e-06,
+      "loss": 0.3461,
+      "step": 4772
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228422283773257e-06,
+      "loss": 0.9844,
+      "step": 4773
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228053156164038e-06,
+      "loss": 0.9326,
+      "step": 4774
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227683947665503e-06,
+      "loss": 0.9453,
+      "step": 4775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227314658284708e-06,
+      "loss": 0.9585,
+      "step": 4776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226945288028725e-06,
+      "loss": 0.979,
+      "step": 4777
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226575836904616e-06,
+      "loss": 0.9395,
+      "step": 4778
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226206304919452e-06,
+      "loss": 0.8726,
+      "step": 4779
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225836692080302e-06,
+      "loss": 0.8984,
+      "step": 4780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225466998394235e-06,
+      "loss": 0.9077,
+      "step": 4781
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225097223868327e-06,
+      "loss": 0.9204,
+      "step": 4782
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224727368509651e-06,
+      "loss": 0.9053,
+      "step": 4783
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224357432325284e-06,
+      "loss": 0.9136,
+      "step": 4784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223987415322304e-06,
+      "loss": 0.8799,
+      "step": 4785
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223617317507785e-06,
+      "loss": 0.9272,
+      "step": 4786
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223247138888813e-06,
+      "loss": 0.9316,
+      "step": 4787
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222876879472468e-06,
+      "loss": 0.8672,
+      "step": 4788
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222506539265835e-06,
+      "loss": 0.9395,
+      "step": 4789
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222136118275996e-06,
+      "loss": 0.9497,
+      "step": 4790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221765616510042e-06,
+      "loss": 0.9448,
+      "step": 4791
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221395033975056e-06,
+      "loss": 0.9282,
+      "step": 4792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221024370678132e-06,
+      "loss": 0.9233,
+      "step": 4793
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220653626626363e-06,
+      "loss": 0.9453,
+      "step": 4794
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220282801826836e-06,
+      "loss": 0.9312,
+      "step": 4795
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.21991189628665e-06,
+      "loss": 0.9053,
+      "step": 4796
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219540910012901e-06,
+      "loss": 0.8975,
+      "step": 4797
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219169843012683e-06,
+      "loss": 0.9097,
+      "step": 4798
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218798695293099e-06,
+      "loss": 0.9766,
+      "step": 4799
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218427466861245e-06,
+      "loss": 0.9077,
+      "step": 4800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.218056157724228e-06,
+      "loss": 0.8975,
+      "step": 4801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217684767889151e-06,
+      "loss": 0.8882,
+      "step": 4802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217313297363118e-06,
+      "loss": 0.9351,
+      "step": 4803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216941746153234e-06,
+      "loss": 0.9097,
+      "step": 4804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216570114266612e-06,
+      "loss": 0.9531,
+      "step": 4805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216198401710357e-06,
+      "loss": 0.9272,
+      "step": 4806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215826608491583e-06,
+      "loss": 0.8921,
+      "step": 4807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215454734617404e-06,
+      "loss": 0.8872,
+      "step": 4808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215082780094932e-06,
+      "loss": 0.9014,
+      "step": 4809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214710744931285e-06,
+      "loss": 0.9478,
+      "step": 4810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214338629133579e-06,
+      "loss": 0.8994,
+      "step": 4811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213966432708935e-06,
+      "loss": 0.3806,
+      "step": 4812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213594155664472e-06,
+      "loss": 0.3459,
+      "step": 4813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213221798007313e-06,
+      "loss": 0.9424,
+      "step": 4814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212849359744581e-06,
+      "loss": 0.3862,
+      "step": 4815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212476840883405e-06,
+      "loss": 0.8569,
+      "step": 4816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212104241430906e-06,
+      "loss": 0.9512,
+      "step": 4817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211731561394217e-06,
+      "loss": 0.8701,
+      "step": 4818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211358800780466e-06,
+      "loss": 0.9595,
+      "step": 4819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210985959596783e-06,
+      "loss": 0.9448,
+      "step": 4820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210613037850306e-06,
+      "loss": 0.4009,
+      "step": 4821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210240035548165e-06,
+      "loss": 0.8857,
+      "step": 4822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209866952697497e-06,
+      "loss": 0.9731,
+      "step": 4823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20949378930544e-06,
+      "loss": 0.9639,
+      "step": 4824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209120545379135e-06,
+      "loss": 0.9043,
+      "step": 4825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208747220925723e-06,
+      "loss": 0.918,
+      "step": 4826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208373815952341e-06,
+      "loss": 0.9258,
+      "step": 4827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208000330466137e-06,
+      "loss": 0.8989,
+      "step": 4828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207626764474256e-06,
+      "loss": 0.9019,
+      "step": 4829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207253117983845e-06,
+      "loss": 0.96,
+      "step": 4830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206879391002053e-06,
+      "loss": 0.9004,
+      "step": 4831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206505583536026e-06,
+      "loss": 0.9146,
+      "step": 4832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20613169559292e-06,
+      "loss": 0.9795,
+      "step": 4833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205757727179889e-06,
+      "loss": 0.9282,
+      "step": 4834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205383678304083e-06,
+      "loss": 0.9146,
+      "step": 4835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205009548972662e-06,
+      "loss": 0.9072,
+      "step": 4836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20463533919278e-06,
+      "loss": 0.8486,
+      "step": 4837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.204261048971599e-06,
+      "loss": 0.9165,
+      "step": 4838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20388667831628e-06,
+      "loss": 0.9194,
+      "step": 4839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203512227233981e-06,
+      "loss": 0.9404,
+      "step": 4840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203137695731873e-06,
+      "loss": 0.8999,
+      "step": 4841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202763083817117e-06,
+      "loss": 0.8999,
+      "step": 4842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202388391496879e-06,
+      "loss": 0.9487,
+      "step": 4843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202013618778329e-06,
+      "loss": 0.9517,
+      "step": 4844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201638765668639e-06,
+      "loss": 0.8989,
+      "step": 4845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201263832174976e-06,
+      "loss": 0.856,
+      "step": 4846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200888818304516e-06,
+      "loss": 0.9336,
+      "step": 4847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200513724064434e-06,
+      "loss": 0.9639,
+      "step": 4848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200138549461905e-06,
+      "loss": 0.9336,
+      "step": 4849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199763294504108e-06,
+      "loss": 0.9814,
+      "step": 4850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199387959198219e-06,
+      "loss": 0.9028,
+      "step": 4851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199012543551422e-06,
+      "loss": 0.9751,
+      "step": 4852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198637047570898e-06,
+      "loss": 0.9639,
+      "step": 4853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19826147126383e-06,
+      "loss": 0.9878,
+      "step": 4854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197885814637406e-06,
+      "loss": 0.8765,
+      "step": 4855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197510077698812e-06,
+      "loss": 0.9492,
+      "step": 4856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197134260455233e-06,
+      "loss": 0.9058,
+      "step": 4857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196758362913865e-06,
+      "loss": 0.9307,
+      "step": 4858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196382385081894e-06,
+      "loss": 0.9385,
+      "step": 4859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196006326966516e-06,
+      "loss": 0.9214,
+      "step": 4860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195630188574926e-06,
+      "loss": 0.9248,
+      "step": 4861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195253969914317e-06,
+      "loss": 0.9097,
+      "step": 4862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19487767099189e-06,
+      "loss": 0.8882,
+      "step": 4863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194501291814842e-06,
+      "loss": 0.9053,
+      "step": 4864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194124832390375e-06,
+      "loss": 0.9497,
+      "step": 4865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19374829272569e-06,
+      "loss": 0.9097,
+      "step": 4866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193371672827994e-06,
+      "loss": 0.8994,
+      "step": 4867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192994972704488e-06,
+      "loss": 0.8945,
+      "step": 4868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192618192362382e-06,
+      "loss": 0.9346,
+      "step": 4869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192241331808883e-06,
+      "loss": 0.8477,
+      "step": 4870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191864391051199e-06,
+      "loss": 0.9722,
+      "step": 4871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191487370096546e-06,
+      "loss": 0.8438,
+      "step": 4872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191110268952132e-06,
+      "loss": 0.9023,
+      "step": 4873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190733087625176e-06,
+      "loss": 0.9146,
+      "step": 4874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190355826122895e-06,
+      "loss": 0.9185,
+      "step": 4875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1899784844525e-06,
+      "loss": 0.9165,
+      "step": 4876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189601062621216e-06,
+      "loss": 0.917,
+      "step": 4877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18922356063626e-06,
+      "loss": 0.9106,
+      "step": 4878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188845978504857e-06,
+      "loss": 0.9312,
+      "step": 4879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18846831623423e-06,
+      "loss": 0.9448,
+      "step": 4880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188090573831603e-06,
+      "loss": 0.9336,
+      "step": 4881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187712751304204e-06,
+      "loss": 0.894,
+      "step": 4882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187334848659261e-06,
+      "loss": 0.9517,
+      "step": 4883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186956865904004e-06,
+      "loss": 0.9858,
+      "step": 4884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186578803045665e-06,
+      "loss": 0.9556,
+      "step": 4885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186200660091475e-06,
+      "loss": 0.9316,
+      "step": 4886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18582243704867e-06,
+      "loss": 0.8677,
+      "step": 4887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185444133924487e-06,
+      "loss": 0.9092,
+      "step": 4888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18506575072616e-06,
+      "loss": 0.8999,
+      "step": 4889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184687287460932e-06,
+      "loss": 0.9312,
+      "step": 4890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184308744136042e-06,
+      "loss": 0.9126,
+      "step": 4891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183930120758732e-06,
+      "loss": 0.9048,
+      "step": 4892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183551417336246e-06,
+      "loss": 0.9448,
+      "step": 4893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183172633875828e-06,
+      "loss": 0.8926,
+      "step": 4894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182793770384726e-06,
+      "loss": 0.9478,
+      "step": 4895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182414826870187e-06,
+      "loss": 0.9336,
+      "step": 4896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182035803339464e-06,
+      "loss": 0.8579,
+      "step": 4897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181656699799804e-06,
+      "loss": 0.8711,
+      "step": 4898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181277516258462e-06,
+      "loss": 0.9492,
+      "step": 4899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180898252722695e-06,
+      "loss": 0.9019,
+      "step": 4900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180518909199752e-06,
+      "loss": 0.9253,
+      "step": 4901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180139485696897e-06,
+      "loss": 0.9678,
+      "step": 4902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179759982221385e-06,
+      "loss": 0.9189,
+      "step": 4903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17938039878048e-06,
+      "loss": 0.9253,
+      "step": 4904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179000735381441e-06,
+      "loss": 0.8857,
+      "step": 4905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178620992031532e-06,
+      "loss": 0.8965,
+      "step": 4906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17824116873802e-06,
+      "loss": 0.8755,
+      "step": 4907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177861265508172e-06,
+      "loss": 0.9146,
+      "step": 4908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177481282349252e-06,
+      "loss": 0.9282,
+      "step": 4909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177101219268532e-06,
+      "loss": 0.8789,
+      "step": 4910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176721076273285e-06,
+      "loss": 0.9326,
+      "step": 4911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17634085337078e-06,
+      "loss": 0.3748,
+      "step": 4912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175960550568294e-06,
+      "loss": 0.8965,
+      "step": 4913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175580167873103e-06,
+      "loss": 0.8823,
+      "step": 4914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175199705292484e-06,
+      "loss": 0.9175,
+      "step": 4915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174819162833715e-06,
+      "loss": 0.9644,
+      "step": 4916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174438540504075e-06,
+      "loss": 0.9224,
+      "step": 4917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174057838310849e-06,
+      "loss": 0.3818,
+      "step": 4918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173677056261318e-06,
+      "loss": 0.8623,
+      "step": 4919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17329619436277e-06,
+      "loss": 0.9233,
+      "step": 4920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172915252622488e-06,
+      "loss": 0.8892,
+      "step": 4921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172534231047763e-06,
+      "loss": 0.8765,
+      "step": 4922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172153129645882e-06,
+      "loss": 0.9619,
+      "step": 4923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171771948424138e-06,
+      "loss": 0.8599,
+      "step": 4924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171390687389821e-06,
+      "loss": 0.9365,
+      "step": 4925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171009346550227e-06,
+      "loss": 0.8926,
+      "step": 4926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170627925912652e-06,
+      "loss": 0.8521,
+      "step": 4927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170246425484392e-06,
+      "loss": 0.9395,
+      "step": 4928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169864845272748e-06,
+      "loss": 0.3337,
+      "step": 4929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16948318528502e-06,
+      "loss": 0.9106,
+      "step": 4930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169101445528506e-06,
+      "loss": 0.9272,
+      "step": 4931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168719626010512e-06,
+      "loss": 0.9277,
+      "step": 4932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16833772673834e-06,
+      "loss": 0.896,
+      "step": 4933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167955747719304e-06,
+      "loss": 0.9478,
+      "step": 4934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167573688960702e-06,
+      "loss": 0.8564,
+      "step": 4935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167191550469851e-06,
+      "loss": 0.9146,
+      "step": 4936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166809332254059e-06,
+      "loss": 0.9756,
+      "step": 4937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166427034320636e-06,
+      "loss": 0.9385,
+      "step": 4938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1660446566769e-06,
+      "loss": 0.9541,
+      "step": 4939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165662199330165e-06,
+      "loss": 0.8818,
+      "step": 4940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165279662287748e-06,
+      "loss": 0.9404,
+      "step": 4941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164897045556967e-06,
+      "loss": 0.9302,
+      "step": 4942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164514349145141e-06,
+      "loss": 0.856,
+      "step": 4943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164131573059594e-06,
+      "loss": 0.9287,
+      "step": 4944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163748717307649e-06,
+      "loss": 0.9292,
+      "step": 4945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163365781896629e-06,
+      "loss": 0.877,
+      "step": 4946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162982766833861e-06,
+      "loss": 0.9253,
+      "step": 4947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162599672126673e-06,
+      "loss": 0.9097,
+      "step": 4948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162216497782393e-06,
+      "loss": 0.9639,
+      "step": 4949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161833243808354e-06,
+      "loss": 0.9258,
+      "step": 4950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161449910211885e-06,
+      "loss": 0.9258,
+      "step": 4951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161066497000324e-06,
+      "loss": 0.8711,
+      "step": 4952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.160683004181e-06,
+      "loss": 0.936,
+      "step": 4953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.160299431761259e-06,
+      "loss": 0.9399,
+      "step": 4954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159915779748429e-06,
+      "loss": 0.957,
+      "step": 4955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159532048149859e-06,
+      "loss": 0.8682,
+      "step": 4956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159148236972883e-06,
+      "loss": 0.8853,
+      "step": 4957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158764346224849e-06,
+      "loss": 0.9009,
+      "step": 4958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158380375913099e-06,
+      "loss": 0.9268,
+      "step": 4959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15799632604498e-06,
+      "loss": 0.8799,
+      "step": 4960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15761219662784e-06,
+      "loss": 0.9531,
+      "step": 4961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157227987669026e-06,
+      "loss": 0.9233,
+      "step": 4962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15684369917589e-06,
+      "loss": 0.8564,
+      "step": 4963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156459331155783e-06,
+      "loss": 0.978,
+      "step": 4964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156074883616059e-06,
+      "loss": 0.979,
+      "step": 4965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155690356564076e-06,
+      "loss": 0.9087,
+      "step": 4966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155305750007186e-06,
+      "loss": 0.9092,
+      "step": 4967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154921063952748e-06,
+      "loss": 0.9346,
+      "step": 4968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154536298408124e-06,
+      "loss": 0.8643,
+      "step": 4969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154151453380672e-06,
+      "loss": 0.9053,
+      "step": 4970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153766528877759e-06,
+      "loss": 0.9404,
+      "step": 4971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153381524906745e-06,
+      "loss": 0.3566,
+      "step": 4972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152996441475e-06,
+      "loss": 0.9453,
+      "step": 4973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152611278589886e-06,
+      "loss": 0.9663,
+      "step": 4974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152226036258776e-06,
+      "loss": 0.9165,
+      "step": 4975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151840714489039e-06,
+      "loss": 0.9502,
+      "step": 4976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151455313288047e-06,
+      "loss": 0.8833,
+      "step": 4977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15106983266317e-06,
+      "loss": 0.9341,
+      "step": 4978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15068427262179e-06,
+      "loss": 0.8979,
+      "step": 4979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.150298633171278e-06,
+      "loss": 0.9565,
+      "step": 4980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149912914319012e-06,
+      "loss": 0.8994,
+      "step": 4981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149527116072372e-06,
+      "loss": 1.0093,
+      "step": 4982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14914123843874e-06,
+      "loss": 0.9512,
+      "step": 4983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148755281425499e-06,
+      "loss": 0.915,
+      "step": 4984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14836924504003e-06,
+      "loss": 0.9556,
+      "step": 4985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147983129289721e-06,
+      "loss": 0.8848,
+      "step": 4986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14759693418196e-06,
+      "loss": 0.3204,
+      "step": 4987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14721065972413e-06,
+      "loss": 0.9077,
+      "step": 4988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146824305923629e-06,
+      "loss": 0.8867,
+      "step": 4989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146437872787842e-06,
+      "loss": 0.9346,
+      "step": 4990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146051360324166e-06,
+      "loss": 0.8945,
+      "step": 4991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145664768539992e-06,
+      "loss": 0.9209,
+      "step": 4992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14527809744272e-06,
+      "loss": 0.9741,
+      "step": 4993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144891347039746e-06,
+      "loss": 0.9277,
+      "step": 4994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144504517338469e-06,
+      "loss": 0.8975,
+      "step": 4995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144117608346289e-06,
+      "loss": 0.9116,
+      "step": 4996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143730620070609e-06,
+      "loss": 0.9546,
+      "step": 4997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143343552518833e-06,
+      "loss": 0.9321,
+      "step": 4998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142956405698365e-06,
+      "loss": 0.9316,
+      "step": 4999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142569179616612e-06,
+      "loss": 0.9707,
+      "step": 5000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142181874280984e-06,
+      "loss": 0.9414,
+      "step": 5001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141794489698888e-06,
+      "loss": 0.9595,
+      "step": 5002
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141407025877737e-06,
+      "loss": 0.8467,
+      "step": 5003
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141019482824946e-06,
+      "loss": 0.9092,
+      "step": 5004
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140631860547924e-06,
+      "loss": 0.9131,
+      "step": 5005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140244159054091e-06,
+      "loss": 0.9341,
+      "step": 5006
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139856378350863e-06,
+      "loss": 0.9312,
+      "step": 5007
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139468518445659e-06,
+      "loss": 0.9663,
+      "step": 5008
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139080579345898e-06,
+      "loss": 0.8428,
+      "step": 5009
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138692561059004e-06,
+      "loss": 0.9277,
+      "step": 5010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1383044635924e-06,
+      "loss": 0.8994,
+      "step": 5011
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13791628695351e-06,
+      "loss": 0.9263,
+      "step": 5012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13752803114976e-06,
+      "loss": 0.8794,
+      "step": 5013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137139696188581e-06,
+      "loss": 0.9175,
+      "step": 5014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136751282077399e-06,
+      "loss": 0.9551,
+      "step": 5015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136362788823648e-06,
+      "loss": 0.9097,
+      "step": 5016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135974216434756e-06,
+      "loss": 0.8887,
+      "step": 5017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13558556491816e-06,
+      "loss": 0.937,
+      "step": 5018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135196834281297e-06,
+      "loss": 0.8452,
+      "step": 5019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1348080245316e-06,
+      "loss": 0.917,
+      "step": 5020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.134419135676511e-06,
+      "loss": 0.9209,
+      "step": 5021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13403016772347e-06,
+      "loss": 0.9062,
+      "step": 5022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133641120679917e-06,
+      "loss": 0.9443,
+      "step": 5023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133251994553295e-06,
+      "loss": 0.9341,
+      "step": 5024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132862789351048e-06,
+      "loss": 0.8867,
+      "step": 5025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132473505080626e-06,
+      "loss": 0.9131,
+      "step": 5026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132084141749473e-06,
+      "loss": 0.8037,
+      "step": 5027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131694699365036e-06,
+      "loss": 0.9341,
+      "step": 5028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131305177934773e-06,
+      "loss": 0.3176,
+      "step": 5029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13091557746613e-06,
+      "loss": 0.9316,
+      "step": 5030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130525897966561e-06,
+      "loss": 0.8857,
+      "step": 5031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130136139443525e-06,
+      "loss": 0.9468,
+      "step": 5032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129746301904476e-06,
+      "loss": 0.9263,
+      "step": 5033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129356385356871e-06,
+      "loss": 0.9434,
+      "step": 5034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128966389808173e-06,
+      "loss": 0.9067,
+      "step": 5035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12857631526584e-06,
+      "loss": 0.937,
+      "step": 5036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128186161737338e-06,
+      "loss": 0.9507,
+      "step": 5037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12779592923013e-06,
+      "loss": 0.9209,
+      "step": 5038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12740561775168e-06,
+      "loss": 0.918,
+      "step": 5039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127015227309455e-06,
+      "loss": 0.9282,
+      "step": 5040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126624757910926e-06,
+      "loss": 0.9536,
+      "step": 5041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126234209563565e-06,
+      "loss": 0.8887,
+      "step": 5042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125843582274838e-06,
+      "loss": 0.9365,
+      "step": 5043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125452876052225e-06,
+      "loss": 0.9126,
+      "step": 5044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125062090903196e-06,
+      "loss": 0.8647,
+      "step": 5045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124671226835228e-06,
+      "loss": 0.9648,
+      "step": 5046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124280283855801e-06,
+      "loss": 0.8916,
+      "step": 5047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123889261972394e-06,
+      "loss": 0.9365,
+      "step": 5048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123498161192485e-06,
+      "loss": 0.8784,
+      "step": 5049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12310698152356e-06,
+      "loss": 0.9492,
+      "step": 5050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1227157229731e-06,
+      "loss": 0.9883,
+      "step": 5051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122324385548592e-06,
+      "loss": 0.939,
+      "step": 5052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121932969257524e-06,
+      "loss": 0.9326,
+      "step": 5053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121541474107381e-06,
+      "loss": 0.8979,
+      "step": 5054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121149900105655e-06,
+      "loss": 0.9834,
+      "step": 5055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120758247259838e-06,
+      "loss": 0.9561,
+      "step": 5056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120366515577422e-06,
+      "loss": 0.8657,
+      "step": 5057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119974705065902e-06,
+      "loss": 0.9722,
+      "step": 5058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119582815732773e-06,
+      "loss": 0.9683,
+      "step": 5059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119190847585533e-06,
+      "loss": 0.9531,
+      "step": 5060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118798800631682e-06,
+      "loss": 0.9165,
+      "step": 5061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118406674878719e-06,
+      "loss": 0.8711,
+      "step": 5062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118014470334146e-06,
+      "loss": 0.9097,
+      "step": 5063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117622187005467e-06,
+      "loss": 0.8833,
+      "step": 5064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117229824900187e-06,
+      "loss": 0.9277,
+      "step": 5065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116837384025814e-06,
+      "loss": 0.8999,
+      "step": 5066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116444864389853e-06,
+      "loss": 0.9243,
+      "step": 5067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116052265999815e-06,
+      "loss": 0.9062,
+      "step": 5068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115659588863213e-06,
+      "loss": 0.9229,
+      "step": 5069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115266832987555e-06,
+      "loss": 0.9009,
+      "step": 5070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11487399838036e-06,
+      "loss": 0.979,
+      "step": 5071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114481085049139e-06,
+      "loss": 1.043,
+      "step": 5072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114088093001412e-06,
+      "loss": 0.9502,
+      "step": 5073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113695022244698e-06,
+      "loss": 0.9971,
+      "step": 5074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113301872786514e-06,
+      "loss": 0.9116,
+      "step": 5075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112908644634385e-06,
+      "loss": 0.8979,
+      "step": 5076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11251533779583e-06,
+      "loss": 0.8813,
+      "step": 5077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11212195227838e-06,
+      "loss": 0.9077,
+      "step": 5078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111728488089552e-06,
+      "loss": 0.9175,
+      "step": 5079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111334945236883e-06,
+      "loss": 0.9277,
+      "step": 5080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110941323727896e-06,
+      "loss": 0.8643,
+      "step": 5081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110547623570123e-06,
+      "loss": 0.9336,
+      "step": 5082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1101538447711e-06,
+      "loss": 0.3164,
+      "step": 5083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109759987338354e-06,
+      "loss": 0.9443,
+      "step": 5084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109366051279422e-06,
+      "loss": 0.9624,
+      "step": 5085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108972036601846e-06,
+      "loss": 0.8608,
+      "step": 5086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108577943313158e-06,
+      "loss": 0.9277,
+      "step": 5087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108183771420898e-06,
+      "loss": 0.939,
+      "step": 5088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10778952093261e-06,
+      "loss": 0.8833,
+      "step": 5089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107395191855837e-06,
+      "loss": 0.9219,
+      "step": 5090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107000784198118e-06,
+      "loss": 0.9126,
+      "step": 5091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106606297967005e-06,
+      "loss": 0.9556,
+      "step": 5092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106211733170041e-06,
+      "loss": 0.896,
+      "step": 5093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105817089814778e-06,
+      "loss": 0.8726,
+      "step": 5094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105422367908762e-06,
+      "loss": 0.9204,
+      "step": 5095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105027567459547e-06,
+      "loss": 0.9385,
+      "step": 5096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.104632688474687e-06,
+      "loss": 0.9019,
+      "step": 5097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.104237730961735e-06,
+      "loss": 0.938,
+      "step": 5098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103842694928248e-06,
+      "loss": 0.9702,
+      "step": 5099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103447580381782e-06,
+      "loss": 0.9453,
+      "step": 5100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1030523873299e-06,
+      "loss": 0.8872,
+      "step": 5101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102657115780159e-06,
+      "loss": 0.9263,
+      "step": 5102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102261765740121e-06,
+      "loss": 0.8716,
+      "step": 5103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101866337217352e-06,
+      "loss": 0.9351,
+      "step": 5104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101470830219417e-06,
+      "loss": 0.8794,
+      "step": 5105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10107524475388e-06,
+      "loss": 0.9521,
+      "step": 5106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100679580828314e-06,
+      "loss": 0.9771,
+      "step": 5107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100283838450283e-06,
+      "loss": 0.957,
+      "step": 5108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099888017627363e-06,
+      "loss": 0.8955,
+      "step": 5109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099492118367123e-06,
+      "loss": 0.9053,
+      "step": 5110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09909614067714e-06,
+      "loss": 0.8618,
+      "step": 5111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09870008456499e-06,
+      "loss": 0.9448,
+      "step": 5112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098303950038246e-06,
+      "loss": 0.939,
+      "step": 5113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097907737104488e-06,
+      "loss": 0.8525,
+      "step": 5114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097511445771301e-06,
+      "loss": 0.8481,
+      "step": 5115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097115076046261e-06,
+      "loss": 0.9062,
+      "step": 5116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096718627936954e-06,
+      "loss": 0.937,
+      "step": 5117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096322101450964e-06,
+      "loss": 0.8711,
+      "step": 5118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095925496595876e-06,
+      "loss": 0.9048,
+      "step": 5119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09552881337928e-06,
+      "loss": 0.9146,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095132051808765e-06,
+      "loss": 0.9458,
+      "step": 5121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09473521189192e-06,
+      "loss": 0.9448,
+      "step": 5122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094338293636337e-06,
+      "loss": 0.9424,
+      "step": 5123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093941297049611e-06,
+      "loss": 0.9312,
+      "step": 5124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093544222139338e-06,
+      "loss": 0.9126,
+      "step": 5125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093147068913111e-06,
+      "loss": 0.9126,
+      "step": 5126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092749837378534e-06,
+      "loss": 0.9546,
+      "step": 5127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0923525275432e-06,
+      "loss": 0.9458,
+      "step": 5128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091955139414714e-06,
+      "loss": 0.9067,
+      "step": 5129
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091557673000679e-06,
+      "loss": 0.9414,
+      "step": 5130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091160128308697e-06,
+      "loss": 0.9395,
+      "step": 5131
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090762505346377e-06,
+      "loss": 0.9712,
+      "step": 5132
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090364804121323e-06,
+      "loss": 0.9087,
+      "step": 5133
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089967024641145e-06,
+      "loss": 0.9409,
+      "step": 5134
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089569166913452e-06,
+      "loss": 0.9961,
+      "step": 5135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089171230945857e-06,
+      "loss": 0.9209,
+      "step": 5136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088773216745971e-06,
+      "loss": 0.9414,
+      "step": 5137
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088375124321413e-06,
+      "loss": 0.9111,
+      "step": 5138
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087976953679795e-06,
+      "loss": 0.9478,
+      "step": 5139
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087578704828737e-06,
+      "loss": 0.9561,
+      "step": 5140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087180377775855e-06,
+      "loss": 0.917,
+      "step": 5141
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086781972528773e-06,
+      "loss": 0.9321,
+      "step": 5142
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086383489095111e-06,
+      "loss": 0.8931,
+      "step": 5143
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085984927482497e-06,
+      "loss": 0.9756,
+      "step": 5144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085586287698549e-06,
+      "loss": 0.8965,
+      "step": 5145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085187569750898e-06,
+      "loss": 0.8662,
+      "step": 5146
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084788773647172e-06,
+      "loss": 0.9297,
+      "step": 5147
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084389899395e-06,
+      "loss": 0.9404,
+      "step": 5148
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083990947002013e-06,
+      "loss": 0.9438,
+      "step": 5149
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083591916475842e-06,
+      "loss": 0.8662,
+      "step": 5150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083192807824124e-06,
+      "loss": 0.9258,
+      "step": 5151
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082793621054492e-06,
+      "loss": 0.9248,
+      "step": 5152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082394356174586e-06,
+      "loss": 0.9351,
+      "step": 5153
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081995013192044e-06,
+      "loss": 0.897,
+      "step": 5154
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081595592114503e-06,
+      "loss": 0.9058,
+      "step": 5155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081196092949607e-06,
+      "loss": 0.9517,
+      "step": 5156
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080796515704997e-06,
+      "loss": 0.9277,
+      "step": 5157
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080396860388322e-06,
+      "loss": 0.8857,
+      "step": 5158
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079997127007223e-06,
+      "loss": 0.8628,
+      "step": 5159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07959731556935e-06,
+      "loss": 0.9238,
+      "step": 5160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079197426082353e-06,
+      "loss": 0.8799,
+      "step": 5161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07879745855388e-06,
+      "loss": 0.9185,
+      "step": 5162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078397412991586e-06,
+      "loss": 0.9297,
+      "step": 5163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077997289403122e-06,
+      "loss": 0.8818,
+      "step": 5164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077597087796142e-06,
+      "loss": 0.9478,
+      "step": 5165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077196808178305e-06,
+      "loss": 0.8931,
+      "step": 5166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076796450557269e-06,
+      "loss": 0.9473,
+      "step": 5167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076396014940691e-06,
+      "loss": 0.9297,
+      "step": 5168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075995501336237e-06,
+      "loss": 0.3679,
+      "step": 5169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075594909751561e-06,
+      "loss": 0.9243,
+      "step": 5170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075194240194337e-06,
+      "loss": 0.9062,
+      "step": 5171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074793492672222e-06,
+      "loss": 0.9268,
+      "step": 5172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074392667192888e-06,
+      "loss": 0.9468,
+      "step": 5173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073991763764003e-06,
+      "loss": 0.8604,
+      "step": 5174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073590782393234e-06,
+      "loss": 0.896,
+      "step": 5175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073189723088254e-06,
+      "loss": 0.9746,
+      "step": 5176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072788585856738e-06,
+      "loss": 0.8994,
+      "step": 5177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072387370706355e-06,
+      "loss": 0.8872,
+      "step": 5178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071986077644789e-06,
+      "loss": 0.9487,
+      "step": 5179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071584706679707e-06,
+      "loss": 0.9521,
+      "step": 5180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071183257818798e-06,
+      "loss": 0.875,
+      "step": 5181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070781731069737e-06,
+      "loss": 0.9536,
+      "step": 5182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070380126440205e-06,
+      "loss": 0.9072,
+      "step": 5183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069978443937888e-06,
+      "loss": 0.9106,
+      "step": 5184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06957668357047e-06,
+      "loss": 0.8691,
+      "step": 5185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069174845345639e-06,
+      "loss": 0.8662,
+      "step": 5186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068772929271078e-06,
+      "loss": 0.9419,
+      "step": 5187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068370935354481e-06,
+      "loss": 0.9487,
+      "step": 5188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067968863603539e-06,
+      "loss": 0.9678,
+      "step": 5189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067566714025938e-06,
+      "loss": 0.9697,
+      "step": 5190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067164486629378e-06,
+      "loss": 0.9277,
+      "step": 5191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066762181421552e-06,
+      "loss": 0.2963,
+      "step": 5192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066359798410157e-06,
+      "loss": 0.8726,
+      "step": 5193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065957337602892e-06,
+      "loss": 0.9062,
+      "step": 5194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065554799007457e-06,
+      "loss": 0.9253,
+      "step": 5195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06515218263155e-06,
+      "loss": 0.8574,
+      "step": 5196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064749488482876e-06,
+      "loss": 0.8994,
+      "step": 5197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064346716569141e-06,
+      "loss": 0.9229,
+      "step": 5198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063943866898046e-06,
+      "loss": 0.9717,
+      "step": 5199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063540939477303e-06,
+      "loss": 0.9136,
+      "step": 5200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063137934314618e-06,
+      "loss": 0.8872,
+      "step": 5201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062734851417702e-06,
+      "loss": 0.9321,
+      "step": 5202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062331690794265e-06,
+      "loss": 0.9399,
+      "step": 5203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061928452452022e-06,
+      "loss": 0.8667,
+      "step": 5204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061525136398687e-06,
+      "loss": 0.9229,
+      "step": 5205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061121742641975e-06,
+      "loss": 0.9438,
+      "step": 5206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060718271189606e-06,
+      "loss": 0.8579,
+      "step": 5207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060314722049297e-06,
+      "loss": 0.8936,
+      "step": 5208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059911095228768e-06,
+      "loss": 0.8564,
+      "step": 5209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059507390735744e-06,
+      "loss": 0.918,
+      "step": 5210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059103608577946e-06,
+      "loss": 0.4216,
+      "step": 5211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0586997487631e-06,
+      "loss": 0.9287,
+      "step": 5212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05829581129893e-06,
+      "loss": 0.9131,
+      "step": 5213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057891796193168e-06,
+      "loss": 0.8989,
+      "step": 5214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05748770345354e-06,
+      "loss": 0.9048,
+      "step": 5215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05708353308778e-06,
+      "loss": 0.9521,
+      "step": 5216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056679285103618e-06,
+      "loss": 0.9438,
+      "step": 5217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05627495950879e-06,
+      "loss": 0.9072,
+      "step": 5218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.055870556311028e-06,
+      "loss": 0.8472,
+      "step": 5219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.055466075518073e-06,
+      "loss": 0.8779,
+      "step": 5220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05506151713766e-06,
+      "loss": 1.0229,
+      "step": 5221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054656881177531e-06,
+      "loss": 0.915,
+      "step": 5222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054252167645426e-06,
+      "loss": 0.8887,
+      "step": 5223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05384737654909e-06,
+      "loss": 0.9146,
+      "step": 5224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053442507896261e-06,
+      "loss": 0.9028,
+      "step": 5225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053037561694694e-06,
+      "loss": 0.9321,
+      "step": 5226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05263253795213e-06,
+      "loss": 0.9023,
+      "step": 5227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05222743667632e-06,
+      "loss": 0.9033,
+      "step": 5228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051822257875014e-06,
+      "loss": 0.9326,
+      "step": 5229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051417001555962e-06,
+      "loss": 0.9111,
+      "step": 5230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051011667726918e-06,
+      "loss": 0.9097,
+      "step": 5231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050606256395639e-06,
+      "loss": 0.915,
+      "step": 5232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05020076756988e-06,
+      "loss": 0.9258,
+      "step": 5233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049795201257396e-06,
+      "loss": 0.9155,
+      "step": 5234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049389557465949e-06,
+      "loss": 0.9443,
+      "step": 5235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048983836203298e-06,
+      "loss": 0.9863,
+      "step": 5236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048578037477206e-06,
+      "loss": 0.9043,
+      "step": 5237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048172161295437e-06,
+      "loss": 0.8618,
+      "step": 5238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047766207665753e-06,
+      "loss": 0.9302,
+      "step": 5239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047360176595927e-06,
+      "loss": 0.9629,
+      "step": 5240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046954068093721e-06,
+      "loss": 0.8857,
+      "step": 5241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046547882166906e-06,
+      "loss": 0.3464,
+      "step": 5242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046141618823254e-06,
+      "loss": 0.9668,
+      "step": 5243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045735278070537e-06,
+      "loss": 0.9146,
+      "step": 5244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04532885991653e-06,
+      "loss": 0.8989,
+      "step": 5245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044922364369004e-06,
+      "loss": 0.8501,
+      "step": 5246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044515791435741e-06,
+      "loss": 0.9575,
+      "step": 5247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044109141124517e-06,
+      "loss": 0.9473,
+      "step": 5248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043702413443114e-06,
+      "loss": 0.8198,
+      "step": 5249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04329560839931e-06,
+      "loss": 0.936,
+      "step": 5250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04288872600089e-06,
+      "loss": 0.936,
+      "step": 5251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04248176625564e-06,
+      "loss": 0.9189,
+      "step": 5252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.042074729171342e-06,
+      "loss": 0.877,
+      "step": 5253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041667614755783e-06,
+      "loss": 0.8892,
+      "step": 5254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041260423016756e-06,
+      "loss": 0.9556,
+      "step": 5255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040853153962048e-06,
+      "loss": 0.8735,
+      "step": 5256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040445807599453e-06,
+      "loss": 0.9478,
+      "step": 5257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04003838393676e-06,
+      "loss": 0.9414,
+      "step": 5258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039630882981769e-06,
+      "loss": 1.0181,
+      "step": 5259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039223304742272e-06,
+      "loss": 0.9282,
+      "step": 5260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03881564922607e-06,
+      "loss": 0.9536,
+      "step": 5261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038407916440959e-06,
+      "loss": 0.9141,
+      "step": 5262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03800010639474e-06,
+      "loss": 0.9395,
+      "step": 5263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.037592219095217e-06,
+      "loss": 0.8833,
+      "step": 5264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.037184254550192e-06,
+      "loss": 0.9365,
+      "step": 5265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03677621276747e-06,
+      "loss": 0.9307,
+      "step": 5266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03636809375486e-06,
+      "loss": 0.9727,
+      "step": 5267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.035959897520166e-06,
+      "loss": 0.8979,
+      "step": 5268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.0355516240712e-06,
+      "loss": 0.8979,
+      "step": 5269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03514327341577e-06,
+      "loss": 0.8936,
+      "step": 5270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034734845561693e-06,
+      "loss": 0.8838,
+      "step": 5271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03432634051678e-06,
+      "loss": 0.8936,
+      "step": 5272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033917758288846e-06,
+      "loss": 0.9224,
+      "step": 5273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033509098885708e-06,
+      "loss": 0.8784,
+      "step": 5274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033100362315187e-06,
+      "loss": 0.8462,
+      "step": 5275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.0326915485851e-06,
+      "loss": 0.9478,
+      "step": 5276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032282657703268e-06,
+      "loss": 0.8848,
+      "step": 5277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031873689677514e-06,
+      "loss": 0.9644,
+      "step": 5278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031464644515667e-06,
+      "loss": 0.9253,
+      "step": 5279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031055522225545e-06,
+      "loss": 0.9009,
+      "step": 5280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030646322814978e-06,
+      "loss": 0.9043,
+      "step": 5281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030237046291799e-06,
+      "loss": 0.916,
+      "step": 5282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02982769266383e-06,
+      "loss": 0.8896,
+      "step": 5283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029418261938911e-06,
+      "loss": 0.8921,
+      "step": 5284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029008754124868e-06,
+      "loss": 0.9004,
+      "step": 5285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02859916922954e-06,
+      "loss": 0.9165,
+      "step": 5286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028189507260762e-06,
+      "loss": 0.9351,
+      "step": 5287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027779768226372e-06,
+      "loss": 0.9019,
+      "step": 5288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027369952134206e-06,
+      "loss": 0.8848,
+      "step": 5289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026960058992107e-06,
+      "loss": 0.8813,
+      "step": 5290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026550088807916e-06,
+      "loss": 0.9736,
+      "step": 5291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026140041589478e-06,
+      "loss": 0.9448,
+      "step": 5292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025729917344636e-06,
+      "loss": 0.8838,
+      "step": 5293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025319716081235e-06,
+      "loss": 0.9048,
+      "step": 5294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024909437807128e-06,
+      "loss": 0.8838,
+      "step": 5295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02449908253016e-06,
+      "loss": 0.9517,
+      "step": 5296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024088650258182e-06,
+      "loss": 0.9136,
+      "step": 5297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023678140999047e-06,
+      "loss": 0.875,
+      "step": 5298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023267554760609e-06,
+      "loss": 0.4122,
+      "step": 5299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022856891550721e-06,
+      "loss": 0.8965,
+      "step": 5300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022446151377244e-06,
+      "loss": 0.9155,
+      "step": 5301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022035334248032e-06,
+      "loss": 0.2855,
+      "step": 5302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021624440170948e-06,
+      "loss": 0.9492,
+      "step": 5303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021213469153848e-06,
+      "loss": 0.9351,
+      "step": 5304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020802421204601e-06,
+      "loss": 0.8892,
+      "step": 5305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020391296331066e-06,
+      "loss": 0.9121,
+      "step": 5306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01998009454111e-06,
+      "loss": 0.9404,
+      "step": 5307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.0195688158426e-06,
+      "loss": 0.9971,
+      "step": 5308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019157460243406e-06,
+      "loss": 0.9429,
+      "step": 5309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018746027751394e-06,
+      "loss": 0.9297,
+      "step": 5310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01833451837444e-06,
+      "loss": 0.8921,
+      "step": 5311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017922932120413e-06,
+      "loss": 0.9141,
+      "step": 5312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01751126899719e-06,
+      "loss": 0.8682,
+      "step": 5313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017099529012644e-06,
+      "loss": 0.9375,
+      "step": 5314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016687712174657e-06,
+      "loss": 0.9116,
+      "step": 5315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016275818491102e-06,
+      "loss": 0.8745,
+      "step": 5316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015863847969862e-06,
+      "loss": 0.9214,
+      "step": 5317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015451800618818e-06,
+      "loss": 0.9072,
+      "step": 5318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015039676445856e-06,
+      "loss": 0.9082,
+      "step": 5319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014627475458855e-06,
+      "loss": 0.9321,
+      "step": 5320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014215197665706e-06,
+      "loss": 0.9453,
+      "step": 5321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013802843074294e-06,
+      "loss": 0.9517,
+      "step": 5322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013390411692509e-06,
+      "loss": 0.9229,
+      "step": 5323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01297790352824e-06,
+      "loss": 0.8545,
+      "step": 5324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01256531858938e-06,
+      "loss": 0.8867,
+      "step": 5325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012152656883824e-06,
+      "loss": 0.9233,
+      "step": 5326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011739918419465e-06,
+      "loss": 0.9146,
+      "step": 5327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011327103204199e-06,
+      "loss": 0.9272,
+      "step": 5328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010914211245925e-06,
+      "loss": 0.9204,
+      "step": 5329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01050124255254e-06,
+      "loss": 0.8667,
+      "step": 5330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010088197131947e-06,
+      "loss": 0.9463,
+      "step": 5331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00967507499205e-06,
+      "loss": 0.9595,
+      "step": 5332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009261876140749e-06,
+      "loss": 0.8232,
+      "step": 5333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00884860058595e-06,
+      "loss": 0.8223,
+      "step": 5334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00843524833556e-06,
+      "loss": 0.9927,
+      "step": 5335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.008021819397488e-06,
+      "loss": 0.8975,
+      "step": 5336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007608313779642e-06,
+      "loss": 0.8887,
+      "step": 5337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007194731489934e-06,
+      "loss": 0.9146,
+      "step": 5338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006781072536278e-06,
+      "loss": 0.8936,
+      "step": 5339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006367336926583e-06,
+      "loss": 0.8975,
+      "step": 5340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005953524668768e-06,
+      "loss": 0.8765,
+      "step": 5341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005539635770752e-06,
+      "loss": 0.8882,
+      "step": 5342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005125670240449e-06,
+      "loss": 0.9287,
+      "step": 5343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00471162808578e-06,
+      "loss": 0.8994,
+      "step": 5344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004297509314668e-06,
+      "loss": 0.8779,
+      "step": 5345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003883313935036e-06,
+      "loss": 0.8833,
+      "step": 5346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003469041954804e-06,
+      "loss": 0.9219,
+      "step": 5347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003054693381904e-06,
+      "loss": 0.9272,
+      "step": 5348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002640268224257e-06,
+      "loss": 0.9126,
+      "step": 5349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002225766489794e-06,
+      "loss": 0.9673,
+      "step": 5350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.001811188186447e-06,
+      "loss": 0.9038,
+      "step": 5351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.001396533322146e-06,
+      "loss": 0.8906,
+      "step": 5352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000981801904823e-06,
+      "loss": 0.9219,
+      "step": 5353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000566993942413e-06,
+      "loss": 0.9204,
+      "step": 5354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000152109442853e-06,
+      "loss": 0.9775,
+      "step": 5355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99973714841408e-06,
+      "loss": 0.9038,
+      "step": 5356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99932211086403e-06,
+      "loss": 0.9189,
+      "step": 5357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998906996800648e-06,
+      "loss": 0.9023,
+      "step": 5358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998491806231876e-06,
+      "loss": 0.9995,
+      "step": 5359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998076539165652e-06,
+      "loss": 0.8921,
+      "step": 5360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997661195609923e-06,
+      "loss": 0.8936,
+      "step": 5361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997245775572637e-06,
+      "loss": 0.9121,
+      "step": 5362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99683027906174e-06,
+      "loss": 1.0151,
+      "step": 5363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996414706085183e-06,
+      "loss": 0.9404,
+      "step": 5364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995999056650914e-06,
+      "loss": 0.8945,
+      "step": 5365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995583330766887e-06,
+      "loss": 0.957,
+      "step": 5366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995167528441053e-06,
+      "loss": 0.9644,
+      "step": 5367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994751649681369e-06,
+      "loss": 0.8945,
+      "step": 5368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994335694495792e-06,
+      "loss": 0.8843,
+      "step": 5369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993919662892278e-06,
+      "loss": 0.9131,
+      "step": 5370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993503554878786e-06,
+      "loss": 0.9272,
+      "step": 5371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99308737046328e-06,
+      "loss": 0.8413,
+      "step": 5372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99267110965372e-06,
+      "loss": 0.8413,
+      "step": 5373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.992254772458069e-06,
+      "loss": 0.8828,
+      "step": 5374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991838358884295e-06,
+      "loss": 0.8584,
+      "step": 5375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991421868940361e-06,
+      "loss": 0.9067,
+      "step": 5376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991005302634238e-06,
+      "loss": 0.8979,
+      "step": 5377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.990588659973895e-06,
+      "loss": 0.9922,
+      "step": 5378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9901719409673e-06,
+      "loss": 0.896,
+      "step": 5379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989755145622431e-06,
+      "loss": 0.9482,
+      "step": 5380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989338273947258e-06,
+      "loss": 0.9287,
+      "step": 5381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988921325949756e-06,
+      "loss": 0.9258,
+      "step": 5382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988504301637905e-06,
+      "loss": 0.9233,
+      "step": 5383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988087201019683e-06,
+      "loss": 0.8745,
+      "step": 5384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987670024103066e-06,
+      "loss": 0.9487,
+      "step": 5385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987252770896037e-06,
+      "loss": 0.9204,
+      "step": 5386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986835441406581e-06,
+      "loss": 0.9448,
+      "step": 5387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986418035642679e-06,
+      "loss": 0.9712,
+      "step": 5388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98600055361232e-06,
+      "loss": 0.9355,
+      "step": 5389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985582995323488e-06,
+      "loss": 0.9473,
+      "step": 5390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985165360784174e-06,
+      "loss": 0.9316,
+      "step": 5391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984747650002366e-06,
+      "loss": 0.9053,
+      "step": 5392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984329862986056e-06,
+      "loss": 0.9141,
+      "step": 5393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983911999743237e-06,
+      "loss": 0.875,
+      "step": 5394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983494060281905e-06,
+      "loss": 0.9595,
+      "step": 5395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983076044610052e-06,
+      "loss": 0.9556,
+      "step": 5396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98265795273568e-06,
+      "loss": 0.9424,
+      "step": 5397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982239784666786e-06,
+      "loss": 0.9302,
+      "step": 5398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981821540411369e-06,
+      "loss": 0.8818,
+      "step": 5399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98140321997743e-06,
+      "loss": 0.9409,
+      "step": 5400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980984823372973e-06,
+      "loss": 0.8711,
+      "step": 5401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980566350606006e-06,
+      "loss": 0.9126,
+      "step": 5402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980147801684529e-06,
+      "loss": 0.8491,
+      "step": 5403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979729176616553e-06,
+      "loss": 0.8882,
+      "step": 5404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979310475410089e-06,
+      "loss": 0.9033,
+      "step": 5405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978891698073144e-06,
+      "loss": 0.9434,
+      "step": 5406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97847284461373e-06,
+      "loss": 0.9346,
+      "step": 5407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978053915039862e-06,
+      "loss": 0.9297,
+      "step": 5408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977634909359554e-06,
+      "loss": 0.8809,
+      "step": 5409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977215827580822e-06,
+      "loss": 0.9341,
+      "step": 5410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976796669711685e-06,
+      "loss": 0.8384,
+      "step": 5411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976377435760158e-06,
+      "loss": 0.9512,
+      "step": 5412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975958125734267e-06,
+      "loss": 0.9175,
+      "step": 5413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97553873964203e-06,
+      "loss": 0.876,
+      "step": 5414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975119277491475e-06,
+      "loss": 0.959,
+      "step": 5415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974699739290622e-06,
+      "loss": 0.9414,
+      "step": 5416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974280125047501e-06,
+      "loss": 0.9141,
+      "step": 5417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973860434770138e-06,
+      "loss": 0.9458,
+      "step": 5418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973440668466563e-06,
+      "loss": 0.9282,
+      "step": 5419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973020826144805e-06,
+      "loss": 0.896,
+      "step": 5420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9726009078129e-06,
+      "loss": 0.9131,
+      "step": 5421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972180913478878e-06,
+      "loss": 0.9136,
+      "step": 5422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971760843150777e-06,
+      "loss": 0.895,
+      "step": 5423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971340696836632e-06,
+      "loss": 0.8198,
+      "step": 5424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970920474544482e-06,
+      "loss": 0.853,
+      "step": 5425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970500176282366e-06,
+      "loss": 0.9048,
+      "step": 5426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970079802058322e-06,
+      "loss": 0.8926,
+      "step": 5427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969659351880398e-06,
+      "loss": 0.9258,
+      "step": 5428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969238825756636e-06,
+      "loss": 0.9365,
+      "step": 5429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96881822369508e-06,
+      "loss": 0.957,
+      "step": 5430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.968397545703775e-06,
+      "loss": 0.9849,
+      "step": 5431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967976791790774e-06,
+      "loss": 0.8076,
+      "step": 5432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967555961964123e-06,
+      "loss": 0.8589,
+      "step": 5433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967135056231875e-06,
+      "loss": 0.8677,
+      "step": 5434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966714074602081e-06,
+      "loss": 0.9609,
+      "step": 5435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966293017082796e-06,
+      "loss": 0.8594,
+      "step": 5436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965871883682076e-06,
+      "loss": 0.9307,
+      "step": 5437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965450674407979e-06,
+      "loss": 0.8223,
+      "step": 5438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965029389268559e-06,
+      "loss": 0.8711,
+      "step": 5439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96460802827188e-06,
+      "loss": 0.8276,
+      "step": 5440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.964186591426003e-06,
+      "loss": 0.8794,
+      "step": 5441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.963765078738988e-06,
+      "loss": 0.9126,
+      "step": 5442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9633434902189e-06,
+      "loss": 0.9414,
+      "step": 5443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962921825873808e-06,
+      "loss": 0.9102,
+      "step": 5444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962500085711777e-06,
+      "loss": 0.9189,
+      "step": 5445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962078269740873e-06,
+      "loss": 0.8574,
+      "step": 5446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96165637796917e-06,
+      "loss": 0.8794,
+      "step": 5447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961234410404739e-06,
+      "loss": 0.9419,
+      "step": 5448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960812367055646e-06,
+      "loss": 0.8887,
+      "step": 5449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960390247929975e-06,
+      "loss": 0.9121,
+      "step": 5450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959968053035799e-06,
+      "loss": 0.9409,
+      "step": 5451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959545782381192e-06,
+      "loss": 0.9673,
+      "step": 5452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959123435974236e-06,
+      "loss": 0.8965,
+      "step": 5453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958701013823007e-06,
+      "loss": 0.9346,
+      "step": 5454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958278515935591e-06,
+      "loss": 0.8115,
+      "step": 5455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95785594232007e-06,
+      "loss": 0.9917,
+      "step": 5456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95743329298453e-06,
+      "loss": 0.9165,
+      "step": 5457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957010567937053e-06,
+      "loss": 0.9053,
+      "step": 5458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956587767185727e-06,
+      "loss": 0.9316,
+      "step": 5459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956164890738643e-06,
+      "loss": 0.8469,
+      "step": 5460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955741938603892e-06,
+      "loss": 0.9434,
+      "step": 5461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955318910789564e-06,
+      "loss": 0.9331,
+      "step": 5462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95489580730375e-06,
+      "loss": 0.9458,
+      "step": 5463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954472628154549e-06,
+      "loss": 0.9116,
+      "step": 5464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954049373350056e-06,
+      "loss": 0.8691,
+      "step": 5465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953626042898366e-06,
+      "loss": 0.9521,
+      "step": 5466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95320263680758e-06,
+      "loss": 0.9297,
+      "step": 5467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952779155085798e-06,
+      "loss": 0.8984,
+      "step": 5468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952355597741123e-06,
+      "loss": 0.8896,
+      "step": 5469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951931964781657e-06,
+      "loss": 0.9087,
+      "step": 5470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951508256215504e-06,
+      "loss": 0.9453,
+      "step": 5471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951084472050771e-06,
+      "loss": 0.9121,
+      "step": 5472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950660612295568e-06,
+      "loss": 0.9365,
+      "step": 5473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950236676958e-06,
+      "loss": 0.3831,
+      "step": 5474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94981266604618e-06,
+      "loss": 0.9507,
+      "step": 5475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.949388579568221e-06,
+      "loss": 0.9507,
+      "step": 5476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948964417532233e-06,
+      "loss": 0.9722,
+      "step": 5477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948540179946334e-06,
+      "loss": 0.3815,
+      "step": 5478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948115866818638e-06,
+      "loss": 0.9424,
+      "step": 5479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947691478157266e-06,
+      "loss": 0.916,
+      "step": 5480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947267013970334e-06,
+      "loss": 0.9995,
+      "step": 5481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946842474265965e-06,
+      "loss": 0.9214,
+      "step": 5482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946417859052278e-06,
+      "loss": 0.876,
+      "step": 5483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945993168337401e-06,
+      "loss": 0.9062,
+      "step": 5484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945568402129454e-06,
+      "loss": 0.856,
+      "step": 5485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945143560436569e-06,
+      "loss": 0.9253,
+      "step": 5486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944718643266868e-06,
+      "loss": 0.9302,
+      "step": 5487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944293650628485e-06,
+      "loss": 0.9292,
+      "step": 5488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943868582529549e-06,
+      "loss": 0.8149,
+      "step": 5489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943443438978192e-06,
+      "loss": 0.8901,
+      "step": 5490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943018219982547e-06,
+      "loss": 0.8857,
+      "step": 5491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942592925550751e-06,
+      "loss": 0.978,
+      "step": 5492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942167555690939e-06,
+      "loss": 0.8433,
+      "step": 5493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94174211041125e-06,
+      "loss": 0.8521,
+      "step": 5494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941316589719822e-06,
+      "loss": 0.9829,
+      "step": 5495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9408909936248e-06,
+      "loss": 0.9346,
+      "step": 5496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94046532213432e-06,
+      "loss": 0.8677,
+      "step": 5497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94003957525653e-06,
+      "loss": 0.9194,
+      "step": 5498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939613752999575e-06,
+      "loss": 0.9053,
+      "step": 5499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939187855371599e-06,
+      "loss": 0.896,
+      "step": 5500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938761882380753e-06,
+      "loss": 0.9678,
+      "step": 5501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938335834035186e-06,
+      "loss": 0.9233,
+      "step": 5502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937909710343049e-06,
+      "loss": 0.9248,
+      "step": 5503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937483511312493e-06,
+      "loss": 0.9507,
+      "step": 5504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937057236951672e-06,
+      "loss": 0.9243,
+      "step": 5505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936630887268742e-06,
+      "loss": 0.894,
+      "step": 5506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93620446227186e-06,
+      "loss": 0.8901,
+      "step": 5507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935777961969186e-06,
+      "loss": 0.9058,
+      "step": 5508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935351386368875e-06,
+      "loss": 0.3864,
+      "step": 5509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934924735479091e-06,
+      "loss": 0.9019,
+      "step": 5510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934498009307998e-06,
+      "loss": 0.9404,
+      "step": 5511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934071207863754e-06,
+      "loss": 0.9116,
+      "step": 5512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933644331154531e-06,
+      "loss": 0.9351,
+      "step": 5513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933217379188494e-06,
+      "loss": 1.0059,
+      "step": 5514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93279035197381e-06,
+      "loss": 0.9243,
+      "step": 5515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93236324951865e-06,
+      "loss": 0.9761,
+      "step": 5516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931936071831184e-06,
+      "loss": 0.8984,
+      "step": 5517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931508818919586e-06,
+      "loss": 0.9395,
+      "step": 5518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931081490792026e-06,
+      "loss": 0.9175,
+      "step": 5519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930654087456685e-06,
+      "loss": 0.8623,
+      "step": 5520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930226608921736e-06,
+      "loss": 0.8887,
+      "step": 5521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92979905519536e-06,
+      "loss": 0.9116,
+      "step": 5522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929371426285733e-06,
+      "loss": 0.8789,
+      "step": 5523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928943722201041e-06,
+      "loss": 0.9482,
+      "step": 5524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928515942949464e-06,
+      "loss": 0.9458,
+      "step": 5525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928088088539186e-06,
+      "loss": 0.9434,
+      "step": 5526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927660158978392e-06,
+      "loss": 0.8682,
+      "step": 5527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927232154275273e-06,
+      "loss": 0.896,
+      "step": 5528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926804074438011e-06,
+      "loss": 0.9326,
+      "step": 5529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926375919474801e-06,
+      "loss": 0.9512,
+      "step": 5530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925947689393833e-06,
+      "loss": 0.9297,
+      "step": 5531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925519384203298e-06,
+      "loss": 0.8774,
+      "step": 5532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925091003911392e-06,
+      "loss": 0.8726,
+      "step": 5533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924662548526308e-06,
+      "loss": 0.8608,
+      "step": 5534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924234018056248e-06,
+      "loss": 0.9263,
+      "step": 5535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923805412509405e-06,
+      "loss": 0.9023,
+      "step": 5536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923376731893983e-06,
+      "loss": 0.9731,
+      "step": 5537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922947976218181e-06,
+      "loss": 0.9019,
+      "step": 5538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9225191454902e-06,
+      "loss": 0.9346,
+      "step": 5539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92209023971825e-06,
+      "loss": 0.8711,
+      "step": 5540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921661258910533e-06,
+      "loss": 0.8784,
+      "step": 5541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921232203075256e-06,
+      "loss": 1.0068,
+      "step": 5542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920803072220627e-06,
+      "loss": 0.3851,
+      "step": 5543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920373866354856e-06,
+      "loss": 0.9688,
+      "step": 5544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919944585486157e-06,
+      "loss": 0.9453,
+      "step": 5545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91951522962274e-06,
+      "loss": 0.9077,
+      "step": 5546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919085798772821e-06,
+      "loss": 0.9238,
+      "step": 5547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918656292944615e-06,
+      "loss": 0.9355,
+      "step": 5548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918226712146338e-06,
+      "loss": 0.8599,
+      "step": 5549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91779705638621e-06,
+      "loss": 0.8716,
+      "step": 5550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917367325672452e-06,
+      "loss": 0.9277,
+      "step": 5551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916937520013283e-06,
+      "loss": 0.8853,
+      "step": 5552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916507639416928e-06,
+      "loss": 0.8545,
+      "step": 5553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916077683891608e-06,
+      "loss": 0.9336,
+      "step": 5554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915647653445551e-06,
+      "loss": 0.8921,
+      "step": 5555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915217548086986e-06,
+      "loss": 0.9248,
+      "step": 5556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914787367824138e-06,
+      "loss": 1.0137,
+      "step": 5557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914357112665241e-06,
+      "loss": 0.9131,
+      "step": 5558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913926782618523e-06,
+      "loss": 0.9229,
+      "step": 5559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913496377692218e-06,
+      "loss": 0.8906,
+      "step": 5560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913065897894559e-06,
+      "loss": 0.8511,
+      "step": 5561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912635343233784e-06,
+      "loss": 0.8794,
+      "step": 5562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91220471371813e-06,
+      "loss": 0.9312,
+      "step": 5563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911774009355835e-06,
+      "loss": 0.8745,
+      "step": 5564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911343230155138e-06,
+      "loss": 0.3394,
+      "step": 5565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910912376124283e-06,
+      "loss": 0.9126,
+      "step": 5566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910481447271511e-06,
+      "loss": 0.938,
+      "step": 5567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910050443605066e-06,
+      "loss": 0.9673,
+      "step": 5568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909619365133195e-06,
+      "loss": 0.9634,
+      "step": 5569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909188211864143e-06,
+      "loss": 0.8877,
+      "step": 5570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.908756983806164e-06,
+      "loss": 0.8735,
+      "step": 5571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9083256809675e-06,
+      "loss": 0.3591,
+      "step": 5572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90789430335641e-06,
+      "loss": 0.8887,
+      "step": 5573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.907462850981144e-06,
+      "loss": 0.9121,
+      "step": 5574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.907031323849955e-06,
+      "loss": 0.9219,
+      "step": 5575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9065997219711e-06,
+      "loss": 0.9185,
+      "step": 5576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906168045352838e-06,
+      "loss": 0.9556,
+      "step": 5577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905736294003425e-06,
+      "loss": 0.9092,
+      "step": 5578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905304467931122e-06,
+      "loss": 0.9185,
+      "step": 5579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90487256714419e-06,
+      "loss": 0.8765,
+      "step": 5580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.904440591650894e-06,
+      "loss": 0.8691,
+      "step": 5581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.904008541459494e-06,
+      "loss": 0.874,
+      "step": 5582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90357641657826e-06,
+      "loss": 0.8955,
+      "step": 5583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90314421701546e-06,
+      "loss": 0.8823,
+      "step": 5584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902711942779358e-06,
+      "loss": 0.9302,
+      "step": 5585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902279593878229e-06,
+      "loss": 0.9053,
+      "step": 5586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90184717032034e-06,
+      "loss": 0.9883,
+      "step": 5587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901414672113966e-06,
+      "loss": 0.9106,
+      "step": 5588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900982099267383e-06,
+      "loss": 0.9209,
+      "step": 5589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900549451788864e-06,
+      "loss": 0.9053,
+      "step": 5590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900116729686687e-06,
+      "loss": 0.9531,
+      "step": 5591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899683932969133e-06,
+      "loss": 0.897,
+      "step": 5592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89925106164448e-06,
+      "loss": 0.8955,
+      "step": 5593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898818115721009e-06,
+      "loss": 0.9824,
+      "step": 5594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898385095207003e-06,
+      "loss": 0.9282,
+      "step": 5595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897952000110748e-06,
+      "loss": 0.9521,
+      "step": 5596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897518830440527e-06,
+      "loss": 0.9248,
+      "step": 5597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897085586204632e-06,
+      "loss": 0.9771,
+      "step": 5598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.896652267411347e-06,
+      "loss": 0.8755,
+      "step": 5599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.896218874068964e-06,
+      "loss": 0.9155,
+      "step": 5600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895785406185774e-06,
+      "loss": 0.9199,
+      "step": 5601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895351863770072e-06,
+      "loss": 0.9282,
+      "step": 5602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894918246830147e-06,
+      "loss": 0.8633,
+      "step": 5603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894484555374301e-06,
+      "loss": 0.9072,
+      "step": 5604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894050789410829e-06,
+      "loss": 0.917,
+      "step": 5605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893616948948027e-06,
+      "loss": 0.9746,
+      "step": 5606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893183033994199e-06,
+      "loss": 0.9365,
+      "step": 5607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892749044557642e-06,
+      "loss": 0.8774,
+      "step": 5608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892314980646663e-06,
+      "loss": 0.9082,
+      "step": 5609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891880842269567e-06,
+      "loss": 0.9189,
+      "step": 5610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891446629434654e-06,
+      "loss": 0.897,
+      "step": 5611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891012342150238e-06,
+      "loss": 0.9263,
+      "step": 5612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890577980424624e-06,
+      "loss": 0.8818,
+      "step": 5613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890143544266122e-06,
+      "loss": 0.9019,
+      "step": 5614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889709033683043e-06,
+      "loss": 0.9355,
+      "step": 5615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8892744486837e-06,
+      "loss": 0.9688,
+      "step": 5616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88883978927641e-06,
+      "loss": 0.9258,
+      "step": 5617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888405055469487e-06,
+      "loss": 0.9189,
+      "step": 5618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887970247271247e-06,
+      "loss": 0.9316,
+      "step": 5619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88753536469001e-06,
+      "loss": 0.8721,
+      "step": 5620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887100407734093e-06,
+      "loss": 0.8813,
+      "step": 5621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886665376411823e-06,
+      "loss": 0.9492,
+      "step": 5622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886230270731518e-06,
+      "loss": 0.9473,
+      "step": 5623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885795090701506e-06,
+      "loss": 0.8936,
+      "step": 5624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885359836330107e-06,
+      "loss": 0.8901,
+      "step": 5625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884924507625657e-06,
+      "loss": 0.9121,
+      "step": 5626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884489104596474e-06,
+      "loss": 0.8857,
+      "step": 5627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884053627250898e-06,
+      "loss": 0.8457,
+      "step": 5628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883618075597252e-06,
+      "loss": 0.9028,
+      "step": 5629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883182449643873e-06,
+      "loss": 0.9146,
+      "step": 5630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882746749399095e-06,
+      "loss": 0.8828,
+      "step": 5631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882310974871253e-06,
+      "loss": 0.9004,
+      "step": 5632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.881875126068686e-06,
+      "loss": 0.9165,
+      "step": 5633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88143920299973e-06,
+      "loss": 0.9487,
+      "step": 5634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.881003205672725e-06,
+      "loss": 0.9331,
+      "step": 5635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880567134096013e-06,
+      "loss": 0.9365,
+      "step": 5636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880130988277934e-06,
+      "loss": 0.9443,
+      "step": 5637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87969476822684e-06,
+      "loss": 0.8628,
+      "step": 5638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.879258473951068e-06,
+      "loss": 0.9492,
+      "step": 5639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87882210545897e-06,
+      "loss": 0.3625,
+      "step": 5640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878385662758894e-06,
+      "loss": 0.8726,
+      "step": 5641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877949145859186e-06,
+      "loss": 0.8965,
+      "step": 5642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8775125547682e-06,
+      "loss": 0.9072,
+      "step": 5643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87707588949429e-06,
+      "loss": 0.8823,
+      "step": 5644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.876639150045809e-06,
+      "loss": 0.9365,
+      "step": 5645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.876202336431112e-06,
+      "loss": 0.8574,
+      "step": 5646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875765448658556e-06,
+      "loss": 0.9023,
+      "step": 5647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875328486736498e-06,
+      "loss": 0.8848,
+      "step": 5648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.874891450673303e-06,
+      "loss": 0.9502,
+      "step": 5649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.874454340477325e-06,
+      "loss": 0.875,
+      "step": 5650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87401715615693e-06,
+      "loss": 0.9116,
+      "step": 5651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873579897720483e-06,
+      "loss": 0.8721,
+      "step": 5652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873142565176349e-06,
+      "loss": 0.9453,
+      "step": 5653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872705158532895e-06,
+      "loss": 0.939,
+      "step": 5654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872267677798488e-06,
+      "loss": 0.8994,
+      "step": 5655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871830122981499e-06,
+      "loss": 0.9248,
+      "step": 5656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871392494090296e-06,
+      "loss": 0.8955,
+      "step": 5657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870954791133256e-06,
+      "loss": 0.9604,
+      "step": 5658
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87051701411875e-06,
+      "loss": 0.8867,
+      "step": 5659
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870079163055156e-06,
+      "loss": 0.8604,
+      "step": 5660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86964123795085e-06,
+      "loss": 0.9458,
+      "step": 5661
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869203238814207e-06,
+      "loss": 0.9053,
+      "step": 5662
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86876516565361e-06,
+      "loss": 0.9082,
+      "step": 5663
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868327018477438e-06,
+      "loss": 1.0044,
+      "step": 5664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867888797294076e-06,
+      "loss": 0.9038,
+      "step": 5665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867450502111905e-06,
+      "loss": 0.9331,
+      "step": 5666
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867012132939312e-06,
+      "loss": 0.8818,
+      "step": 5667
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866573689784683e-06,
+      "loss": 0.8984,
+      "step": 5668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866135172656406e-06,
+      "loss": 0.9282,
+      "step": 5669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86569658156287e-06,
+      "loss": 0.8589,
+      "step": 5670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.865257916512468e-06,
+      "loss": 0.9468,
+      "step": 5671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864819177513593e-06,
+      "loss": 0.9561,
+      "step": 5672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864380364574634e-06,
+      "loss": 0.8857,
+      "step": 5673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86394147770399e-06,
+      "loss": 0.9917,
+      "step": 5674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863502516910058e-06,
+      "loss": 0.8984,
+      "step": 5675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863063482201232e-06,
+      "loss": 0.874,
+      "step": 5676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862624373585917e-06,
+      "loss": 0.9229,
+      "step": 5677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86218519107251e-06,
+      "loss": 0.9121,
+      "step": 5678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861745934669415e-06,
+      "loss": 0.9531,
+      "step": 5679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861306604385034e-06,
+      "loss": 0.4067,
+      "step": 5680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.860867200227772e-06,
+      "loss": 0.9033,
+      "step": 5681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.860427722206037e-06,
+      "loss": 0.9155,
+      "step": 5682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859988170328238e-06,
+      "loss": 0.9253,
+      "step": 5683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859548544602781e-06,
+      "loss": 0.8813,
+      "step": 5684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85910884503808e-06,
+      "loss": 0.917,
+      "step": 5685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858669071642545e-06,
+      "loss": 0.9424,
+      "step": 5686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85822922442459e-06,
+      "loss": 0.9751,
+      "step": 5687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85778930339263e-06,
+      "loss": 0.9346,
+      "step": 5688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857349308555081e-06,
+      "loss": 0.8726,
+      "step": 5689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856909239920362e-06,
+      "loss": 0.9458,
+      "step": 5690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856469097496891e-06,
+      "loss": 0.8384,
+      "step": 5691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85602888129309e-06,
+      "loss": 0.9443,
+      "step": 5692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85558859131738e-06,
+      "loss": 0.897,
+      "step": 5693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855148227578184e-06,
+      "loss": 0.8394,
+      "step": 5694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854707790083928e-06,
+      "loss": 0.8984,
+      "step": 5695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854267278843037e-06,
+      "loss": 0.8936,
+      "step": 5696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.853826693863938e-06,
+      "loss": 0.9697,
+      "step": 5697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.853386035155063e-06,
+      "loss": 0.915,
+      "step": 5698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852945302724843e-06,
+      "loss": 0.9346,
+      "step": 5699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852504496581705e-06,
+      "loss": 0.9248,
+      "step": 5700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852063616734086e-06,
+      "loss": 0.9009,
+      "step": 5701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85162266319042e-06,
+      "loss": 0.856,
+      "step": 5702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.851181635959144e-06,
+      "loss": 0.9634,
+      "step": 5703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850740535048693e-06,
+      "loss": 0.9258,
+      "step": 5704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85029936046751e-06,
+      "loss": 0.8994,
+      "step": 5705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84985811222403e-06,
+      "loss": 0.9727,
+      "step": 5706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8494167903267e-06,
+      "loss": 0.8574,
+      "step": 5707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848975394783958e-06,
+      "loss": 0.9014,
+      "step": 5708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848533925604254e-06,
+      "loss": 0.939,
+      "step": 5709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848092382796031e-06,
+      "loss": 0.9761,
+      "step": 5710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.847650766367736e-06,
+      "loss": 0.9878,
+      "step": 5711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84720907632782e-06,
+      "loss": 0.9819,
+      "step": 5712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846767312684731e-06,
+      "loss": 0.8896,
+      "step": 5713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846325475446922e-06,
+      "loss": 0.9097,
+      "step": 5714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845883564622847e-06,
+      "loss": 0.9629,
+      "step": 5715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845441580220954e-06,
+      "loss": 0.8574,
+      "step": 5716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844999522249708e-06,
+      "loss": 0.9194,
+      "step": 5717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844557390717563e-06,
+      "loss": 0.9141,
+      "step": 5718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844115185632974e-06,
+      "loss": 0.9077,
+      "step": 5719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843672907004406e-06,
+      "loss": 0.9043,
+      "step": 5720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843230554840318e-06,
+      "loss": 0.9302,
+      "step": 5721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842788129149172e-06,
+      "loss": 0.9302,
+      "step": 5722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842345629939434e-06,
+      "loss": 0.9375,
+      "step": 5723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841903057219572e-06,
+      "loss": 0.9346,
+      "step": 5724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841460410998047e-06,
+      "loss": 0.8833,
+      "step": 5725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841017691283333e-06,
+      "loss": 0.3955,
+      "step": 5726
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840574898083896e-06,
+      "loss": 0.8589,
+      "step": 5727
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84013203140821e-06,
+      "loss": 0.875,
+      "step": 5728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83968909126475e-06,
+      "loss": 0.9888,
+      "step": 5729
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839246077661982e-06,
+      "loss": 0.9351,
+      "step": 5730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83880299060839e-06,
+      "loss": 0.9268,
+      "step": 5731
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838359830112445e-06,
+      "loss": 0.9517,
+      "step": 5732
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837916596182631e-06,
+      "loss": 0.939,
+      "step": 5733
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837473288827424e-06,
+      "loss": 0.3669,
+      "step": 5734
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837029908055306e-06,
+      "loss": 0.9375,
+      "step": 5735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.836586453874757e-06,
+      "loss": 0.917,
+      "step": 5736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.836142926294267e-06,
+      "loss": 0.9038,
+      "step": 5737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835699325322315e-06,
+      "loss": 0.917,
+      "step": 5738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835255650967392e-06,
+      "loss": 0.9614,
+      "step": 5739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834811903237983e-06,
+      "loss": 0.8911,
+      "step": 5740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83436808214258e-06,
+      "loss": 0.8862,
+      "step": 5741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833924187689675e-06,
+      "loss": 0.9097,
+      "step": 5742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833480219887757e-06,
+      "loss": 0.9722,
+      "step": 5743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833036178745323e-06,
+      "loss": 0.9097,
+      "step": 5744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832592064270867e-06,
+      "loss": 0.8892,
+      "step": 5745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832147876472883e-06,
+      "loss": 0.9067,
+      "step": 5746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831703615359875e-06,
+      "loss": 0.9448,
+      "step": 5747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831259280940336e-06,
+      "loss": 0.9009,
+      "step": 5748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83081487322277e-06,
+      "loss": 0.8989,
+      "step": 5749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83037039221568e-06,
+      "loss": 0.8838,
+      "step": 5750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.829925837927567e-06,
+      "loss": 0.9111,
+      "step": 5751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82948121036694e-06,
+      "loss": 0.8735,
+      "step": 5752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.8290365095423e-06,
+      "loss": 0.35,
+      "step": 5753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828591735462158e-06,
+      "loss": 0.8809,
+      "step": 5754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828146888135025e-06,
+      "loss": 0.8794,
+      "step": 5755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827701967569408e-06,
+      "loss": 0.9448,
+      "step": 5756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82725697377382e-06,
+      "loss": 0.9121,
+      "step": 5757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826811906756778e-06,
+      "loss": 0.8979,
+      "step": 5758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826366766526792e-06,
+      "loss": 0.9321,
+      "step": 5759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825921553092381e-06,
+      "loss": 0.8569,
+      "step": 5760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82547626646206e-06,
+      "loss": 0.8965,
+      "step": 5761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825030906644351e-06,
+      "loss": 0.9395,
+      "step": 5762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824585473647774e-06,
+      "loss": 0.877,
+      "step": 5763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824139967480848e-06,
+      "loss": 0.9019,
+      "step": 5764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823694388152101e-06,
+      "loss": 0.9067,
+      "step": 5765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823248735670053e-06,
+      "loss": 0.8984,
+      "step": 5766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822803010043232e-06,
+      "loss": 0.958,
+      "step": 5767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822357211280167e-06,
+      "loss": 0.8828,
+      "step": 5768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821911339389384e-06,
+      "loss": 0.8296,
+      "step": 5769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821465394379415e-06,
+      "loss": 0.9053,
+      "step": 5770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821019376258792e-06,
+      "loss": 0.8779,
+      "step": 5771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820573285036044e-06,
+      "loss": 0.9209,
+      "step": 5772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820127120719713e-06,
+      "loss": 0.9478,
+      "step": 5773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819680883318328e-06,
+      "loss": 0.9048,
+      "step": 5774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819234572840427e-06,
+      "loss": 0.917,
+      "step": 5775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818788189294552e-06,
+      "loss": 0.8945,
+      "step": 5776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818341732689241e-06,
+      "loss": 0.9565,
+      "step": 5777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817895203033037e-06,
+      "loss": 0.8564,
+      "step": 5778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81744860033448e-06,
+      "loss": 0.8877,
+      "step": 5779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817001924602117e-06,
+      "loss": 0.9058,
+      "step": 5780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.816555175844492e-06,
+      "loss": 0.9106,
+      "step": 5781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81610835407015e-06,
+      "loss": 0.3442,
+      "step": 5782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.815661459287645e-06,
+      "loss": 0.8633,
+      "step": 5783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.815214491505521e-06,
+      "loss": 0.9175,
+      "step": 5784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.814767450732334e-06,
+      "loss": 0.8696,
+      "step": 5785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.814320336976633e-06,
+      "loss": 0.9727,
+      "step": 5786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.813873150246973e-06,
+      "loss": 0.9375,
+      "step": 5787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81342589055191e-06,
+      "loss": 0.9209,
+      "step": 5788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812978557900002e-06,
+      "loss": 0.937,
+      "step": 5789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812531152299804e-06,
+      "loss": 0.9199,
+      "step": 5790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812083673759878e-06,
+      "loss": 0.811,
+      "step": 5791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811636122288783e-06,
+      "loss": 0.9004,
+      "step": 5792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811188497895084e-06,
+      "loss": 0.9033,
+      "step": 5793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810740800587344e-06,
+      "loss": 0.3004,
+      "step": 5794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810293030374126e-06,
+      "loss": 0.8867,
+      "step": 5795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809845187263999e-06,
+      "loss": 0.9453,
+      "step": 5796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809397271265528e-06,
+      "loss": 0.835,
+      "step": 5797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808949282387287e-06,
+      "loss": 0.9302,
+      "step": 5798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808501220637842e-06,
+      "loss": 0.9492,
+      "step": 5799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808053086025767e-06,
+      "loss": 0.8315,
+      "step": 5800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807604878559638e-06,
+      "loss": 0.8677,
+      "step": 5801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807156598248026e-06,
+      "loss": 0.938,
+      "step": 5802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806708245099509e-06,
+      "loss": 0.8794,
+      "step": 5803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806259819122666e-06,
+      "loss": 0.9375,
+      "step": 5804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805811320326073e-06,
+      "loss": 0.937,
+      "step": 5805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805362748718314e-06,
+      "loss": 0.9482,
+      "step": 5806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.804914104307966e-06,
+      "loss": 0.9199,
+      "step": 5807
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80446538710362e-06,
+      "loss": 0.9253,
+      "step": 5808
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.804016597113853e-06,
+      "loss": 0.9473,
+      "step": 5809
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803567734347253e-06,
+      "loss": 0.8545,
+      "step": 5810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803118798812412e-06,
+      "loss": 0.9287,
+      "step": 5811
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802669790517913e-06,
+      "loss": 0.8726,
+      "step": 5812
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802220709472349e-06,
+      "loss": 0.9561,
+      "step": 5813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801771555684312e-06,
+      "loss": 0.3149,
+      "step": 5814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801322329162394e-06,
+      "loss": 0.9517,
+      "step": 5815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80087302991519e-06,
+      "loss": 0.3159,
+      "step": 5816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800423657951294e-06,
+      "loss": 0.877,
+      "step": 5817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799974213279306e-06,
+      "loss": 0.9639,
+      "step": 5818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799524695907822e-06,
+      "loss": 0.8789,
+      "step": 5819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799075105845446e-06,
+      "loss": 0.9585,
+      "step": 5820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798625443100774e-06,
+      "loss": 0.8286,
+      "step": 5821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798175707682412e-06,
+      "loss": 0.9136,
+      "step": 5822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.797725899598964e-06,
+      "loss": 0.8989,
+      "step": 5823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.797276018859035e-06,
+      "loss": 0.8667,
+      "step": 5824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796826065471232e-06,
+      "loss": 0.9448,
+      "step": 5825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796376039444163e-06,
+      "loss": 0.9253,
+      "step": 5826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795925940786438e-06,
+      "loss": 0.9033,
+      "step": 5827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795475769506668e-06,
+      "loss": 0.9146,
+      "step": 5828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795025525613466e-06,
+      "loss": 0.9492,
+      "step": 5829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.794575209115445e-06,
+      "loss": 0.9141,
+      "step": 5830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79412482002122e-06,
+      "loss": 0.8633,
+      "step": 5831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793674358339412e-06,
+      "loss": 0.8911,
+      "step": 5832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793223824078631e-06,
+      "loss": 0.9458,
+      "step": 5833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792773217247504e-06,
+      "loss": 0.8687,
+      "step": 5834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792322537854646e-06,
+      "loss": 0.8818,
+      "step": 5835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.791871785908685e-06,
+      "loss": 0.8984,
+      "step": 5836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79142096141824e-06,
+      "loss": 0.957,
+      "step": 5837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790970064391937e-06,
+      "loss": 0.8921,
+      "step": 5838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790519094838403e-06,
+      "loss": 0.8809,
+      "step": 5839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790068052766268e-06,
+      "loss": 0.9365,
+      "step": 5840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789616938184155e-06,
+      "loss": 0.8657,
+      "step": 5841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7891657511007e-06,
+      "loss": 0.8794,
+      "step": 5842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788714491524533e-06,
+      "loss": 0.9321,
+      "step": 5843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788263159464287e-06,
+      "loss": 0.8535,
+      "step": 5844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7878117549286e-06,
+      "loss": 0.8394,
+      "step": 5845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787360277926101e-06,
+      "loss": 0.8828,
+      "step": 5846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786908728465435e-06,
+      "loss": 0.9233,
+      "step": 5847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786457106555236e-06,
+      "loss": 0.9336,
+      "step": 5848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786005412204146e-06,
+      "loss": 0.9165,
+      "step": 5849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785553645420807e-06,
+      "loss": 0.8462,
+      "step": 5850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78510180621386e-06,
+      "loss": 0.3894,
+      "step": 5851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784649894591953e-06,
+      "loss": 0.8389,
+      "step": 5852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784197910563728e-06,
+      "loss": 0.9355,
+      "step": 5853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783745854137834e-06,
+      "loss": 0.8887,
+      "step": 5854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78329372532292e-06,
+      "loss": 0.915,
+      "step": 5855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782841524127637e-06,
+      "loss": 0.8535,
+      "step": 5856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782389250560632e-06,
+      "loss": 0.8936,
+      "step": 5857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78193690463056e-06,
+      "loss": 0.9316,
+      "step": 5858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781484486346076e-06,
+      "loss": 0.9731,
+      "step": 5859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781031995715836e-06,
+      "loss": 0.9326,
+      "step": 5860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780579432748497e-06,
+      "loss": 0.9507,
+      "step": 5861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780126797452713e-06,
+      "loss": 0.9131,
+      "step": 5862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779674089837148e-06,
+      "loss": 0.8916,
+      "step": 5863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779221309910461e-06,
+      "loss": 0.8584,
+      "step": 5864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778768457681316e-06,
+      "loss": 0.3285,
+      "step": 5865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778315533158375e-06,
+      "loss": 0.415,
+      "step": 5866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777862536350304e-06,
+      "loss": 0.863,
+      "step": 5867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77740946726577e-06,
+      "loss": 0.3933,
+      "step": 5868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77695632591344e-06,
+      "loss": 0.8062,
+      "step": 5869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776503112301984e-06,
+      "loss": 0.8921,
+      "step": 5870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776049826440073e-06,
+      "loss": 0.9727,
+      "step": 5871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775596468336379e-06,
+      "loss": 0.9346,
+      "step": 5872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775143037999573e-06,
+      "loss": 0.8691,
+      "step": 5873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774689535438333e-06,
+      "loss": 0.917,
+      "step": 5874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774235960661332e-06,
+      "loss": 0.8906,
+      "step": 5875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773782313677252e-06,
+      "loss": 0.8657,
+      "step": 5876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77332859449477e-06,
+      "loss": 0.9521,
+      "step": 5877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772874803122564e-06,
+      "loss": 0.9111,
+      "step": 5878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772420939569319e-06,
+      "loss": 0.9126,
+      "step": 5879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771967003843715e-06,
+      "loss": 0.9385,
+      "step": 5880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771512995954441e-06,
+      "loss": 0.9087,
+      "step": 5881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771058915910178e-06,
+      "loss": 0.9453,
+      "step": 5882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770604763719616e-06,
+      "loss": 0.9771,
+      "step": 5883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770150539391444e-06,
+      "loss": 0.7649,
+      "step": 5884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76969624293435e-06,
+      "loss": 0.8892,
+      "step": 5885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.769241874357029e-06,
+      "loss": 0.365,
+      "step": 5886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76878743366817e-06,
+      "loss": 1.0024,
+      "step": 5887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768332920876468e-06,
+      "loss": 0.918,
+      "step": 5888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767878335990619e-06,
+      "loss": 0.9102,
+      "step": 5889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76742367901932e-06,
+      "loss": 0.9453,
+      "step": 5890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766968949971269e-06,
+      "loss": 0.8945,
+      "step": 5891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766514148855166e-06,
+      "loss": 0.9473,
+      "step": 5892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766059275679712e-06,
+      "loss": 0.9517,
+      "step": 5893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.765604330453611e-06,
+      "loss": 0.8955,
+      "step": 5894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.765149313185564e-06,
+      "loss": 0.9966,
+      "step": 5895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764694223884279e-06,
+      "loss": 0.9131,
+      "step": 5896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76423906255846e-06,
+      "loss": 0.8413,
+      "step": 5897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763783829216815e-06,
+      "loss": 0.9263,
+      "step": 5898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763328523868056e-06,
+      "loss": 0.8984,
+      "step": 5899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762873146520892e-06,
+      "loss": 0.9292,
+      "step": 5900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762417697184034e-06,
+      "loss": 0.9424,
+      "step": 5901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761962175866196e-06,
+      "loss": 0.9268,
+      "step": 5902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761506582576095e-06,
+      "loss": 0.9355,
+      "step": 5903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761050917322446e-06,
+      "loss": 0.8799,
+      "step": 5904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760595180113965e-06,
+      "loss": 0.876,
+      "step": 5905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760139370959373e-06,
+      "loss": 0.8892,
+      "step": 5906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75968348986739e-06,
+      "loss": 0.896,
+      "step": 5907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759227536846736e-06,
+      "loss": 0.8481,
+      "step": 5908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758771511906136e-06,
+      "loss": 0.9048,
+      "step": 5909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758315415054314e-06,
+      "loss": 0.9546,
+      "step": 5910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757859246299996e-06,
+      "loss": 0.8032,
+      "step": 5911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757403005651909e-06,
+      "loss": 0.8901,
+      "step": 5912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756946693118782e-06,
+      "loss": 0.9121,
+      "step": 5913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756490308709343e-06,
+      "loss": 0.9663,
+      "step": 5914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756033852432326e-06,
+      "loss": 0.9263,
+      "step": 5915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755577324296462e-06,
+      "loss": 0.9258,
+      "step": 5916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755120724310485e-06,
+      "loss": 0.8906,
+      "step": 5917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754664052483133e-06,
+      "loss": 0.9341,
+      "step": 5918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754207308823141e-06,
+      "loss": 0.8989,
+      "step": 5919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753750493339247e-06,
+      "loss": 0.9336,
+      "step": 5920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75329360604019e-06,
+      "loss": 0.9087,
+      "step": 5921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752836646934712e-06,
+      "loss": 0.9546,
+      "step": 5922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752379616031557e-06,
+      "loss": 0.8745,
+      "step": 5923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751922513339464e-06,
+      "loss": 0.834,
+      "step": 5924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751465338867184e-06,
+      "loss": 0.9238,
+      "step": 5925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75100809262346e-06,
+      "loss": 0.9316,
+      "step": 5926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.750550774617036e-06,
+      "loss": 0.9346,
+      "step": 5927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75009338485667e-06,
+      "loss": 0.918,
+      "step": 5928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749635923351108e-06,
+      "loss": 0.9131,
+      "step": 5929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7491783901091e-06,
+      "loss": 0.9785,
+      "step": 5930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.748720785139404e-06,
+      "loss": 0.8545,
+      "step": 5931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.748263108450769e-06,
+      "loss": 0.9019,
+      "step": 5932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747805360051957e-06,
+      "loss": 0.9512,
+      "step": 5933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747347539951722e-06,
+      "loss": 0.8989,
+      "step": 5934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746889648158823e-06,
+      "loss": 0.8545,
+      "step": 5935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74643168468202e-06,
+      "loss": 0.8979,
+      "step": 5936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745973649530077e-06,
+      "loss": 0.8979,
+      "step": 5937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745515542711754e-06,
+      "loss": 0.9424,
+      "step": 5938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745057364235817e-06,
+      "loss": 0.9492,
+      "step": 5939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74459911411103e-06,
+      "loss": 0.9355,
+      "step": 5940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744140792346162e-06,
+      "loss": 0.9521,
+      "step": 5941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74368239894998e-06,
+      "loss": 0.9414,
+      "step": 5942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743223933931256e-06,
+      "loss": 0.896,
+      "step": 5943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742765397298759e-06,
+      "loss": 0.8638,
+      "step": 5944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742306789061261e-06,
+      "loss": 0.8945,
+      "step": 5945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741848109227537e-06,
+      "loss": 0.8828,
+      "step": 5946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741389357806362e-06,
+      "loss": 0.875,
+      "step": 5947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740930534806515e-06,
+      "loss": 0.8853,
+      "step": 5948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74047164023677e-06,
+      "loss": 0.376,
+      "step": 5949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740012674105908e-06,
+      "loss": 0.8516,
+      "step": 5950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739553636422713e-06,
+      "loss": 0.9131,
+      "step": 5951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739094527195962e-06,
+      "loss": 0.9199,
+      "step": 5952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73863534643444e-06,
+      "loss": 0.9155,
+      "step": 5953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738176094146933e-06,
+      "loss": 0.9297,
+      "step": 5954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737716770342227e-06,
+      "loss": 0.9282,
+      "step": 5955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737257375029109e-06,
+      "loss": 0.9263,
+      "step": 5956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.736797908216367e-06,
+      "loss": 0.9077,
+      "step": 5957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.736338369912795e-06,
+      "loss": 0.9106,
+      "step": 5958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73587876012718e-06,
+      "loss": 0.8862,
+      "step": 5959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735419078868318e-06,
+      "loss": 0.9521,
+      "step": 5960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734959326145005e-06,
+      "loss": 0.894,
+      "step": 5961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734499501966032e-06,
+      "loss": 0.8687,
+      "step": 5962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7340396063402e-06,
+      "loss": 0.9595,
+      "step": 5963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733579639276306e-06,
+      "loss": 0.9478,
+      "step": 5964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733119600783149e-06,
+      "loss": 0.8608,
+      "step": 5965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732659490869533e-06,
+      "loss": 0.8857,
+      "step": 5966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732199309544257e-06,
+      "loss": 0.8862,
+      "step": 5967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731739056816129e-06,
+      "loss": 0.9395,
+      "step": 5968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73127873269395e-06,
+      "loss": 0.8975,
+      "step": 5969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73081833718653e-06,
+      "loss": 0.8525,
+      "step": 5970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.730357870302678e-06,
+      "loss": 0.8789,
+      "step": 5971
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.7298973320512e-06,
+      "loss": 0.958,
+      "step": 5972
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72943672244091e-06,
+      "loss": 0.9316,
+      "step": 5973
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728976041480616e-06,
+      "loss": 0.9033,
+      "step": 5974
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728515289179136e-06,
+      "loss": 0.8882,
+      "step": 5975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728054465545282e-06,
+      "loss": 0.9043,
+      "step": 5976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727593570587872e-06,
+      "loss": 0.9453,
+      "step": 5977
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727132604315723e-06,
+      "loss": 0.8638,
+      "step": 5978
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726671566737654e-06,
+      "loss": 0.9678,
+      "step": 5979
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726210457862486e-06,
+      "loss": 0.8223,
+      "step": 5980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725749277699041e-06,
+      "loss": 0.8535,
+      "step": 5981
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725288026256139e-06,
+      "loss": 0.9004,
+      "step": 5982
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724826703542607e-06,
+      "loss": 0.9297,
+      "step": 5983
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724365309567272e-06,
+      "loss": 0.9165,
+      "step": 5984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723903844338958e-06,
+      "loss": 0.9131,
+      "step": 5985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723442307866498e-06,
+      "loss": 0.8125,
+      "step": 5986
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722980700158718e-06,
+      "loss": 0.8774,
+      "step": 5987
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722519021224449e-06,
+      "loss": 0.9126,
+      "step": 5988
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722057271072527e-06,
+      "loss": 0.958,
+      "step": 5989
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721595449711783e-06,
+      "loss": 0.9048,
+      "step": 5990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721133557151053e-06,
+      "loss": 0.9219,
+      "step": 5991
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.720671593399177e-06,
+      "loss": 0.9307,
+      "step": 5992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.720209558464986e-06,
+      "loss": 0.9219,
+      "step": 5993
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719747452357328e-06,
+      "loss": 0.9136,
+      "step": 5994
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71928527508504e-06,
+      "loss": 0.8892,
+      "step": 5995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71882302665696e-06,
+      "loss": 0.9365,
+      "step": 5996
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718360707081939e-06,
+      "loss": 0.915,
+      "step": 5997
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717898316368816e-06,
+      "loss": 0.9355,
+      "step": 5998
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717435854526439e-06,
+      "loss": 0.8447,
+      "step": 5999
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716973321563657e-06,
+      "loss": 0.9053,
+      "step": 6000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71651071748932e-06,
+      "loss": 0.835,
+      "step": 6001
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716048042312274e-06,
+      "loss": 0.9048,
+      "step": 6002
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715585296041373e-06,
+      "loss": 0.8584,
+      "step": 6003
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715122478685473e-06,
+      "loss": 0.8701,
+      "step": 6004
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714659590253423e-06,
+      "loss": 0.875,
+      "step": 6005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714196630754083e-06,
+      "loss": 0.9033,
+      "step": 6006
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713733600196308e-06,
+      "loss": 0.4058,
+      "step": 6007
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713270498588958e-06,
+      "loss": 0.936,
+      "step": 6008
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71280732594089e-06,
+      "loss": 0.9092,
+      "step": 6009
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.712344082260968e-06,
+      "loss": 0.8862,
+      "step": 6010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711880767558054e-06,
+      "loss": 0.9053,
+      "step": 6011
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711417381841012e-06,
+      "loss": 0.9209,
+      "step": 6012
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710953925118709e-06,
+      "loss": 0.8569,
+      "step": 6013
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710490397400007e-06,
+      "loss": 0.3577,
+      "step": 6014
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710026798693776e-06,
+      "loss": 0.9487,
+      "step": 6015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709563129008887e-06,
+      "loss": 0.9727,
+      "step": 6016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70909938835421e-06,
+      "loss": 0.9175,
+      "step": 6017
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708635576738617e-06,
+      "loss": 0.9224,
+      "step": 6018
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708171694170981e-06,
+      "loss": 0.9268,
+      "step": 6019
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.707707740660179e-06,
+      "loss": 0.8838,
+      "step": 6020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.707243716215083e-06,
+      "loss": 0.9326,
+      "step": 6021
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706779620844574e-06,
+      "loss": 0.834,
+      "step": 6022
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706315454557528e-06,
+      "loss": 0.8853,
+      "step": 6023
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705851217362829e-06,
+      "loss": 0.9355,
+      "step": 6024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705386909269355e-06,
+      "loss": 0.3303,
+      "step": 6025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704922530285992e-06,
+      "loss": 1.0015,
+      "step": 6026
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704458080421622e-06,
+      "loss": 0.9224,
+      "step": 6027
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70399355968513e-06,
+      "loss": 0.8975,
+      "step": 6028
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703528968085406e-06,
+      "loss": 0.9395,
+      "step": 6029
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703064305631337e-06,
+      "loss": 0.8589,
+      "step": 6030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702599572331812e-06,
+      "loss": 0.9185,
+      "step": 6031
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702134768195722e-06,
+      "loss": 0.8745,
+      "step": 6032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701669893231963e-06,
+      "loss": 0.9082,
+      "step": 6033
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701204947449421e-06,
+      "loss": 0.8867,
+      "step": 6034
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700739930857001e-06,
+      "loss": 0.938,
+      "step": 6035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700274843463592e-06,
+      "loss": 0.9385,
+      "step": 6036
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699809685278093e-06,
+      "loss": 0.9233,
+      "step": 6037
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699344456309407e-06,
+      "loss": 0.9185,
+      "step": 6038
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69887915656643e-06,
+      "loss": 0.9302,
+      "step": 6039
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698413786058068e-06,
+      "loss": 0.9004,
+      "step": 6040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697948344793221e-06,
+      "loss": 0.8901,
+      "step": 6041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697482832780795e-06,
+      "loss": 0.9277,
+      "step": 6042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697017250029697e-06,
+      "loss": 0.9541,
+      "step": 6043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696551596548832e-06,
+      "loss": 0.8828,
+      "step": 6044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696085872347112e-06,
+      "loss": 0.9644,
+      "step": 6045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.695620077433441e-06,
+      "loss": 0.9453,
+      "step": 6046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.695154211816738e-06,
+      "loss": 0.8359,
+      "step": 6047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69468827550591e-06,
+      "loss": 0.9209,
+      "step": 6048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694222268509874e-06,
+      "loss": 0.9053,
+      "step": 6049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693756190837544e-06,
+      "loss": 0.9282,
+      "step": 6050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69329004249784e-06,
+      "loss": 0.9248,
+      "step": 6051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692823823499674e-06,
+      "loss": 0.9058,
+      "step": 6052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692357533851972e-06,
+      "loss": 0.4821,
+      "step": 6053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691891173563649e-06,
+      "loss": 0.9033,
+      "step": 6054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691424742643633e-06,
+      "loss": 0.9551,
+      "step": 6055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690958241100843e-06,
+      "loss": 0.9141,
+      "step": 6056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690491668944205e-06,
+      "loss": 0.8843,
+      "step": 6057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690025026182647e-06,
+      "loss": 0.8965,
+      "step": 6058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689558312825098e-06,
+      "loss": 0.9443,
+      "step": 6059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68909152888048e-06,
+      "loss": 0.9595,
+      "step": 6060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.688624674357731e-06,
+      "loss": 0.8647,
+      "step": 6061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.688157749265777e-06,
+      "loss": 0.9453,
+      "step": 6062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687690753613554e-06,
+      "loss": 0.8911,
+      "step": 6063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687223687409997e-06,
+      "loss": 0.9541,
+      "step": 6064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68675655066404e-06,
+      "loss": 0.939,
+      "step": 6065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68628934338462e-06,
+      "loss": 0.9316,
+      "step": 6066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.685822065580678e-06,
+      "loss": 0.9619,
+      "step": 6067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68535471726115e-06,
+      "loss": 0.8862,
+      "step": 6068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68488729843498e-06,
+      "loss": 0.9019,
+      "step": 6069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684419809111106e-06,
+      "loss": 0.9131,
+      "step": 6070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683952249298478e-06,
+      "loss": 0.9819,
+      "step": 6071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683484619006039e-06,
+      "loss": 0.8945,
+      "step": 6072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683016918242733e-06,
+      "loss": 0.9487,
+      "step": 6073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682549147017509e-06,
+      "loss": 0.9351,
+      "step": 6074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682081305339319e-06,
+      "loss": 0.9219,
+      "step": 6075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68161339321711e-06,
+      "loss": 0.9102,
+      "step": 6076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.681145410659837e-06,
+      "loss": 0.9497,
+      "step": 6077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68067735767645e-06,
+      "loss": 0.9331,
+      "step": 6078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680209234275905e-06,
+      "loss": 0.9038,
+      "step": 6079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679741040467158e-06,
+      "loss": 0.9419,
+      "step": 6080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679272776259167e-06,
+      "loss": 0.9038,
+      "step": 6081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67880444166089e-06,
+      "loss": 0.8379,
+      "step": 6082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678336036681287e-06,
+      "loss": 0.3843,
+      "step": 6083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677867561329319e-06,
+      "loss": 0.8682,
+      "step": 6084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67739901561395e-06,
+      "loss": 0.8984,
+      "step": 6085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67693039954414e-06,
+      "loss": 0.9717,
+      "step": 6086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67646171312886e-06,
+      "loss": 0.8423,
+      "step": 6087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675992956377074e-06,
+      "loss": 0.9619,
+      "step": 6088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675524129297749e-06,
+      "loss": 0.897,
+      "step": 6089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675055231899857e-06,
+      "loss": 0.9316,
+      "step": 6090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674586264192365e-06,
+      "loss": 0.8823,
+      "step": 6091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674117226184248e-06,
+      "loss": 0.9829,
+      "step": 6092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67364811788448e-06,
+      "loss": 0.9009,
+      "step": 6093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.673178939302033e-06,
+      "loss": 0.9424,
+      "step": 6094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672709690445887e-06,
+      "loss": 0.8574,
+      "step": 6095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672240371325015e-06,
+      "loss": 0.9175,
+      "step": 6096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6717709819484e-06,
+      "loss": 0.9053,
+      "step": 6097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.671301522325017e-06,
+      "loss": 0.9058,
+      "step": 6098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670831992463852e-06,
+      "loss": 0.8999,
+      "step": 6099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670362392373888e-06,
+      "loss": 0.9253,
+      "step": 6100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669892722064106e-06,
+      "loss": 0.9297,
+      "step": 6101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669422981543494e-06,
+      "loss": 0.9146,
+      "step": 6102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668953170821037e-06,
+      "loss": 0.8999,
+      "step": 6103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668483289905725e-06,
+      "loss": 0.9219,
+      "step": 6104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668013338806548e-06,
+      "loss": 0.9673,
+      "step": 6105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667543317532495e-06,
+      "loss": 0.9463,
+      "step": 6106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667073226092558e-06,
+      "loss": 0.875,
+      "step": 6107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666603064495732e-06,
+      "loss": 0.8901,
+      "step": 6108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666132832751014e-06,
+      "loss": 0.936,
+      "step": 6109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665662530867393e-06,
+      "loss": 0.9258,
+      "step": 6110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665192158853875e-06,
+      "loss": 0.9375,
+      "step": 6111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664721716719454e-06,
+      "loss": 0.958,
+      "step": 6112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664251204473134e-06,
+      "loss": 0.9399,
+      "step": 6113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663780622123911e-06,
+      "loss": 0.9634,
+      "step": 6114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663309969680794e-06,
+      "loss": 0.9482,
+      "step": 6115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662839247152782e-06,
+      "loss": 0.8887,
+      "step": 6116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662368454548884e-06,
+      "loss": 0.9604,
+      "step": 6117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661897591878108e-06,
+      "loss": 0.9341,
+      "step": 6118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661426659149458e-06,
+      "loss": 0.9414,
+      "step": 6119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660955656371949e-06,
+      "loss": 0.8965,
+      "step": 6120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660484583554588e-06,
+      "loss": 0.9165,
+      "step": 6121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660013440706388e-06,
+      "loss": 0.9326,
+      "step": 6122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659542227836366e-06,
+      "loss": 0.9351,
+      "step": 6123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659070944953532e-06,
+      "loss": 0.896,
+      "step": 6124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658599592066905e-06,
+      "loss": 0.8354,
+      "step": 6125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658128169185505e-06,
+      "loss": 0.9136,
+      "step": 6126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657656676318346e-06,
+      "loss": 0.3195,
+      "step": 6127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65718511347445e-06,
+      "loss": 0.8784,
+      "step": 6128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656713480662844e-06,
+      "loss": 0.8657,
+      "step": 6129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656241777892544e-06,
+      "loss": 0.9316,
+      "step": 6130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.655770005172578e-06,
+      "loss": 0.8638,
+      "step": 6131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.655298162511971e-06,
+      "loss": 0.9238,
+      "step": 6132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65482624991975e-06,
+      "loss": 0.9302,
+      "step": 6133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654354267404943e-06,
+      "loss": 0.8687,
+      "step": 6134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653882214976582e-06,
+      "loss": 0.9824,
+      "step": 6135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653410092643693e-06,
+      "loss": 0.8984,
+      "step": 6136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652937900415314e-06,
+      "loss": 0.917,
+      "step": 6137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652465638300478e-06,
+      "loss": 0.9321,
+      "step": 6138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651993306308217e-06,
+      "loss": 0.8408,
+      "step": 6139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65152090444757e-06,
+      "loss": 0.9653,
+      "step": 6140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651048432727573e-06,
+      "loss": 0.9023,
+      "step": 6141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650575891157267e-06,
+      "loss": 0.9453,
+      "step": 6142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650103279745692e-06,
+      "loss": 0.9268,
+      "step": 6143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64963059850189e-06,
+      "loss": 0.9258,
+      "step": 6144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6491578474349e-06,
+      "loss": 0.8994,
+      "step": 6145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.648685026553774e-06,
+      "loss": 0.9297,
+      "step": 6146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.648212135867552e-06,
+      "loss": 0.4628,
+      "step": 6147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647739175385284e-06,
+      "loss": 0.8701,
+      "step": 6148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647266145116015e-06,
+      "loss": 0.8809,
+      "step": 6149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6467930450688e-06,
+      "loss": 0.8696,
+      "step": 6150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.646319875252686e-06,
+      "loss": 0.8501,
+      "step": 6151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645846635676728e-06,
+      "loss": 0.9141,
+      "step": 6152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645373326349977e-06,
+      "loss": 0.9355,
+      "step": 6153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64489994728149e-06,
+      "loss": 0.9194,
+      "step": 6154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644426498480325e-06,
+      "loss": 0.9321,
+      "step": 6155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643952979955538e-06,
+      "loss": 0.8833,
+      "step": 6156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643479391716187e-06,
+      "loss": 0.9316,
+      "step": 6157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643005733771334e-06,
+      "loss": 0.9443,
+      "step": 6158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642532006130042e-06,
+      "loss": 0.3322,
+      "step": 6159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642058208801373e-06,
+      "loss": 0.9673,
+      "step": 6160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64158434179439e-06,
+      "loss": 0.999,
+      "step": 6161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64111040511816e-06,
+      "loss": 0.8911,
+      "step": 6162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640636398781751e-06,
+      "loss": 0.2838,
+      "step": 6163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64016232279423e-06,
+      "loss": 0.897,
+      "step": 6164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63968817716467e-06,
+      "loss": 0.9683,
+      "step": 6165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639213961902137e-06,
+      "loss": 0.8901,
+      "step": 6166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638739677015705e-06,
+      "loss": 0.9067,
+      "step": 6167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638265322514452e-06,
+      "loss": 0.8931,
+      "step": 6168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63779089840745e-06,
+      "loss": 0.9355,
+      "step": 6169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.637316404703775e-06,
+      "loss": 0.9204,
+      "step": 6170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636841841412505e-06,
+      "loss": 0.9282,
+      "step": 6171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636367208542719e-06,
+      "loss": 0.873,
+      "step": 6172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6358925061035e-06,
+      "loss": 0.8877,
+      "step": 6173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.635417734103928e-06,
+      "loss": 0.9712,
+      "step": 6174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634942892553085e-06,
+      "loss": 0.8716,
+      "step": 6175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634467981460056e-06,
+      "loss": 0.876,
+      "step": 6176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63399300083393e-06,
+      "loss": 0.856,
+      "step": 6177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633517950683787e-06,
+      "loss": 0.8691,
+      "step": 6178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633042831018724e-06,
+      "loss": 0.9214,
+      "step": 6179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.632567641847824e-06,
+      "loss": 0.9419,
+      "step": 6180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.632092383180182e-06,
+      "loss": 0.9448,
+      "step": 6181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.631617055024889e-06,
+      "loss": 0.9331,
+      "step": 6182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.631141657391037e-06,
+      "loss": 0.9321,
+      "step": 6183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630666190287726e-06,
+      "loss": 0.8359,
+      "step": 6184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630190653724049e-06,
+      "loss": 0.9287,
+      "step": 6185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629715047709104e-06,
+      "loss": 0.9097,
+      "step": 6186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62923937225199e-06,
+      "loss": 0.9575,
+      "step": 6187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62876362736181e-06,
+      "loss": 0.9434,
+      "step": 6188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62828781304766e-06,
+      "loss": 0.9268,
+      "step": 6189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627811929318648e-06,
+      "loss": 0.9229,
+      "step": 6190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627335976183878e-06,
+      "loss": 0.8896,
+      "step": 6191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626859953652454e-06,
+      "loss": 0.9536,
+      "step": 6192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626383861733485e-06,
+      "loss": 0.9819,
+      "step": 6193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625907700436077e-06,
+      "loss": 0.8765,
+      "step": 6194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625431469769342e-06,
+      "loss": 0.9219,
+      "step": 6195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624955169742389e-06,
+      "loss": 0.8911,
+      "step": 6196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624478800364332e-06,
+      "loss": 0.917,
+      "step": 6197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624002361644284e-06,
+      "loss": 0.9546,
+      "step": 6198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62352585359136e-06,
+      "loss": 0.9277,
+      "step": 6199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623049276214677e-06,
+      "loss": 0.8989,
+      "step": 6200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622572629523351e-06,
+      "loss": 0.8628,
+      "step": 6201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622095913526502e-06,
+      "loss": 0.8955,
+      "step": 6202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621619128233252e-06,
+      "loss": 0.8657,
+      "step": 6203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621142273652721e-06,
+      "loss": 0.9478,
+      "step": 6204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62066534979403e-06,
+      "loss": 0.2922,
+      "step": 6205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.620188356666307e-06,
+      "loss": 0.8286,
+      "step": 6206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619711294278677e-06,
+      "loss": 0.9404,
+      "step": 6207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619234162640265e-06,
+      "loss": 0.9272,
+      "step": 6208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.6187569617602e-06,
+      "loss": 0.8823,
+      "step": 6209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61827969164761e-06,
+      "loss": 0.9204,
+      "step": 6210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617802352311631e-06,
+      "loss": 0.9272,
+      "step": 6211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617324943761391e-06,
+      "loss": 0.9155,
+      "step": 6212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.616847466006024e-06,
+      "loss": 0.8794,
+      "step": 6213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.616369919054668e-06,
+      "loss": 0.9199,
+      "step": 6214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615892302916453e-06,
+      "loss": 0.8779,
+      "step": 6215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615414617600523e-06,
+      "loss": 0.9966,
+      "step": 6216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.614936863116012e-06,
+      "loss": 0.9541,
+      "step": 6217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.614459039472064e-06,
+      "loss": 0.9062,
+      "step": 6218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61398114667782e-06,
+      "loss": 0.9058,
+      "step": 6219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.613503184742419e-06,
+      "loss": 0.8315,
+      "step": 6220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61302515367501e-06,
+      "loss": 0.8989,
+      "step": 6221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612547053484735e-06,
+      "loss": 0.8853,
+      "step": 6222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612068884180742e-06,
+      "loss": 0.9385,
+      "step": 6223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611590645772181e-06,
+      "loss": 0.9028,
+      "step": 6224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611112338268197e-06,
+      "loss": 0.9536,
+      "step": 6225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610633961677947e-06,
+      "loss": 0.9434,
+      "step": 6226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610155516010577e-06,
+      "loss": 0.916,
+      "step": 6227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609677001275246e-06,
+      "loss": 0.9209,
+      "step": 6228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609198417481103e-06,
+      "loss": 0.8813,
+      "step": 6229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608719764637307e-06,
+      "loss": 0.8618,
+      "step": 6230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608241042753016e-06,
+      "loss": 0.915,
+      "step": 6231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.607762251837388e-06,
+      "loss": 0.9189,
+      "step": 6232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.607283391899583e-06,
+      "loss": 0.8848,
+      "step": 6233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606804462948764e-06,
+      "loss": 0.9033,
+      "step": 6234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606325464994089e-06,
+      "loss": 0.9272,
+      "step": 6235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.605846398044727e-06,
+      "loss": 0.9058,
+      "step": 6236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60536726210984e-06,
+      "loss": 0.3274,
+      "step": 6237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604888057198597e-06,
+      "loss": 0.9214,
+      "step": 6238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604408783320166e-06,
+      "loss": 0.9087,
+      "step": 6239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603929440483714e-06,
+      "loss": 0.9146,
+      "step": 6240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603450028698411e-06,
+      "loss": 0.8518,
+      "step": 6241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602970547973432e-06,
+      "loss": 0.9795,
+      "step": 6242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60249099831795e-06,
+      "loss": 0.9287,
+      "step": 6243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602011379741137e-06,
+      "loss": 0.8843,
+      "step": 6244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60153169225217e-06,
+      "loss": 0.9414,
+      "step": 6245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601051935860226e-06,
+      "loss": 0.9238,
+      "step": 6246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600572110574485e-06,
+      "loss": 0.9219,
+      "step": 6247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600092216404125e-06,
+      "loss": 0.96,
+      "step": 6248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59961225335833e-06,
+      "loss": 0.9019,
+      "step": 6249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599132221446277e-06,
+      "loss": 0.9346,
+      "step": 6250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.598652120677155e-06,
+      "loss": 0.8652,
+      "step": 6251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.598171951060147e-06,
+      "loss": 0.981,
+      "step": 6252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59769171260444e-06,
+      "loss": 0.8892,
+      "step": 6253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59721140531922e-06,
+      "loss": 0.8652,
+      "step": 6254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596731029213678e-06,
+      "loss": 0.9473,
+      "step": 6255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596250584297004e-06,
+      "loss": 0.9023,
+      "step": 6256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59577007057839e-06,
+      "loss": 0.8984,
+      "step": 6257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595289488067027e-06,
+      "loss": 0.9453,
+      "step": 6258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59480883677211e-06,
+      "loss": 0.9409,
+      "step": 6259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.594328116702837e-06,
+      "loss": 0.9136,
+      "step": 6260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593847327868403e-06,
+      "loss": 0.8838,
+      "step": 6261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593366470278007e-06,
+      "loss": 0.9365,
+      "step": 6262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592885543940846e-06,
+      "loss": 0.9395,
+      "step": 6263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592404548866123e-06,
+      "loss": 0.8945,
+      "step": 6264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59192348506304e-06,
+      "loss": 0.9526,
+      "step": 6265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591442352540802e-06,
+      "loss": 0.9004,
+      "step": 6266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590961151308612e-06,
+      "loss": 0.8818,
+      "step": 6267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590479881375676e-06,
+      "loss": 0.8994,
+      "step": 6268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589998542751203e-06,
+      "loss": 0.8853,
+      "step": 6269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5895171354444e-06,
+      "loss": 0.843,
+      "step": 6270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589035659464478e-06,
+      "loss": 0.2959,
+      "step": 6271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588554114820646e-06,
+      "loss": 0.8901,
+      "step": 6272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588072501522123e-06,
+      "loss": 0.8926,
+      "step": 6273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587590819578116e-06,
+      "loss": 0.9292,
+      "step": 6274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587109068997844e-06,
+      "loss": 0.917,
+      "step": 6275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586627249790523e-06,
+      "loss": 0.8975,
+      "step": 6276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586145361965371e-06,
+      "loss": 0.9658,
+      "step": 6277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585663405531608e-06,
+      "loss": 0.8877,
+      "step": 6278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585181380498454e-06,
+      "loss": 0.9727,
+      "step": 6279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584699286875128e-06,
+      "loss": 0.9209,
+      "step": 6280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584217124670857e-06,
+      "loss": 0.9336,
+      "step": 6281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.583734893894866e-06,
+      "loss": 0.8682,
+      "step": 6282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.583252594556376e-06,
+      "loss": 0.8975,
+      "step": 6283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58277022666462e-06,
+      "loss": 0.8857,
+      "step": 6284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582287790228823e-06,
+      "loss": 0.9102,
+      "step": 6285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581805285258216e-06,
+      "loss": 0.9326,
+      "step": 6286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58132271176203e-06,
+      "loss": 0.897,
+      "step": 6287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580840069749496e-06,
+      "loss": 0.9478,
+      "step": 6288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58035735922985e-06,
+      "loss": 0.853,
+      "step": 6289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579874580212324e-06,
+      "loss": 0.9385,
+      "step": 6290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579391732706157e-06,
+      "loss": 0.9077,
+      "step": 6291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578908816720586e-06,
+      "loss": 0.8506,
+      "step": 6292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578425832264849e-06,
+      "loss": 0.9561,
+      "step": 6293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577942779348189e-06,
+      "loss": 0.936,
+      "step": 6294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577459657979844e-06,
+      "loss": 0.9028,
+      "step": 6295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576976468169058e-06,
+      "loss": 0.9082,
+      "step": 6296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576493209925075e-06,
+      "loss": 0.9839,
+      "step": 6297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576009883257143e-06,
+      "loss": 0.8623,
+      "step": 6298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575526488174506e-06,
+      "loss": 0.8608,
+      "step": 6299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575043024686412e-06,
+      "loss": 0.8774,
+      "step": 6300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574559492802112e-06,
+      "loss": 0.9536,
+      "step": 6301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574075892530856e-06,
+      "loss": 0.8545,
+      "step": 6302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573592223881897e-06,
+      "loss": 0.9272,
+      "step": 6303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573108486864487e-06,
+      "loss": 0.9077,
+      "step": 6304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57262468148788e-06,
+      "loss": 0.9927,
+      "step": 6305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572140807761335e-06,
+      "loss": 0.9102,
+      "step": 6306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571656865694106e-06,
+      "loss": 0.9619,
+      "step": 6307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571172855295452e-06,
+      "loss": 0.8872,
+      "step": 6308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570688776574636e-06,
+      "loss": 0.8535,
+      "step": 6309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570204629540916e-06,
+      "loss": 0.9185,
+      "step": 6310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569720414203555e-06,
+      "loss": 0.8403,
+      "step": 6311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569236130571819e-06,
+      "loss": 0.8745,
+      "step": 6312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568751778654968e-06,
+      "loss": 0.9346,
+      "step": 6313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568267358462277e-06,
+      "loss": 0.8916,
+      "step": 6314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.567782870003006e-06,
+      "loss": 0.8818,
+      "step": 6315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.567298313286424e-06,
+      "loss": 0.8643,
+      "step": 6316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.566813688321807e-06,
+      "loss": 0.9087,
+      "step": 6317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.566328995118423e-06,
+      "loss": 0.9146,
+      "step": 6318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565844233685545e-06,
+      "loss": 0.9229,
+      "step": 6319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565359404032448e-06,
+      "loss": 0.9497,
+      "step": 6320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564874506168408e-06,
+      "loss": 0.8916,
+      "step": 6321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564389540102699e-06,
+      "loss": 0.873,
+      "step": 6322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563904505844602e-06,
+      "loss": 0.939,
+      "step": 6323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563419403403397e-06,
+      "loss": 0.853,
+      "step": 6324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562934232788364e-06,
+      "loss": 0.9263,
+      "step": 6325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562448994008784e-06,
+      "loss": 0.8823,
+      "step": 6326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56196368707394e-06,
+      "loss": 0.9033,
+      "step": 6327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561478311993118e-06,
+      "loss": 0.9146,
+      "step": 6328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560992868775603e-06,
+      "loss": 0.9463,
+      "step": 6329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560507357430685e-06,
+      "loss": 0.9028,
+      "step": 6330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56002177796765e-06,
+      "loss": 0.8359,
+      "step": 6331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559536130395788e-06,
+      "loss": 0.8867,
+      "step": 6332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559050414724391e-06,
+      "loss": 0.9023,
+      "step": 6333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558564630962752e-06,
+      "loss": 0.9019,
+      "step": 6334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558078779120163e-06,
+      "loss": 0.8838,
+      "step": 6335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55759285920592e-06,
+      "loss": 0.9302,
+      "step": 6336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557106871229319e-06,
+      "loss": 0.939,
+      "step": 6337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556620815199659e-06,
+      "loss": 0.8774,
+      "step": 6338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556134691126237e-06,
+      "loss": 0.9312,
+      "step": 6339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.555648499018356e-06,
+      "loss": 0.8838,
+      "step": 6340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.555162238885316e-06,
+      "loss": 0.9624,
+      "step": 6341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554675910736421e-06,
+      "loss": 0.8926,
+      "step": 6342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554189514580972e-06,
+      "loss": 0.9038,
+      "step": 6343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55370305042828e-06,
+      "loss": 0.9229,
+      "step": 6344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.553216518287646e-06,
+      "loss": 0.9399,
+      "step": 6345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552729918168383e-06,
+      "loss": 0.9468,
+      "step": 6346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552243250079796e-06,
+      "loss": 0.9023,
+      "step": 6347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5517565140312e-06,
+      "loss": 0.8652,
+      "step": 6348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551269710031905e-06,
+      "loss": 0.9058,
+      "step": 6349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550782838091222e-06,
+      "loss": 0.9136,
+      "step": 6350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55029589821847e-06,
+      "loss": 0.8999,
+      "step": 6351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.549808890422963e-06,
+      "loss": 0.9165,
+      "step": 6352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.549321814714018e-06,
+      "loss": 0.9165,
+      "step": 6353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548834671100951e-06,
+      "loss": 0.9116,
+      "step": 6354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548347459593087e-06,
+      "loss": 0.9214,
+      "step": 6355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547860180199745e-06,
+      "loss": 0.8613,
+      "step": 6356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547372832930246e-06,
+      "loss": 0.853,
+      "step": 6357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546885417793916e-06,
+      "loss": 0.896,
+      "step": 6358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546397934800077e-06,
+      "loss": 0.9521,
+      "step": 6359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545910383958058e-06,
+      "loss": 0.897,
+      "step": 6360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545422765277185e-06,
+      "loss": 0.8823,
+      "step": 6361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544935078766789e-06,
+      "loss": 0.897,
+      "step": 6362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544447324436198e-06,
+      "loss": 0.9214,
+      "step": 6363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.543959502294743e-06,
+      "loss": 0.9302,
+      "step": 6364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.54347161235176e-06,
+      "loss": 0.9058,
+      "step": 6365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542983654616582e-06,
+      "loss": 0.9609,
+      "step": 6366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542495629098542e-06,
+      "loss": 0.9399,
+      "step": 6367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542007535806977e-06,
+      "loss": 0.9888,
+      "step": 6368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.541519374751228e-06,
+      "loss": 0.3362,
+      "step": 6369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.541031145940634e-06,
+      "loss": 0.8623,
+      "step": 6370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540542849384531e-06,
+      "loss": 0.937,
+      "step": 6371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540054485092268e-06,
+      "loss": 0.9014,
+      "step": 6372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539566053073181e-06,
+      "loss": 0.9517,
+      "step": 6373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539077553336619e-06,
+      "loss": 0.3484,
+      "step": 6374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.538588985891927e-06,
+      "loss": 0.8877,
+      "step": 6375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.538100350748451e-06,
+      "loss": 0.9517,
+      "step": 6376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53761164791554e-06,
+      "loss": 0.9019,
+      "step": 6377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537122877402543e-06,
+      "loss": 0.917,
+      "step": 6378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536634039218812e-06,
+      "loss": 0.9287,
+      "step": 6379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536145133373699e-06,
+      "loss": 0.9609,
+      "step": 6380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535656159876557e-06,
+      "loss": 0.9219,
+      "step": 6381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535167118736742e-06,
+      "loss": 0.8823,
+      "step": 6382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53467800996361e-06,
+      "loss": 0.8491,
+      "step": 6383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534188833566515e-06,
+      "loss": 0.8335,
+      "step": 6384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53369958955482e-06,
+      "loss": 0.8413,
+      "step": 6385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.533210277937882e-06,
+      "loss": 0.8574,
+      "step": 6386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.532720898725066e-06,
+      "loss": 0.9316,
+      "step": 6387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53223145192573e-06,
+      "loss": 0.8931,
+      "step": 6388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531741937549243e-06,
+      "loss": 0.938,
+      "step": 6389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531252355604965e-06,
+      "loss": 0.938,
+      "step": 6390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530762706102265e-06,
+      "loss": 0.9561,
+      "step": 6391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530272989050512e-06,
+      "loss": 0.9136,
+      "step": 6392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.529783204459074e-06,
+      "loss": 0.9341,
+      "step": 6393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52929335233732e-06,
+      "loss": 0.8618,
+      "step": 6394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528803432694621e-06,
+      "loss": 0.9219,
+      "step": 6395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528313445540355e-06,
+      "loss": 0.9087,
+      "step": 6396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.527823390883892e-06,
+      "loss": 0.873,
+      "step": 6397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.527333268734607e-06,
+      "loss": 0.9331,
+      "step": 6398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526843079101878e-06,
+      "loss": 0.916,
+      "step": 6399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526352821995086e-06,
+      "loss": 0.8887,
+      "step": 6400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.525862497423605e-06,
+      "loss": 0.897,
+      "step": 6401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52537210539682e-06,
+      "loss": 0.8901,
+      "step": 6402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524881645924108e-06,
+      "loss": 0.8105,
+      "step": 6403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52439111901486e-06,
+      "loss": 0.9126,
+      "step": 6404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523900524678455e-06,
+      "loss": 0.9736,
+      "step": 6405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523409862924278e-06,
+      "loss": 0.9102,
+      "step": 6406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52291913376172e-06,
+      "loss": 0.9023,
+      "step": 6407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522428337200165e-06,
+      "loss": 0.9121,
+      "step": 6408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.521937473249009e-06,
+      "loss": 0.9316,
+      "step": 6409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.521446541917638e-06,
+      "loss": 0.8799,
+      "step": 6410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520955543215443e-06,
+      "loss": 0.8892,
+      "step": 6411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520464477151822e-06,
+      "loss": 0.9258,
+      "step": 6412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51997334373617e-06,
+      "loss": 0.8857,
+      "step": 6413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519482142977877e-06,
+      "loss": 0.8984,
+      "step": 6414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518990874886349e-06,
+      "loss": 0.8604,
+      "step": 6415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518499539470977e-06,
+      "loss": 0.8901,
+      "step": 6416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518008136741164e-06,
+      "loss": 0.8696,
+      "step": 6417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517516666706314e-06,
+      "loss": 0.8921,
+      "step": 6418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517025129375826e-06,
+      "loss": 0.896,
+      "step": 6419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516533524759106e-06,
+      "loss": 0.8896,
+      "step": 6420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516041852865557e-06,
+      "loss": 0.9404,
+      "step": 6421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51555011370459e-06,
+      "loss": 0.8691,
+      "step": 6422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.515058307285606e-06,
+      "loss": 0.9014,
+      "step": 6423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514566433618018e-06,
+      "loss": 0.8687,
+      "step": 6424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514074492711238e-06,
+      "loss": 0.8721,
+      "step": 6425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513582484574673e-06,
+      "loss": 0.9146,
+      "step": 6426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513090409217738e-06,
+      "loss": 0.8608,
+      "step": 6427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51259826664985e-06,
+      "loss": 0.8945,
+      "step": 6428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51210605688042e-06,
+      "loss": 0.3165,
+      "step": 6429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511613779918868e-06,
+      "loss": 0.8789,
+      "step": 6430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51112143577461e-06,
+      "loss": 0.9209,
+      "step": 6431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.510629024457067e-06,
+      "loss": 0.9097,
+      "step": 6432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.510136545975657e-06,
+      "loss": 0.9541,
+      "step": 6433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509644000339805e-06,
+      "loss": 0.9609,
+      "step": 6434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509151387558933e-06,
+      "loss": 0.9004,
+      "step": 6435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508658707642465e-06,
+      "loss": 0.9165,
+      "step": 6436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508165960599825e-06,
+      "loss": 0.8877,
+      "step": 6437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.507673146440446e-06,
+      "loss": 0.916,
+      "step": 6438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.50718026517375e-06,
+      "loss": 0.9028,
+      "step": 6439
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506687316809168e-06,
+      "loss": 0.917,
+      "step": 6440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506194301356132e-06,
+      "loss": 0.8755,
+      "step": 6441
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505701218824076e-06,
+      "loss": 0.8965,
+      "step": 6442
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505208069222429e-06,
+      "loss": 0.9409,
+      "step": 6443
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50471485256063e-06,
+      "loss": 0.9292,
+      "step": 6444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.504221568848111e-06,
+      "loss": 0.876,
+      "step": 6445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.503728218094314e-06,
+      "loss": 0.8271,
+      "step": 6446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.503234800308673e-06,
+      "loss": 0.9224,
+      "step": 6447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502741315500631e-06,
+      "loss": 0.9775,
+      "step": 6448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502247763679626e-06,
+      "loss": 0.915,
+      "step": 6449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501754144855104e-06,
+      "loss": 0.8779,
+      "step": 6450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501260459036507e-06,
+      "loss": 0.9033,
+      "step": 6451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500766706233281e-06,
+      "loss": 0.9004,
+      "step": 6452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50027288645487e-06,
+      "loss": 0.9175,
+      "step": 6453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.499778999710723e-06,
+      "loss": 0.9072,
+      "step": 6454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49928504601029e-06,
+      "loss": 0.9028,
+      "step": 6455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49879102536302e-06,
+      "loss": 0.9126,
+      "step": 6456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498296937778364e-06,
+      "loss": 0.3257,
+      "step": 6457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497802783265773e-06,
+      "loss": 0.9624,
+      "step": 6458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497308561834704e-06,
+      "loss": 0.9146,
+      "step": 6459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49681427349461e-06,
+      "loss": 0.8569,
+      "step": 6460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.496319918254951e-06,
+      "loss": 0.9175,
+      "step": 6461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.495825496125182e-06,
+      "loss": 0.8975,
+      "step": 6462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.495331007114763e-06,
+      "loss": 0.8945,
+      "step": 6463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494836451233152e-06,
+      "loss": 0.8906,
+      "step": 6464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494341828489812e-06,
+      "loss": 0.8906,
+      "step": 6465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49384713889421e-06,
+      "loss": 0.9272,
+      "step": 6466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.493352382455804e-06,
+      "loss": 0.8857,
+      "step": 6467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492857559184062e-06,
+      "loss": 0.9014,
+      "step": 6468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492362669088452e-06,
+      "loss": 0.8804,
+      "step": 6469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491867712178438e-06,
+      "loss": 0.9492,
+      "step": 6470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491372688463495e-06,
+      "loss": 0.916,
+      "step": 6471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490877597953091e-06,
+      "loss": 0.9062,
+      "step": 6472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490382440656697e-06,
+      "loss": 0.9053,
+      "step": 6473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489887216583786e-06,
+      "loss": 0.917,
+      "step": 6474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489391925743834e-06,
+      "loss": 0.9351,
+      "step": 6475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488896568146316e-06,
+      "loss": 0.9146,
+      "step": 6476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488401143800708e-06,
+      "loss": 0.9058,
+      "step": 6477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487905652716488e-06,
+      "loss": 0.9053,
+      "step": 6478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487410094903139e-06,
+      "loss": 0.9038,
+      "step": 6479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486914470370138e-06,
+      "loss": 0.9238,
+      "step": 6480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48641877912697e-06,
+      "loss": 0.8931,
+      "step": 6481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.485923021183116e-06,
+      "loss": 0.9062,
+      "step": 6482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48542719654806e-06,
+      "loss": 0.9258,
+      "step": 6483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484931305231293e-06,
+      "loss": 0.9229,
+      "step": 6484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484435347242296e-06,
+      "loss": 0.8867,
+      "step": 6485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.483939322590558e-06,
+      "loss": 0.9194,
+      "step": 6486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.483443231285574e-06,
+      "loss": 0.8447,
+      "step": 6487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48294707333683e-06,
+      "loss": 0.9502,
+      "step": 6488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482450848753822e-06,
+      "loss": 0.9121,
+      "step": 6489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48195455754604e-06,
+      "loss": 0.9004,
+      "step": 6490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.481458199722981e-06,
+      "loss": 0.8975,
+      "step": 6491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480961775294141e-06,
+      "loss": 0.8965,
+      "step": 6492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480465284269015e-06,
+      "loss": 0.9492,
+      "step": 6493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.479968726657105e-06,
+      "loss": 0.8809,
+      "step": 6494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47947210246791e-06,
+      "loss": 0.916,
+      "step": 6495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47897541171093e-06,
+      "loss": 0.8574,
+      "step": 6496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.478478654395668e-06,
+      "loss": 1.019,
+      "step": 6497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477981830531628e-06,
+      "loss": 0.9058,
+      "step": 6498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477484940128315e-06,
+      "loss": 0.8945,
+      "step": 6499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476987983195235e-06,
+      "loss": 0.9111,
+      "step": 6500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476490959741896e-06,
+      "loss": 0.8398,
+      "step": 6501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475993869777807e-06,
+      "loss": 0.9702,
+      "step": 6502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475496713312479e-06,
+      "loss": 0.895,
+      "step": 6503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47499949035542e-06,
+      "loss": 0.9126,
+      "step": 6504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.474502200916146e-06,
+      "loss": 0.8682,
+      "step": 6505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47400484500417e-06,
+      "loss": 0.9121,
+      "step": 6506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.473507422629005e-06,
+      "loss": 0.9717,
+      "step": 6507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.473009933800173e-06,
+      "loss": 0.8315,
+      "step": 6508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472512378527186e-06,
+      "loss": 0.9136,
+      "step": 6509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472014756819564e-06,
+      "loss": 0.9448,
+      "step": 6510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47151706868683e-06,
+      "loss": 0.9033,
+      "step": 6511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471019314138504e-06,
+      "loss": 0.9336,
+      "step": 6512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470521493184109e-06,
+      "loss": 0.8921,
+      "step": 6513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470023605833168e-06,
+      "loss": 0.8843,
+      "step": 6514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469525652095208e-06,
+      "loss": 0.9282,
+      "step": 6515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469027631979756e-06,
+      "loss": 0.8501,
+      "step": 6516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468529545496338e-06,
+      "loss": 0.937,
+      "step": 6517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468031392654483e-06,
+      "loss": 0.9521,
+      "step": 6518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467533173463723e-06,
+      "loss": 0.9541,
+      "step": 6519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.46703488793359e-06,
+      "loss": 0.9365,
+      "step": 6520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466536536073618e-06,
+      "loss": 0.9517,
+      "step": 6521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466038117893336e-06,
+      "loss": 0.9272,
+      "step": 6522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465539633402284e-06,
+      "loss": 0.8882,
+      "step": 6523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465041082609998e-06,
+      "loss": 0.9609,
+      "step": 6524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464542465526016e-06,
+      "loss": 0.9253,
+      "step": 6525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464043782159878e-06,
+      "loss": 0.8687,
+      "step": 6526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463545032521123e-06,
+      "loss": 0.3656,
+      "step": 6527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463046216619294e-06,
+      "loss": 0.9116,
+      "step": 6528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462547334463933e-06,
+      "loss": 0.9077,
+      "step": 6529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462048386064586e-06,
+      "loss": 0.9224,
+      "step": 6530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461549371430797e-06,
+      "loss": 0.8853,
+      "step": 6531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461050290572114e-06,
+      "loss": 0.8564,
+      "step": 6532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460551143498085e-06,
+      "loss": 0.9702,
+      "step": 6533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460051930218262e-06,
+      "loss": 0.8872,
+      "step": 6534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45955265074219e-06,
+      "loss": 0.9004,
+      "step": 6535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459053305079426e-06,
+      "loss": 0.9229,
+      "step": 6536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45855389323952e-06,
+      "loss": 0.9395,
+      "step": 6537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45805441523203e-06,
+      "loss": 0.9146,
+      "step": 6538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457554871066509e-06,
+      "loss": 0.9863,
+      "step": 6539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457055260752514e-06,
+      "loss": 0.9697,
+      "step": 6540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.456555584299603e-06,
+      "loss": 0.9819,
+      "step": 6541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45605584171734e-06,
+      "loss": 0.9268,
+      "step": 6542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.455556033015281e-06,
+      "loss": 0.9072,
+      "step": 6543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45505615820299e-06,
+      "loss": 0.9639,
+      "step": 6544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454556217290031e-06,
+      "loss": 0.9399,
+      "step": 6545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454056210285967e-06,
+      "loss": 0.8433,
+      "step": 6546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453556137200365e-06,
+      "loss": 0.9321,
+      "step": 6547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45305599804279e-06,
+      "loss": 0.9336,
+      "step": 6548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.452555792822816e-06,
+      "loss": 0.9395,
+      "step": 6549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.452055521550007e-06,
+      "loss": 0.8545,
+      "step": 6550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451555184233938e-06,
+      "loss": 0.8828,
+      "step": 6551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451054780884175e-06,
+      "loss": 0.9561,
+      "step": 6552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4505543115103e-06,
+      "loss": 0.834,
+      "step": 6553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450053776121882e-06,
+      "loss": 0.9058,
+      "step": 6554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4495531747285e-06,
+      "loss": 0.9028,
+      "step": 6555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.449052507339727e-06,
+      "loss": 0.9004,
+      "step": 6556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448551773965148e-06,
+      "loss": 0.8965,
+      "step": 6557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448050974614337e-06,
+      "loss": 0.8599,
+      "step": 6558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44755010929688e-06,
+      "loss": 0.9307,
+      "step": 6559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447049178022352e-06,
+      "loss": 0.9058,
+      "step": 6560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.446548180800346e-06,
+      "loss": 0.4414,
+      "step": 6561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44604711764044e-06,
+      "loss": 0.9575,
+      "step": 6562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445545988552224e-06,
+      "loss": 0.833,
+      "step": 6563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44504479354528e-06,
+      "loss": 0.896,
+      "step": 6564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444543532629204e-06,
+      "loss": 0.874,
+      "step": 6565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444042205813582e-06,
+      "loss": 0.9243,
+      "step": 6566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443540813108005e-06,
+      "loss": 0.8779,
+      "step": 6567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443039354522064e-06,
+      "loss": 0.8716,
+      "step": 6568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442537830065356e-06,
+      "loss": 0.9365,
+      "step": 6569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442036239747477e-06,
+      "loss": 0.873,
+      "step": 6570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.441534583578016e-06,
+      "loss": 0.8721,
+      "step": 6571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44103286156658e-06,
+      "loss": 0.8882,
+      "step": 6572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440531073722759e-06,
+      "loss": 0.9165,
+      "step": 6573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44002922005616e-06,
+      "loss": 0.8604,
+      "step": 6574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43952730057638e-06,
+      "loss": 0.9053,
+      "step": 6575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439025315293022e-06,
+      "loss": 0.8965,
+      "step": 6576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43852326421569e-06,
+      "loss": 0.9189,
+      "step": 6577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.438021147353992e-06,
+      "loss": 0.8906,
+      "step": 6578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43751896471753e-06,
+      "loss": 0.9336,
+      "step": 6579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437016716315915e-06,
+      "loss": 0.9263,
+      "step": 6580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436514402158751e-06,
+      "loss": 0.8867,
+      "step": 6581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436012022255655e-06,
+      "loss": 0.9185,
+      "step": 6582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435509576616233e-06,
+      "loss": 0.9219,
+      "step": 6583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435007065250098e-06,
+      "loss": 0.8896,
+      "step": 6584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.434504488166864e-06,
+      "loss": 0.8506,
+      "step": 6585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43400184537615e-06,
+      "loss": 0.9248,
+      "step": 6586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433499136887569e-06,
+      "loss": 0.9521,
+      "step": 6587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.432996362710737e-06,
+      "loss": 0.9321,
+      "step": 6588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.432493522855276e-06,
+      "loss": 0.8706,
+      "step": 6589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431990617330805e-06,
+      "loss": 0.8438,
+      "step": 6590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431487646146946e-06,
+      "loss": 0.9595,
+      "step": 6591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.430984609313319e-06,
+      "loss": 0.8809,
+      "step": 6592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.430481506839552e-06,
+      "loss": 0.8945,
+      "step": 6593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429978338735266e-06,
+      "loss": 0.8838,
+      "step": 6594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429475105010093e-06,
+      "loss": 0.873,
+      "step": 6595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428971805673654e-06,
+      "loss": 0.9595,
+      "step": 6596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42846844073558e-06,
+      "loss": 0.874,
+      "step": 6597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427965010205504e-06,
+      "loss": 0.8984,
+      "step": 6598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427461514093056e-06,
+      "loss": 0.9155,
+      "step": 6599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426957952407867e-06,
+      "loss": 0.9453,
+      "step": 6600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426454325159572e-06,
+      "loss": 0.939,
+      "step": 6601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.425950632357807e-06,
+      "loss": 0.9404,
+      "step": 6602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.425446874012209e-06,
+      "loss": 0.9209,
+      "step": 6603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424943050132413e-06,
+      "loss": 0.8843,
+      "step": 6604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424439160728058e-06,
+      "loss": 0.9067,
+      "step": 6605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423935205808789e-06,
+      "loss": 0.9131,
+      "step": 6606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423431185384241e-06,
+      "loss": 0.8486,
+      "step": 6607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.422927099464063e-06,
+      "loss": 0.915,
+      "step": 6608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.422422948057892e-06,
+      "loss": 0.9663,
+      "step": 6609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.421918731175378e-06,
+      "loss": 0.9004,
+      "step": 6610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.421414448826169e-06,
+      "loss": 0.9478,
+      "step": 6611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420910101019907e-06,
+      "loss": 0.9072,
+      "step": 6612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420405687766247e-06,
+      "loss": 0.897,
+      "step": 6613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419901209074833e-06,
+      "loss": 0.9458,
+      "step": 6614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419396664955321e-06,
+      "loss": 0.8979,
+      "step": 6615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418892055417364e-06,
+      "loss": 0.8721,
+      "step": 6616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418387380470613e-06,
+      "loss": 0.9087,
+      "step": 6617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417882640124725e-06,
+      "loss": 0.8521,
+      "step": 6618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417377834389356e-06,
+      "loss": 0.8589,
+      "step": 6619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.416872963274166e-06,
+      "loss": 0.9062,
+      "step": 6620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41636802678881e-06,
+      "loss": 0.8882,
+      "step": 6621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415863024942949e-06,
+      "loss": 0.9634,
+      "step": 6622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415357957746247e-06,
+      "loss": 0.8804,
+      "step": 6623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.414852825208366e-06,
+      "loss": 0.8896,
+      "step": 6624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.414347627338967e-06,
+      "loss": 0.9282,
+      "step": 6625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41384236414772e-06,
+      "loss": 0.8926,
+      "step": 6626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.413337035644287e-06,
+      "loss": 0.8989,
+      "step": 6627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41283164183834e-06,
+      "loss": 0.8569,
+      "step": 6628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.412326182739544e-06,
+      "loss": 0.8862,
+      "step": 6629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41182065835757e-06,
+      "loss": 0.8604,
+      "step": 6630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411315068702093e-06,
+      "loss": 0.8867,
+      "step": 6631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410809413782781e-06,
+      "loss": 0.9653,
+      "step": 6632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410303693609312e-06,
+      "loss": 0.9336,
+      "step": 6633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409797908191359e-06,
+      "loss": 0.8496,
+      "step": 6634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409292057538598e-06,
+      "loss": 0.8965,
+      "step": 6635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.408786141660707e-06,
+      "loss": 0.916,
+      "step": 6636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.408280160567368e-06,
+      "loss": 0.8755,
+      "step": 6637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407774114268256e-06,
+      "loss": 0.9136,
+      "step": 6638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407268002773056e-06,
+      "loss": 0.8657,
+      "step": 6639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.406761826091452e-06,
+      "loss": 0.9238,
+      "step": 6640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.406255584233125e-06,
+      "loss": 0.9385,
+      "step": 6641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40574927720776e-06,
+      "loss": 0.8547,
+      "step": 6642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.405242905025045e-06,
+      "loss": 0.9385,
+      "step": 6643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404736467694666e-06,
+      "loss": 0.8643,
+      "step": 6644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404229965226315e-06,
+      "loss": 0.9048,
+      "step": 6645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40372339762968e-06,
+      "loss": 0.8828,
+      "step": 6646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403216764914453e-06,
+      "loss": 0.8257,
+      "step": 6647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.402710067090324e-06,
+      "loss": 0.8813,
+      "step": 6648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.402203304166992e-06,
+      "loss": 0.9438,
+      "step": 6649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401696476154149e-06,
+      "loss": 0.8813,
+      "step": 6650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401189583061492e-06,
+      "loss": 0.8701,
+      "step": 6651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40068262489872e-06,
+      "loss": 0.9331,
+      "step": 6652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.400175601675529e-06,
+      "loss": 0.9189,
+      "step": 6653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39966851340162e-06,
+      "loss": 0.8745,
+      "step": 6654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399161360086697e-06,
+      "loss": 0.9097,
+      "step": 6655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39865414174046e-06,
+      "loss": 0.9629,
+      "step": 6656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.398146858372614e-06,
+      "loss": 0.9131,
+      "step": 6657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.397639509992861e-06,
+      "loss": 0.981,
+      "step": 6658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.397132096610913e-06,
+      "loss": 0.9048,
+      "step": 6659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396624618236476e-06,
+      "loss": 0.896,
+      "step": 6660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396117074879255e-06,
+      "loss": 0.8699,
+      "step": 6661
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395609466548964e-06,
+      "loss": 0.854,
+      "step": 6662
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395101793255311e-06,
+      "loss": 0.9009,
+      "step": 6663
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394594055008013e-06,
+      "loss": 0.9336,
+      "step": 6664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394086251816781e-06,
+      "loss": 0.915,
+      "step": 6665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39357838369133e-06,
+      "loss": 0.9253,
+      "step": 6666
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.393070450641378e-06,
+      "loss": 0.9224,
+      "step": 6667
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392562452676641e-06,
+      "loss": 0.9282,
+      "step": 6668
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392054389806837e-06,
+      "loss": 0.9214,
+      "step": 6669
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39154626204169e-06,
+      "loss": 0.8682,
+      "step": 6670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.391038069390915e-06,
+      "loss": 0.8882,
+      "step": 6671
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39052981186424e-06,
+      "loss": 0.8989,
+      "step": 6672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390021489471389e-06,
+      "loss": 0.9502,
+      "step": 6673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.389513102222083e-06,
+      "loss": 0.8892,
+      "step": 6674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.389004650126049e-06,
+      "loss": 0.9053,
+      "step": 6675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.388496133193017e-06,
+      "loss": 0.9219,
+      "step": 6676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.387987551432714e-06,
+      "loss": 0.8755,
+      "step": 6677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38747890485487e-06,
+      "loss": 0.9062,
+      "step": 6678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386970193469217e-06,
+      "loss": 0.9141,
+      "step": 6679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386461417285488e-06,
+      "loss": 0.9292,
+      "step": 6680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385952576313414e-06,
+      "loss": 0.9062,
+      "step": 6681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385443670562732e-06,
+      "loss": 0.3754,
+      "step": 6682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38493470004318e-06,
+      "loss": 0.9771,
+      "step": 6683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38442566476449e-06,
+      "loss": 0.8984,
+      "step": 6684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383916564736404e-06,
+      "loss": 0.9209,
+      "step": 6685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383407399968664e-06,
+      "loss": 0.9629,
+      "step": 6686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382898170471009e-06,
+      "loss": 0.8945,
+      "step": 6687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382388876253178e-06,
+      "loss": 0.9336,
+      "step": 6688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38187951732492e-06,
+      "loss": 0.8838,
+      "step": 6689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381370093695978e-06,
+      "loss": 0.9097,
+      "step": 6690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.380860605376096e-06,
+      "loss": 0.9067,
+      "step": 6691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.380351052375023e-06,
+      "loss": 0.8794,
+      "step": 6692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379841434702509e-06,
+      "loss": 0.9009,
+      "step": 6693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379331752368302e-06,
+      "loss": 0.9492,
+      "step": 6694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378822005382154e-06,
+      "loss": 0.8809,
+      "step": 6695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378312193753817e-06,
+      "loss": 0.9204,
+      "step": 6696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377802317493043e-06,
+      "loss": 0.959,
+      "step": 6697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377292376609588e-06,
+      "loss": 0.8794,
+      "step": 6698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376782371113208e-06,
+      "loss": 0.8853,
+      "step": 6699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37627230101366e-06,
+      "loss": 0.8062,
+      "step": 6700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375762166320702e-06,
+      "loss": 0.895,
+      "step": 6701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375251967044096e-06,
+      "loss": 0.9404,
+      "step": 6702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3747417031936e-06,
+      "loss": 0.9214,
+      "step": 6703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.374231374778977e-06,
+      "loss": 0.9033,
+      "step": 6704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373720981809991e-06,
+      "loss": 0.8975,
+      "step": 6705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373210524296407e-06,
+      "loss": 0.9263,
+      "step": 6706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37270000224799e-06,
+      "loss": 0.9761,
+      "step": 6707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.372189415674506e-06,
+      "loss": 0.8911,
+      "step": 6708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371678764585726e-06,
+      "loss": 0.8687,
+      "step": 6709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371168048991418e-06,
+      "loss": 0.8574,
+      "step": 6710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370657268901353e-06,
+      "loss": 0.918,
+      "step": 6711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370146424325301e-06,
+      "loss": 0.8848,
+      "step": 6712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.369635515273038e-06,
+      "loss": 0.9473,
+      "step": 6713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.369124541754338e-06,
+      "loss": 0.9697,
+      "step": 6714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368613503778977e-06,
+      "loss": 0.9009,
+      "step": 6715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368102401356732e-06,
+      "loss": 0.3109,
+      "step": 6716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.367591234497378e-06,
+      "loss": 0.9014,
+      "step": 6717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.367080003210699e-06,
+      "loss": 0.9395,
+      "step": 6718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366568707506475e-06,
+      "loss": 0.9297,
+      "step": 6719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366057347394485e-06,
+      "loss": 0.9126,
+      "step": 6720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365545922884513e-06,
+      "loss": 0.894,
+      "step": 6721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365034433986345e-06,
+      "loss": 0.9121,
+      "step": 6722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364522880709764e-06,
+      "loss": 0.8892,
+      "step": 6723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364011263064562e-06,
+      "loss": 0.9165,
+      "step": 6724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.36349958106052e-06,
+      "loss": 0.9321,
+      "step": 6725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362987834707433e-06,
+      "loss": 0.8999,
+      "step": 6726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362476024015088e-06,
+      "loss": 0.8716,
+      "step": 6727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361964148993277e-06,
+      "loss": 0.9194,
+      "step": 6728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361452209651796e-06,
+      "loss": 0.8809,
+      "step": 6729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360940206000438e-06,
+      "loss": 0.9663,
+      "step": 6730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360428138048996e-06,
+      "loss": 0.936,
+      "step": 6731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35991600580727e-06,
+      "loss": 0.9214,
+      "step": 6732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.359403809285054e-06,
+      "loss": 0.9028,
+      "step": 6733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358891548492149e-06,
+      "loss": 0.895,
+      "step": 6734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358379223438359e-06,
+      "loss": 0.8789,
+      "step": 6735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35786683413348e-06,
+      "loss": 0.9668,
+      "step": 6736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.357354380587317e-06,
+      "loss": 0.9058,
+      "step": 6737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.356841862809674e-06,
+      "loss": 0.8833,
+      "step": 6738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.356329280810358e-06,
+      "loss": 0.875,
+      "step": 6739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355816634599174e-06,
+      "loss": 0.9053,
+      "step": 6740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355303924185929e-06,
+      "loss": 0.8574,
+      "step": 6741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.354791149580432e-06,
+      "loss": 0.9395,
+      "step": 6742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.354278310792494e-06,
+      "loss": 0.9521,
+      "step": 6743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353765407831926e-06,
+      "loss": 0.9482,
+      "step": 6744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353252440708543e-06,
+      "loss": 0.958,
+      "step": 6745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.352739409432153e-06,
+      "loss": 0.9502,
+      "step": 6746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.352226314012577e-06,
+      "loss": 0.8721,
+      "step": 6747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35171315445963e-06,
+      "loss": 0.8521,
+      "step": 6748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.351199930783128e-06,
+      "loss": 0.918,
+      "step": 6749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35068664299289e-06,
+      "loss": 0.8394,
+      "step": 6750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350173291098737e-06,
+      "loss": 0.9146,
+      "step": 6751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349659875110489e-06,
+      "loss": 0.9116,
+      "step": 6752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349146395037971e-06,
+      "loss": 0.9692,
+      "step": 6753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.348632850891002e-06,
+      "loss": 0.9268,
+      "step": 6754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.348119242679411e-06,
+      "loss": 0.9399,
+      "step": 6755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347605570413025e-06,
+      "loss": 0.9297,
+      "step": 6756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347091834101667e-06,
+      "loss": 0.9043,
+      "step": 6757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.346578033755168e-06,
+      "loss": 0.8599,
+      "step": 6758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.346064169383356e-06,
+      "loss": 0.8838,
+      "step": 6759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345550240996068e-06,
+      "loss": 0.9043,
+      "step": 6760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345036248603129e-06,
+      "loss": 0.8486,
+      "step": 6761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344522192214374e-06,
+      "loss": 0.8774,
+      "step": 6762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344008071839643e-06,
+      "loss": 0.9497,
+      "step": 6763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343493887488767e-06,
+      "loss": 0.8657,
+      "step": 6764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342979639171583e-06,
+      "loss": 0.8721,
+      "step": 6765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.34246532689793e-06,
+      "loss": 0.9326,
+      "step": 6766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341950950677651e-06,
+      "loss": 0.9043,
+      "step": 6767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341436510520582e-06,
+      "loss": 0.8842,
+      "step": 6768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.340922006436568e-06,
+      "loss": 0.8418,
+      "step": 6769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.340407438435451e-06,
+      "loss": 0.8682,
+      "step": 6770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339892806527075e-06,
+      "loss": 0.9727,
+      "step": 6771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339378110721287e-06,
+      "loss": 0.8687,
+      "step": 6772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338863351027936e-06,
+      "loss": 0.9443,
+      "step": 6773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338348527456865e-06,
+      "loss": 0.3246,
+      "step": 6774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337833640017926e-06,
+      "loss": 0.8984,
+      "step": 6775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337318688720971e-06,
+      "loss": 0.8911,
+      "step": 6776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33680367357585e-06,
+      "loss": 0.8804,
+      "step": 6777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.336288594592417e-06,
+      "loss": 0.8994,
+      "step": 6778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.335773451780525e-06,
+      "loss": 0.8555,
+      "step": 6779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33525824515003e-06,
+      "loss": 0.8779,
+      "step": 6780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33474297471079e-06,
+      "loss": 0.8833,
+      "step": 6781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.334227640472663e-06,
+      "loss": 0.876,
+      "step": 6782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333712242445504e-06,
+      "loss": 0.8491,
+      "step": 6783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333196780639179e-06,
+      "loss": 0.937,
+      "step": 6784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332681255063547e-06,
+      "loss": 0.9302,
+      "step": 6785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332165665728471e-06,
+      "loss": 0.9199,
+      "step": 6786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.331650012643813e-06,
+      "loss": 0.8633,
+      "step": 6787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.331134295819444e-06,
+      "loss": 0.9126,
+      "step": 6788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330618515265223e-06,
+      "loss": 0.9141,
+      "step": 6789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330102670991022e-06,
+      "loss": 0.9521,
+      "step": 6790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32958676300671e-06,
+      "loss": 0.9482,
+      "step": 6791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329070791322158e-06,
+      "loss": 0.9004,
+      "step": 6792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328554755947233e-06,
+      "loss": 0.9287,
+      "step": 6793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328038656891812e-06,
+      "loss": 0.958,
+      "step": 6794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327522494165767e-06,
+      "loss": 0.9229,
+      "step": 6795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327006267778975e-06,
+      "loss": 0.8999,
+      "step": 6796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.326489977741307e-06,
+      "loss": 0.9531,
+      "step": 6797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325973624062646e-06,
+      "loss": 0.9292,
+      "step": 6798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325457206752868e-06,
+      "loss": 0.9146,
+      "step": 6799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324940725821853e-06,
+      "loss": 0.9263,
+      "step": 6800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324424181279483e-06,
+      "loss": 0.9102,
+      "step": 6801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323907573135639e-06,
+      "loss": 0.9507,
+      "step": 6802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323390901400204e-06,
+      "loss": 0.9268,
+      "step": 6803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.322874166083066e-06,
+      "loss": 0.3142,
+      "step": 6804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32235736719411e-06,
+      "loss": 0.98,
+      "step": 6805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32184050474322e-06,
+      "loss": 0.9424,
+      "step": 6806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321323578740288e-06,
+      "loss": 0.8955,
+      "step": 6807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.320806589195201e-06,
+      "loss": 0.9219,
+      "step": 6808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.320289536117851e-06,
+      "loss": 0.9326,
+      "step": 6809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319772419518132e-06,
+      "loss": 0.9336,
+      "step": 6810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319255239405935e-06,
+      "loss": 0.957,
+      "step": 6811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.318737995791154e-06,
+      "loss": 0.9634,
+      "step": 6812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.318220688683684e-06,
+      "loss": 0.8975,
+      "step": 6813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317703318093425e-06,
+      "loss": 0.8755,
+      "step": 6814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317185884030272e-06,
+      "loss": 0.8965,
+      "step": 6815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.316668386504128e-06,
+      "loss": 0.8774,
+      "step": 6816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.316150825524888e-06,
+      "loss": 0.9873,
+      "step": 6817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315633201102461e-06,
+      "loss": 0.896,
+      "step": 6818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315115513246745e-06,
+      "loss": 0.9287,
+      "step": 6819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314597761967644e-06,
+      "loss": 0.9111,
+      "step": 6820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314079947275066e-06,
+      "loss": 0.8887,
+      "step": 6821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.313562069178916e-06,
+      "loss": 0.9131,
+      "step": 6822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3130441276891e-06,
+      "loss": 0.8436,
+      "step": 6823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.312526122815532e-06,
+      "loss": 0.8882,
+      "step": 6824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.312008054568119e-06,
+      "loss": 0.9121,
+      "step": 6825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31148992295677e-06,
+      "loss": 0.8745,
+      "step": 6826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.310971727991404e-06,
+      "loss": 0.8726,
+      "step": 6827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31045346968193e-06,
+      "loss": 0.9663,
+      "step": 6828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309935148038265e-06,
+      "loss": 0.8931,
+      "step": 6829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309416763070324e-06,
+      "loss": 0.8955,
+      "step": 6830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308898314788025e-06,
+      "loss": 0.8887,
+      "step": 6831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308379803201287e-06,
+      "loss": 0.9072,
+      "step": 6832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30786122832003e-06,
+      "loss": 0.8354,
+      "step": 6833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.307342590154178e-06,
+      "loss": 0.8799,
+      "step": 6834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306823888713647e-06,
+      "loss": 0.8184,
+      "step": 6835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306305124008365e-06,
+      "loss": 0.9517,
+      "step": 6836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305786296048257e-06,
+      "loss": 0.9067,
+      "step": 6837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305267404843247e-06,
+      "loss": 0.9531,
+      "step": 6838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304748450403264e-06,
+      "loss": 0.9653,
+      "step": 6839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304229432738233e-06,
+      "loss": 0.9111,
+      "step": 6840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30371035185809e-06,
+      "loss": 0.9492,
+      "step": 6841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30319120777276e-06,
+      "loss": 0.9463,
+      "step": 6842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302672000492176e-06,
+      "loss": 0.9438,
+      "step": 6843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302152730026274e-06,
+      "loss": 0.8794,
+      "step": 6844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301633396384986e-06,
+      "loss": 0.915,
+      "step": 6845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301113999578248e-06,
+      "loss": 0.8911,
+      "step": 6846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300594539615998e-06,
+      "loss": 0.8613,
+      "step": 6847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300075016508173e-06,
+      "loss": 0.8398,
+      "step": 6848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299555430264712e-06,
+      "loss": 0.8562,
+      "step": 6849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299035780895556e-06,
+      "loss": 0.8911,
+      "step": 6850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298516068410649e-06,
+      "loss": 0.8833,
+      "step": 6851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29799629281993e-06,
+      "loss": 0.875,
+      "step": 6852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.297476454133346e-06,
+      "loss": 0.918,
+      "step": 6853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29695655236084e-06,
+      "loss": 0.8945,
+      "step": 6854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29643658751236e-06,
+      "loss": 0.9248,
+      "step": 6855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295916559597855e-06,
+      "loss": 0.9023,
+      "step": 6856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29539646862727e-06,
+      "loss": 0.9185,
+      "step": 6857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.294876314610561e-06,
+      "loss": 0.9146,
+      "step": 6858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.294356097557673e-06,
+      "loss": 0.9263,
+      "step": 6859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.293835817478564e-06,
+      "loss": 0.9067,
+      "step": 6860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.293315474383184e-06,
+      "loss": 0.9238,
+      "step": 6861
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292795068281488e-06,
+      "loss": 1.002,
+      "step": 6862
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292274599183437e-06,
+      "loss": 0.8896,
+      "step": 6863
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291754067098982e-06,
+      "loss": 0.9268,
+      "step": 6864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291233472038085e-06,
+      "loss": 0.9185,
+      "step": 6865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290712814010707e-06,
+      "loss": 0.9194,
+      "step": 6866
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290192093026805e-06,
+      "loss": 0.9302,
+      "step": 6867
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289671309096346e-06,
+      "loss": 0.3282,
+      "step": 6868
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289150462229288e-06,
+      "loss": 0.8726,
+      "step": 6869
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288629552435601e-06,
+      "loss": 0.957,
+      "step": 6870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288108579725248e-06,
+      "loss": 0.9082,
+      "step": 6871
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.287587544108195e-06,
+      "loss": 0.9194,
+      "step": 6872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.287066445594414e-06,
+      "loss": 0.9839,
+      "step": 6873
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.286545284193871e-06,
+      "loss": 0.7998,
+      "step": 6874
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.286024059916538e-06,
+      "loss": 0.9009,
+      "step": 6875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285502772772387e-06,
+      "loss": 0.8862,
+      "step": 6876
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28498142277139e-06,
+      "loss": 0.8545,
+      "step": 6877
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28446000992352e-06,
+      "loss": 0.896,
+      "step": 6878
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283938534238758e-06,
+      "loss": 0.8984,
+      "step": 6879
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283416995727075e-06,
+      "loss": 0.8984,
+      "step": 6880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282895394398452e-06,
+      "loss": 0.8818,
+      "step": 6881
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282373730262866e-06,
+      "loss": 0.9028,
+      "step": 6882
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281852003330298e-06,
+      "loss": 0.8169,
+      "step": 6883
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28133021361073e-06,
+      "loss": 0.8413,
+      "step": 6884
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280808361114142e-06,
+      "loss": 0.9053,
+      "step": 6885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280286445850523e-06,
+      "loss": 0.9302,
+      "step": 6886
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279764467829856e-06,
+      "loss": 0.9038,
+      "step": 6887
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279242427062125e-06,
+      "loss": 0.9097,
+      "step": 6888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278720323557318e-06,
+      "loss": 0.9009,
+      "step": 6889
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278198157325426e-06,
+      "loss": 0.9971,
+      "step": 6890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.277675928376437e-06,
+      "loss": 0.916,
+      "step": 6891
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27715363672034e-06,
+      "loss": 0.3643,
+      "step": 6892
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276631282367133e-06,
+      "loss": 0.4199,
+      "step": 6893
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276108865326806e-06,
+      "loss": 0.9146,
+      "step": 6894
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275586385609353e-06,
+      "loss": 0.8413,
+      "step": 6895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27506384322477e-06,
+      "loss": 0.9058,
+      "step": 6896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274541238183058e-06,
+      "loss": 0.9731,
+      "step": 6897
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274018570494208e-06,
+      "loss": 0.8809,
+      "step": 6898
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273495840168226e-06,
+      "loss": 0.873,
+      "step": 6899
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27297304721511e-06,
+      "loss": 0.9297,
+      "step": 6900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.272450191644863e-06,
+      "loss": 0.8384,
+      "step": 6901
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271927273467484e-06,
+      "loss": 0.8296,
+      "step": 6902
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271404292692983e-06,
+      "loss": 0.8193,
+      "step": 6903
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270881249331362e-06,
+      "loss": 0.9189,
+      "step": 6904
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27035814339263e-06,
+      "loss": 0.8979,
+      "step": 6905
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.26983497488679e-06,
+      "loss": 0.9087,
+      "step": 6906
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.269311743823855e-06,
+      "loss": 0.9302,
+      "step": 6907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268788450213836e-06,
+      "loss": 0.8984,
+      "step": 6908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268265094066744e-06,
+      "loss": 0.9297,
+      "step": 6909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267741675392589e-06,
+      "loss": 0.915,
+      "step": 6910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267218194201385e-06,
+      "loss": 0.8809,
+      "step": 6911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26669465050315e-06,
+      "loss": 0.9111,
+      "step": 6912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2661710443079e-06,
+      "loss": 0.9092,
+      "step": 6913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265647375625649e-06,
+      "loss": 0.9531,
+      "step": 6914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265123644466418e-06,
+      "loss": 0.9316,
+      "step": 6915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264599850840228e-06,
+      "loss": 0.9019,
+      "step": 6916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264075994757097e-06,
+      "loss": 0.877,
+      "step": 6917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263552076227048e-06,
+      "loss": 0.8965,
+      "step": 6918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263028095260105e-06,
+      "loss": 0.9336,
+      "step": 6919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.262504051866296e-06,
+      "loss": 0.8696,
+      "step": 6920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26197994605564e-06,
+      "loss": 0.9033,
+      "step": 6921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26145577783817e-06,
+      "loss": 0.9385,
+      "step": 6922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26093154722391e-06,
+      "loss": 0.8945,
+      "step": 6923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26040725422289e-06,
+      "loss": 0.8921,
+      "step": 6924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.259882898845144e-06,
+      "loss": 0.9673,
+      "step": 6925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.259358481100698e-06,
+      "loss": 0.959,
+      "step": 6926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25883400099959e-06,
+      "loss": 0.9673,
+      "step": 6927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258309458551853e-06,
+      "loss": 0.8418,
+      "step": 6928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25778485376752e-06,
+      "loss": 0.3107,
+      "step": 6929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.257260186656631e-06,
+      "loss": 0.9326,
+      "step": 6930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256735457229221e-06,
+      "loss": 0.9272,
+      "step": 6931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256210665495328e-06,
+      "loss": 0.9204,
+      "step": 6932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255685811464996e-06,
+      "loss": 0.9536,
+      "step": 6933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255160895148263e-06,
+      "loss": 0.8823,
+      "step": 6934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.254635916555173e-06,
+      "loss": 0.7886,
+      "step": 6935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.254110875695768e-06,
+      "loss": 0.8701,
+      "step": 6936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253585772580097e-06,
+      "loss": 0.8984,
+      "step": 6937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253060607218202e-06,
+      "loss": 0.9146,
+      "step": 6938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25253537962013e-06,
+      "loss": 0.8916,
+      "step": 6939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25201008979593e-06,
+      "loss": 0.9429,
+      "step": 6940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.251484737755655e-06,
+      "loss": 0.9175,
+      "step": 6941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250959323509352e-06,
+      "loss": 0.9438,
+      "step": 6942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250433847067076e-06,
+      "loss": 0.9644,
+      "step": 6943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249908308438876e-06,
+      "loss": 0.8394,
+      "step": 6944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249382707634809e-06,
+      "loss": 0.8638,
+      "step": 6945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248857044664932e-06,
+      "loss": 0.9258,
+      "step": 6946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2483313195393e-06,
+      "loss": 0.918,
+      "step": 6947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24780553226797e-06,
+      "loss": 0.9263,
+      "step": 6948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.247279682861005e-06,
+      "loss": 0.938,
+      "step": 6949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24675377132846e-06,
+      "loss": 0.9536,
+      "step": 6950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2462277976804e-06,
+      "loss": 0.3877,
+      "step": 6951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.245701761926886e-06,
+      "loss": 0.8657,
+      "step": 6952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.245175664077983e-06,
+      "loss": 0.8535,
+      "step": 6953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244649504143756e-06,
+      "loss": 0.8608,
+      "step": 6954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244123282134272e-06,
+      "loss": 0.8794,
+      "step": 6955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243596998059595e-06,
+      "loss": 0.8706,
+      "step": 6956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243070651929798e-06,
+      "loss": 0.9409,
+      "step": 6957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242544243754949e-06,
+      "loss": 0.8999,
+      "step": 6958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242017773545119e-06,
+      "loss": 0.9194,
+      "step": 6959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241491241310379e-06,
+      "loss": 0.9014,
+      "step": 6960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240964647060804e-06,
+      "loss": 0.8906,
+      "step": 6961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24043799080647e-06,
+      "loss": 0.9258,
+      "step": 6962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239911272557448e-06,
+      "loss": 0.8882,
+      "step": 6963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239384492323818e-06,
+      "loss": 0.9023,
+      "step": 6964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.238857650115662e-06,
+      "loss": 0.8975,
+      "step": 6965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.238330745943052e-06,
+      "loss": 0.8818,
+      "step": 6966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237803779816073e-06,
+      "loss": 0.9229,
+      "step": 6967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237276751744804e-06,
+      "loss": 0.9116,
+      "step": 6968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236749661739331e-06,
+      "loss": 0.8364,
+      "step": 6969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236222509809737e-06,
+      "loss": 0.9165,
+      "step": 6970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.235695295966105e-06,
+      "loss": 0.8638,
+      "step": 6971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.235168020218525e-06,
+      "loss": 0.8882,
+      "step": 6972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234640682577081e-06,
+      "loss": 0.874,
+      "step": 6973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234113283051865e-06,
+      "loss": 0.8916,
+      "step": 6974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.233585821652965e-06,
+      "loss": 0.741,
+      "step": 6975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.233058298390472e-06,
+      "loss": 1.0044,
+      "step": 6976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23253071327448e-06,
+      "loss": 0.8589,
+      "step": 6977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232003066315082e-06,
+      "loss": 0.9048,
+      "step": 6978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.231475357522372e-06,
+      "loss": 0.9092,
+      "step": 6979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.230947586906445e-06,
+      "loss": 0.8892,
+      "step": 6980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2304197544774e-06,
+      "loss": 0.8857,
+      "step": 6981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229891860245334e-06,
+      "loss": 0.9033,
+      "step": 6982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22936390422035e-06,
+      "loss": 0.8726,
+      "step": 6983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228835886412543e-06,
+      "loss": 1.0039,
+      "step": 6984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228307806832018e-06,
+      "loss": 0.875,
+      "step": 6985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.227779665488877e-06,
+      "loss": 0.9365,
+      "step": 6986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.227251462393225e-06,
+      "loss": 0.9116,
+      "step": 6987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226723197555168e-06,
+      "loss": 0.9111,
+      "step": 6988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226194870984809e-06,
+      "loss": 0.8677,
+      "step": 6989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22566648269226e-06,
+      "loss": 0.936,
+      "step": 6990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.225138032687629e-06,
+      "loss": 0.9453,
+      "step": 6991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.224609520981025e-06,
+      "loss": 0.8906,
+      "step": 6992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.224080947582558e-06,
+      "loss": 0.8823,
+      "step": 6993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.223552312502342e-06,
+      "loss": 0.3265,
+      "step": 6994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.223023615750495e-06,
+      "loss": 0.9185,
+      "step": 6995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222494857337122e-06,
+      "loss": 0.9497,
+      "step": 6996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221966037272347e-06,
+      "loss": 0.8516,
+      "step": 6997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221437155566283e-06,
+      "loss": 0.9399,
+      "step": 6998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220908212229051e-06,
+      "loss": 0.8838,
+      "step": 6999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220379207270769e-06,
+      "loss": 0.9004,
+      "step": 7000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.219850140701557e-06,
+      "loss": 0.9043,
+      "step": 7001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21932101253154e-06,
+      "loss": 0.2881,
+      "step": 7002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218791822770837e-06,
+      "loss": 0.8264,
+      "step": 7003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218262571429577e-06,
+      "loss": 0.9346,
+      "step": 7004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217733258517879e-06,
+      "loss": 0.9551,
+      "step": 7005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217203884045876e-06,
+      "loss": 0.8828,
+      "step": 7006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21667444802369e-06,
+      "loss": 0.8623,
+      "step": 7007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.216144950461458e-06,
+      "loss": 0.9185,
+      "step": 7008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2156153913693e-06,
+      "loss": 0.9395,
+      "step": 7009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215085770757357e-06,
+      "loss": 0.876,
+      "step": 7010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.214556088635754e-06,
+      "loss": 0.959,
+      "step": 7011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.214026345014628e-06,
+      "loss": 0.8491,
+      "step": 7012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.213496539904114e-06,
+      "loss": 0.8374,
+      "step": 7013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212966673314345e-06,
+      "loss": 0.356,
+      "step": 7014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212436745255462e-06,
+      "loss": 0.9092,
+      "step": 7015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211906755737602e-06,
+      "loss": 0.9116,
+      "step": 7016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211376704770906e-06,
+      "loss": 0.9316,
+      "step": 7017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21084659236551e-06,
+      "loss": 0.9229,
+      "step": 7018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210316418531561e-06,
+      "loss": 0.9248,
+      "step": 7019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2097861832792e-06,
+      "loss": 0.8691,
+      "step": 7020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209255886618573e-06,
+      "loss": 0.855,
+      "step": 7021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20872552855982e-06,
+      "loss": 0.9658,
+      "step": 7022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208195109113094e-06,
+      "loss": 0.9644,
+      "step": 7023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.207664628288538e-06,
+      "loss": 0.8887,
+      "step": 7024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.207134086096304e-06,
+      "loss": 0.3289,
+      "step": 7025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20660348254654e-06,
+      "loss": 0.9102,
+      "step": 7026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.206072817649402e-06,
+      "loss": 0.9106,
+      "step": 7027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205542091415036e-06,
+      "loss": 0.8765,
+      "step": 7028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205011303853599e-06,
+      "loss": 0.9688,
+      "step": 7029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.204480454975246e-06,
+      "loss": 0.9019,
+      "step": 7030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203949544790131e-06,
+      "loss": 0.9058,
+      "step": 7031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203418573308412e-06,
+      "loss": 0.8501,
+      "step": 7032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20288754054025e-06,
+      "loss": 0.8318,
+      "step": 7033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2023564464958e-06,
+      "loss": 0.8853,
+      "step": 7034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201825291185226e-06,
+      "loss": 0.8838,
+      "step": 7035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201294074618689e-06,
+      "loss": 0.9106,
+      "step": 7036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200762796806352e-06,
+      "loss": 0.9165,
+      "step": 7037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200231457758378e-06,
+      "loss": 0.9126,
+      "step": 7038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199700057484933e-06,
+      "loss": 0.8608,
+      "step": 7039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199168595996185e-06,
+      "loss": 0.8882,
+      "step": 7040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1986370733023e-06,
+      "loss": 0.8525,
+      "step": 7041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.198105489413445e-06,
+      "loss": 0.8223,
+      "step": 7042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.197573844339796e-06,
+      "loss": 0.9438,
+      "step": 7043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19704213809152e-06,
+      "loss": 0.9111,
+      "step": 7044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19651037067879e-06,
+      "loss": 0.8701,
+      "step": 7045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.195978542111778e-06,
+      "loss": 0.877,
+      "step": 7046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19544665240066e-06,
+      "loss": 0.9175,
+      "step": 7047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194914701555615e-06,
+      "loss": 0.895,
+      "step": 7048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194382689586815e-06,
+      "loss": 0.8638,
+      "step": 7049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19385061650444e-06,
+      "loss": 1.0181,
+      "step": 7050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19331848231867e-06,
+      "loss": 0.896,
+      "step": 7051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.192786287039687e-06,
+      "loss": 0.8945,
+      "step": 7052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.192254030677669e-06,
+      "loss": 0.8516,
+      "step": 7053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191721713242803e-06,
+      "loss": 0.9019,
+      "step": 7054
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19118933474527e-06,
+      "loss": 0.9399,
+      "step": 7055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.190656895195257e-06,
+      "loss": 0.8857,
+      "step": 7056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19012439460295e-06,
+      "loss": 0.8887,
+      "step": 7057
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.189591832978533e-06,
+      "loss": 0.3051,
+      "step": 7058
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.189059210332201e-06,
+      "loss": 0.8916,
+      "step": 7059
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18852652667414e-06,
+      "loss": 0.9282,
+      "step": 7060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.187993782014545e-06,
+      "loss": 0.8906,
+      "step": 7061
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.187460976363602e-06,
+      "loss": 0.8784,
+      "step": 7062
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186928109731508e-06,
+      "loss": 0.8984,
+      "step": 7063
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18639518212846e-06,
+      "loss": 0.8379,
+      "step": 7064
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185862193564648e-06,
+      "loss": 0.9248,
+      "step": 7065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185329144050274e-06,
+      "loss": 0.8921,
+      "step": 7066
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.184796033595533e-06,
+      "loss": 0.8862,
+      "step": 7067
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.184262862210624e-06,
+      "loss": 0.8521,
+      "step": 7068
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18372962990575e-06,
+      "loss": 0.9604,
+      "step": 7069
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183196336691111e-06,
+      "loss": 0.9473,
+      "step": 7070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18266298257691e-06,
+      "loss": 0.8501,
+      "step": 7071
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.182129567573351e-06,
+      "loss": 0.8438,
+      "step": 7072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18159609169064e-06,
+      "loss": 0.8623,
+      "step": 7073
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.181062554938982e-06,
+      "loss": 0.8706,
+      "step": 7074
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.180528957328582e-06,
+      "loss": 0.9395,
+      "step": 7075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179995298869654e-06,
+      "loss": 0.9004,
+      "step": 7076
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179461579572403e-06,
+      "loss": 0.8892,
+      "step": 7077
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178927799447042e-06,
+      "loss": 0.8701,
+      "step": 7078
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178393958503785e-06,
+      "loss": 0.9341,
+      "step": 7079
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177860056752841e-06,
+      "loss": 0.8799,
+      "step": 7080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177326094204428e-06,
+      "loss": 0.9092,
+      "step": 7081
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176792070868757e-06,
+      "loss": 0.8643,
+      "step": 7082
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.17625798675605e-06,
+      "loss": 0.9209,
+      "step": 7083
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175723841876522e-06,
+      "loss": 0.8984,
+      "step": 7084
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175189636240393e-06,
+      "loss": 0.9453,
+      "step": 7085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.174655369857882e-06,
+      "loss": 0.8755,
+      "step": 7086
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.174121042739213e-06,
+      "loss": 0.9033,
+      "step": 7087
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173586654894605e-06,
+      "loss": 0.9492,
+      "step": 7088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173052206334283e-06,
+      "loss": 0.8789,
+      "step": 7089
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.172517697068474e-06,
+      "loss": 0.9897,
+      "step": 7090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1719831271074e-06,
+      "loss": 0.9038,
+      "step": 7091
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.171448496461292e-06,
+      "loss": 0.856,
+      "step": 7092
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170913805140375e-06,
+      "loss": 0.8311,
+      "step": 7093
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170379053154882e-06,
+      "loss": 0.9189,
+      "step": 7094
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16984424051504e-06,
+      "loss": 0.8706,
+      "step": 7095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169309367231085e-06,
+      "loss": 0.9053,
+      "step": 7096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168774433313244e-06,
+      "loss": 0.9189,
+      "step": 7097
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16823943877176e-06,
+      "loss": 0.9404,
+      "step": 7098
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167704383616859e-06,
+      "loss": 0.8809,
+      "step": 7099
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167169267858781e-06,
+      "loss": 0.8228,
+      "step": 7100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.166634091507767e-06,
+      "loss": 0.8457,
+      "step": 7101
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.166098854574052e-06,
+      "loss": 0.8228,
+      "step": 7102
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.165563557067874e-06,
+      "loss": 0.856,
+      "step": 7103
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16502819899948e-06,
+      "loss": 0.8818,
+      "step": 7104
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164492780379107e-06,
+      "loss": 0.9478,
+      "step": 7105
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163957301217002e-06,
+      "loss": 0.8794,
+      "step": 7106
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163421761523407e-06,
+      "loss": 0.8516,
+      "step": 7107
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16288616130857e-06,
+      "loss": 0.9185,
+      "step": 7108
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.162350500582736e-06,
+      "loss": 0.9336,
+      "step": 7109
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161814779356154e-06,
+      "loss": 0.8848,
+      "step": 7110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161278997639072e-06,
+      "loss": 0.8809,
+      "step": 7111
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160743155441743e-06,
+      "loss": 0.9062,
+      "step": 7112
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160207252774414e-06,
+      "loss": 0.8052,
+      "step": 7113
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.159671289647343e-06,
+      "loss": 0.937,
+      "step": 7114
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15913526607078e-06,
+      "loss": 0.9062,
+      "step": 7115
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15859918205498e-06,
+      "loss": 0.9419,
+      "step": 7116
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.158063037610202e-06,
+      "loss": 0.9395,
+      "step": 7117
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157526832746702e-06,
+      "loss": 0.9009,
+      "step": 7118
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156990567474737e-06,
+      "loss": 0.5459,
+      "step": 7119
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156454241804567e-06,
+      "loss": 0.8618,
+      "step": 7120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155917855746456e-06,
+      "loss": 0.9395,
+      "step": 7121
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155381409310662e-06,
+      "loss": 0.8076,
+      "step": 7122
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154844902507448e-06,
+      "loss": 0.8521,
+      "step": 7123
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15430833534708e-06,
+      "loss": 0.8853,
+      "step": 7124
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153771707839824e-06,
+      "loss": 0.8955,
+      "step": 7125
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153235019995945e-06,
+      "loss": 0.9219,
+      "step": 7126
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152698271825712e-06,
+      "loss": 0.9331,
+      "step": 7127
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152161463339392e-06,
+      "loss": 0.9111,
+      "step": 7128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151624594547256e-06,
+      "loss": 0.8613,
+      "step": 7129
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151087665459575e-06,
+      "loss": 0.9038,
+      "step": 7130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15055067608662e-06,
+      "loss": 0.8892,
+      "step": 7131
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.150013626438672e-06,
+      "loss": 0.9199,
+      "step": 7132
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149476516525993e-06,
+      "loss": 0.8491,
+      "step": 7133
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.148939346358868e-06,
+      "loss": 0.9009,
+      "step": 7134
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14840211594757e-06,
+      "loss": 0.9033,
+      "step": 7135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14786482530238e-06,
+      "loss": 0.7986,
+      "step": 7136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.147327474433574e-06,
+      "loss": 0.8975,
+      "step": 7137
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.146790063351435e-06,
+      "loss": 0.9141,
+      "step": 7138
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.146252592066242e-06,
+      "loss": 0.8862,
+      "step": 7139
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14571506058828e-06,
+      "loss": 0.8706,
+      "step": 7140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145177468927831e-06,
+      "loss": 0.9614,
+      "step": 7141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.144639817095181e-06,
+      "loss": 0.8857,
+      "step": 7142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.144102105100618e-06,
+      "loss": 0.8784,
+      "step": 7143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143564332954426e-06,
+      "loss": 0.9463,
+      "step": 7144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143026500666893e-06,
+      "loss": 0.9536,
+      "step": 7145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.14248860824831e-06,
+      "loss": 0.8862,
+      "step": 7146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141950655708973e-06,
+      "loss": 0.9141,
+      "step": 7147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141412643059163e-06,
+      "loss": 0.98,
+      "step": 7148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140874570309183e-06,
+      "loss": 0.9385,
+      "step": 7149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140336437469321e-06,
+      "loss": 0.856,
+      "step": 7150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139798244549875e-06,
+      "loss": 0.8662,
+      "step": 7151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13925999156114e-06,
+      "loss": 0.9321,
+      "step": 7152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138721678513414e-06,
+      "loss": 0.8911,
+      "step": 7153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138183305416997e-06,
+      "loss": 0.9072,
+      "step": 7154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137644872282192e-06,
+      "loss": 0.8955,
+      "step": 7155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137106379119291e-06,
+      "loss": 0.7568,
+      "step": 7156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136567825938604e-06,
+      "loss": 0.8179,
+      "step": 7157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136029212750432e-06,
+      "loss": 0.9028,
+      "step": 7158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135490539565078e-06,
+      "loss": 0.8638,
+      "step": 7159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13495180639285e-06,
+      "loss": 0.8989,
+      "step": 7160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134413013244053e-06,
+      "loss": 0.8643,
+      "step": 7161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133874160128996e-06,
+      "loss": 0.8608,
+      "step": 7162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133335247057989e-06,
+      "loss": 0.8994,
+      "step": 7163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132796274041342e-06,
+      "loss": 0.9082,
+      "step": 7164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132257241089362e-06,
+      "loss": 0.8452,
+      "step": 7165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131718148212368e-06,
+      "loss": 0.9277,
+      "step": 7166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131178995420669e-06,
+      "loss": 0.9312,
+      "step": 7167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130639782724584e-06,
+      "loss": 0.9165,
+      "step": 7168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130100510134427e-06,
+      "loss": 0.9248,
+      "step": 7169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129561177660511e-06,
+      "loss": 0.8564,
+      "step": 7170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129021785313162e-06,
+      "loss": 0.9697,
+      "step": 7171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128482333102694e-06,
+      "loss": 0.8872,
+      "step": 7172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12794282103943e-06,
+      "loss": 0.4127,
+      "step": 7173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127403249133689e-06,
+      "loss": 0.9458,
+      "step": 7174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.126863617395795e-06,
+      "loss": 0.937,
+      "step": 7175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.126323925836076e-06,
+      "loss": 0.8628,
+      "step": 7176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125784174464852e-06,
+      "loss": 0.8442,
+      "step": 7177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12524436329245e-06,
+      "loss": 0.7993,
+      "step": 7178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.124704492329201e-06,
+      "loss": 0.8726,
+      "step": 7179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12416456158543e-06,
+      "loss": 0.8965,
+      "step": 7180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123624571071468e-06,
+      "loss": 0.8975,
+      "step": 7181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123084520797645e-06,
+      "loss": 0.8545,
+      "step": 7182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122544410774294e-06,
+      "loss": 0.894,
+      "step": 7183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122004241011748e-06,
+      "loss": 0.8799,
+      "step": 7184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.121464011520341e-06,
+      "loss": 0.8823,
+      "step": 7185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12092372231041e-06,
+      "loss": 0.9609,
+      "step": 7186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120383373392289e-06,
+      "loss": 0.9609,
+      "step": 7187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.119842964776317e-06,
+      "loss": 0.8359,
+      "step": 7188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.119302496472833e-06,
+      "loss": 0.917,
+      "step": 7189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.118761968492177e-06,
+      "loss": 0.9067,
+      "step": 7190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.118221380844689e-06,
+      "loss": 0.8818,
+      "step": 7191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117680733540714e-06,
+      "loss": 0.8203,
+      "step": 7192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117140026590594e-06,
+      "loss": 0.8721,
+      "step": 7193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116599260004671e-06,
+      "loss": 0.9497,
+      "step": 7194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116058433793294e-06,
+      "loss": 0.9292,
+      "step": 7195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115517547966808e-06,
+      "loss": 0.8647,
+      "step": 7196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.114976602535564e-06,
+      "loss": 0.9268,
+      "step": 7197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.114435597509908e-06,
+      "loss": 0.8848,
+      "step": 7198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11389453290019e-06,
+      "loss": 0.8643,
+      "step": 7199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113353408716764e-06,
+      "loss": 0.8765,
+      "step": 7200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112812224969981e-06,
+      "loss": 0.9658,
+      "step": 7201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112270981670196e-06,
+      "loss": 0.9253,
+      "step": 7202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11172967882776e-06,
+      "loss": 0.8813,
+      "step": 7203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.111188316453035e-06,
+      "loss": 0.9067,
+      "step": 7204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110646894556374e-06,
+      "loss": 0.8594,
+      "step": 7205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110105413148136e-06,
+      "loss": 0.8701,
+      "step": 7206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109563872238679e-06,
+      "loss": 0.9302,
+      "step": 7207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109022271838367e-06,
+      "loss": 0.4397,
+      "step": 7208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10848061195756e-06,
+      "loss": 0.8853,
+      "step": 7209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.107938892606618e-06,
+      "loss": 0.9185,
+      "step": 7210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.107397113795908e-06,
+      "loss": 0.917,
+      "step": 7211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106855275535797e-06,
+      "loss": 0.9185,
+      "step": 7212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106313377836647e-06,
+      "loss": 0.897,
+      "step": 7213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105771420708827e-06,
+      "loss": 0.9604,
+      "step": 7214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105229404162707e-06,
+      "loss": 0.7935,
+      "step": 7215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.104687328208653e-06,
+      "loss": 0.9033,
+      "step": 7216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10414519285704e-06,
+      "loss": 0.8403,
+      "step": 7217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103602998118238e-06,
+      "loss": 0.8315,
+      "step": 7218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103060744002618e-06,
+      "loss": 0.9443,
+      "step": 7219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.102518430520559e-06,
+      "loss": 0.9175,
+      "step": 7220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10197605768243e-06,
+      "loss": 0.8853,
+      "step": 7221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101433625498614e-06,
+      "loss": 0.9355,
+      "step": 7222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.100891133979483e-06,
+      "loss": 0.9375,
+      "step": 7223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10034858313542e-06,
+      "loss": 0.9263,
+      "step": 7224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099805972976803e-06,
+      "loss": 0.8569,
+      "step": 7225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099263303514014e-06,
+      "loss": 0.9253,
+      "step": 7226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09872057475743e-06,
+      "loss": 0.9326,
+      "step": 7227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098177786717442e-06,
+      "loss": 0.9312,
+      "step": 7228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.097634939404431e-06,
+      "loss": 0.9307,
+      "step": 7229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.097092032828782e-06,
+      "loss": 0.9751,
+      "step": 7230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09654906700088e-06,
+      "loss": 0.875,
+      "step": 7231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.096006041931117e-06,
+      "loss": 0.8706,
+      "step": 7232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095462957629878e-06,
+      "loss": 0.9023,
+      "step": 7233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.094919814107557e-06,
+      "loss": 0.8638,
+      "step": 7234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.094376611374542e-06,
+      "loss": 0.9268,
+      "step": 7235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093833349441225e-06,
+      "loss": 0.8818,
+      "step": 7236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093290028318002e-06,
+      "loss": 0.96,
+      "step": 7237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.092746648015265e-06,
+      "loss": 0.9355,
+      "step": 7238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.092203208543411e-06,
+      "loss": 0.9058,
+      "step": 7239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091659709912838e-06,
+      "loss": 0.9106,
+      "step": 7240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091116152133944e-06,
+      "loss": 0.8828,
+      "step": 7241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090572535217128e-06,
+      "loss": 0.8735,
+      "step": 7242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090028859172786e-06,
+      "loss": 0.8779,
+      "step": 7243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089485124011322e-06,
+      "loss": 0.8784,
+      "step": 7244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.088941329743144e-06,
+      "loss": 0.9155,
+      "step": 7245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.088397476378647e-06,
+      "loss": 0.8794,
+      "step": 7246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08785356392824e-06,
+      "loss": 0.8755,
+      "step": 7247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.087309592402328e-06,
+      "loss": 0.8804,
+      "step": 7248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086765561811319e-06,
+      "loss": 0.8721,
+      "step": 7249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086221472165623e-06,
+      "loss": 0.854,
+      "step": 7250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085677323475643e-06,
+      "loss": 0.8774,
+      "step": 7251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085133115751796e-06,
+      "loss": 0.8418,
+      "step": 7252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08458884900449e-06,
+      "loss": 0.9575,
+      "step": 7253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.084044523244138e-06,
+      "loss": 0.8901,
+      "step": 7254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.083500138481155e-06,
+      "loss": 0.9272,
+      "step": 7255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082955694725956e-06,
+      "loss": 0.8921,
+      "step": 7256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082411191988956e-06,
+      "loss": 0.9287,
+      "step": 7257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081866630280574e-06,
+      "loss": 0.8955,
+      "step": 7258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081322009611225e-06,
+      "loss": 0.9712,
+      "step": 7259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080777329991332e-06,
+      "loss": 0.9004,
+      "step": 7260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080232591431311e-06,
+      "loss": 0.9116,
+      "step": 7261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079687793941589e-06,
+      "loss": 0.9331,
+      "step": 7262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079142937532587e-06,
+      "loss": 0.9277,
+      "step": 7263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.078598022214727e-06,
+      "loss": 0.9502,
+      "step": 7264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.078053047998436e-06,
+      "loss": 0.8916,
+      "step": 7265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.077508014894141e-06,
+      "loss": 0.9473,
+      "step": 7266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076962922912266e-06,
+      "loss": 0.8936,
+      "step": 7267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076417772063245e-06,
+      "loss": 0.9004,
+      "step": 7268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075872562357502e-06,
+      "loss": 0.8774,
+      "step": 7269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075327293805468e-06,
+      "loss": 0.8994,
+      "step": 7270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07478196641758e-06,
+      "loss": 0.9126,
+      "step": 7271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074236580204267e-06,
+      "loss": 0.8359,
+      "step": 7272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.073691135175961e-06,
+      "loss": 0.8501,
+      "step": 7273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.073145631343102e-06,
+      "loss": 0.9189,
+      "step": 7274
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072600068716125e-06,
+      "loss": 0.959,
+      "step": 7275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072054447305468e-06,
+      "loss": 0.936,
+      "step": 7276
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.071508767121566e-06,
+      "loss": 0.9238,
+      "step": 7277
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.070963028174862e-06,
+      "loss": 0.9219,
+      "step": 7278
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.070417230475796e-06,
+      "loss": 0.9116,
+      "step": 7279
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06987137403481e-06,
+      "loss": 0.9011,
+      "step": 7280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.069325458862348e-06,
+      "loss": 0.8755,
+      "step": 7281
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.068779484968852e-06,
+      "loss": 0.8628,
+      "step": 7282
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06823345236477e-06,
+      "loss": 0.8506,
+      "step": 7283
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067687361060547e-06,
+      "loss": 0.9048,
+      "step": 7284
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06714121106663e-06,
+      "loss": 0.8286,
+      "step": 7285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066595002393467e-06,
+      "loss": 0.3365,
+      "step": 7286
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066048735051512e-06,
+      "loss": 0.8647,
+      "step": 7287
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06550240905121e-06,
+      "loss": 0.9146,
+      "step": 7288
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06495602440302e-06,
+      "loss": 0.9077,
+      "step": 7289
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.064409581117389e-06,
+      "loss": 0.3558,
+      "step": 7290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063863079204775e-06,
+      "loss": 0.2841,
+      "step": 7291
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063316518675633e-06,
+      "loss": 0.8867,
+      "step": 7292
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062769899540417e-06,
+      "loss": 0.9131,
+      "step": 7293
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062223221809586e-06,
+      "loss": 0.8633,
+      "step": 7294
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061676485493602e-06,
+      "loss": 0.9131,
+      "step": 7295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061129690602919e-06,
+      "loss": 0.9058,
+      "step": 7296
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060582837148003e-06,
+      "loss": 0.8853,
+      "step": 7297
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060035925139314e-06,
+      "loss": 0.9072,
+      "step": 7298
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059488954587316e-06,
+      "loss": 0.3549,
+      "step": 7299
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058941925502475e-06,
+      "loss": 0.8784,
+      "step": 7300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058394837895252e-06,
+      "loss": 0.897,
+      "step": 7301
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057847691776117e-06,
+      "loss": 0.9077,
+      "step": 7302
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057300487155538e-06,
+      "loss": 0.9131,
+      "step": 7303
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.056753224043982e-06,
+      "loss": 0.9346,
+      "step": 7304
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05620590245192e-06,
+      "loss": 0.918,
+      "step": 7305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055658522389821e-06,
+      "loss": 0.8247,
+      "step": 7306
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055111083868163e-06,
+      "loss": 0.9136,
+      "step": 7307
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054563586897414e-06,
+      "loss": 0.8833,
+      "step": 7308
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054016031488052e-06,
+      "loss": 0.8433,
+      "step": 7309
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05346841765055e-06,
+      "loss": 0.8804,
+      "step": 7310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.052920745395382e-06,
+      "loss": 0.8833,
+      "step": 7311
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.052373014733034e-06,
+      "loss": 0.8911,
+      "step": 7312
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05182522567398e-06,
+      "loss": 0.9214,
+      "step": 7313
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.051277378228697e-06,
+      "loss": 0.8721,
+      "step": 7314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050729472407673e-06,
+      "loss": 0.9434,
+      "step": 7315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050181508221384e-06,
+      "loss": 0.8872,
+      "step": 7316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049633485680317e-06,
+      "loss": 0.9331,
+      "step": 7317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049085404794958e-06,
+      "loss": 0.3225,
+      "step": 7318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048537265575787e-06,
+      "loss": 0.9023,
+      "step": 7319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.047989068033296e-06,
+      "loss": 0.9302,
+      "step": 7320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.047440812177971e-06,
+      "loss": 0.8882,
+      "step": 7321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.046892498020303e-06,
+      "loss": 0.8936,
+      "step": 7322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.046344125570778e-06,
+      "loss": 0.9111,
+      "step": 7323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.045795694839888e-06,
+      "loss": 0.8574,
+      "step": 7324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04524720583813e-06,
+      "loss": 0.8823,
+      "step": 7325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044698658575993e-06,
+      "loss": 0.9448,
+      "step": 7326
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044150053063971e-06,
+      "loss": 0.8965,
+      "step": 7327
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043601389312565e-06,
+      "loss": 0.9043,
+      "step": 7328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043052667332264e-06,
+      "loss": 0.8623,
+      "step": 7329
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042503887133575e-06,
+      "loss": 0.8516,
+      "step": 7330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.041955048726989e-06,
+      "loss": 0.9014,
+      "step": 7331
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04140615212301e-06,
+      "loss": 0.8823,
+      "step": 7332
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04085719733214e-06,
+      "loss": 0.897,
+      "step": 7333
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.040308184364878e-06,
+      "loss": 0.9248,
+      "step": 7334
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03975911323173e-06,
+      "loss": 0.9272,
+      "step": 7335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.039209983943201e-06,
+      "loss": 0.9292,
+      "step": 7336
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038660796509796e-06,
+      "loss": 0.9419,
+      "step": 7337
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03811155094202e-06,
+      "loss": 0.9189,
+      "step": 7338
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037562247250383e-06,
+      "loss": 0.9292,
+      "step": 7339
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037012885445394e-06,
+      "loss": 0.9033,
+      "step": 7340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036463465537565e-06,
+      "loss": 0.8828,
+      "step": 7341
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035913987537403e-06,
+      "loss": 0.9331,
+      "step": 7342
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035364451455421e-06,
+      "loss": 0.8994,
+      "step": 7343
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034814857302136e-06,
+      "loss": 0.8857,
+      "step": 7344
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034265205088059e-06,
+      "loss": 0.8989,
+      "step": 7345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.033715494823709e-06,
+      "loss": 0.9048,
+      "step": 7346
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0331657265196e-06,
+      "loss": 0.9077,
+      "step": 7347
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03261590018625e-06,
+      "loss": 0.8667,
+      "step": 7348
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03206601583418e-06,
+      "loss": 0.938,
+      "step": 7349
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.031516073473911e-06,
+      "loss": 0.8828,
+      "step": 7350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03096607311596e-06,
+      "loss": 0.895,
+      "step": 7351
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.030416014770853e-06,
+      "loss": 0.895,
+      "step": 7352
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.029865898449111e-06,
+      "loss": 0.9351,
+      "step": 7353
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.029315724161261e-06,
+      "loss": 0.3376,
+      "step": 7354
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028765491917828e-06,
+      "loss": 0.8608,
+      "step": 7355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028215201729336e-06,
+      "loss": 0.895,
+      "step": 7356
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027664853606317e-06,
+      "loss": 0.9014,
+      "step": 7357
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027114447559299e-06,
+      "loss": 0.8777,
+      "step": 7358
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.026563983598811e-06,
+      "loss": 0.8975,
+      "step": 7359
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.026013461735384e-06,
+      "loss": 0.8687,
+      "step": 7360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02546288197955e-06,
+      "loss": 0.9014,
+      "step": 7361
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024912244341847e-06,
+      "loss": 0.9272,
+      "step": 7362
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024361548832805e-06,
+      "loss": 0.918,
+      "step": 7363
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023810795462959e-06,
+      "loss": 0.8657,
+      "step": 7364
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023259984242848e-06,
+      "loss": 0.9282,
+      "step": 7365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02270911518301e-06,
+      "loss": 0.9058,
+      "step": 7366
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022158188293984e-06,
+      "loss": 0.8926,
+      "step": 7367
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021607203586307e-06,
+      "loss": 0.8481,
+      "step": 7368
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021056161070526e-06,
+      "loss": 0.9292,
+      "step": 7369
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02050506075718e-06,
+      "loss": 0.8789,
+      "step": 7370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019953902656809e-06,
+      "loss": 0.9131,
+      "step": 7371
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019402686779965e-06,
+      "loss": 0.9155,
+      "step": 7372
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018851413137186e-06,
+      "loss": 0.8413,
+      "step": 7373
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018300081739026e-06,
+      "loss": 0.9424,
+      "step": 7374
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.017748692596026e-06,
+      "loss": 0.916,
+      "step": 7375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.017197245718739e-06,
+      "loss": 0.8726,
+      "step": 7376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016645741117715e-06,
+      "loss": 0.8838,
+      "step": 7377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016094178803505e-06,
+      "loss": 0.9248,
+      "step": 7378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01554255878666e-06,
+      "loss": 0.9321,
+      "step": 7379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014990881077732e-06,
+      "loss": 0.8584,
+      "step": 7380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014439145687279e-06,
+      "loss": 0.9014,
+      "step": 7381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013887352625856e-06,
+      "loss": 0.9136,
+      "step": 7382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013335501904015e-06,
+      "loss": 0.9316,
+      "step": 7383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.012783593532321e-06,
+      "loss": 0.9458,
+      "step": 7384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.012231627521329e-06,
+      "loss": 0.835,
+      "step": 7385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.011679603881598e-06,
+      "loss": 0.9199,
+      "step": 7386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.011127522623691e-06,
+      "loss": 0.8867,
+      "step": 7387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010575383758171e-06,
+      "loss": 0.8687,
+      "step": 7388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010023187295597e-06,
+      "loss": 0.9023,
+      "step": 7389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.009470933246537e-06,
+      "loss": 0.8784,
+      "step": 7390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.008918621621557e-06,
+      "loss": 0.9219,
+      "step": 7391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00836625243122e-06,
+      "loss": 0.2889,
+      "step": 7392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007813825686098e-06,
+      "loss": 0.9282,
+      "step": 7393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007261341396757e-06,
+      "loss": 0.9492,
+      "step": 7394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006708799573768e-06,
+      "loss": 0.8774,
+      "step": 7395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006156200227702e-06,
+      "loss": 0.9775,
+      "step": 7396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00560354336913e-06,
+      "loss": 0.9141,
+      "step": 7397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.005050829008627e-06,
+      "loss": 0.897,
+      "step": 7398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.004498057156766e-06,
+      "loss": 0.9419,
+      "step": 7399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.003945227824122e-06,
+      "loss": 0.8862,
+      "step": 7400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.003392341021272e-06,
+      "loss": 0.9185,
+      "step": 7401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002839396758795e-06,
+      "loss": 0.894,
+      "step": 7402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002286395047267e-06,
+      "loss": 0.8804,
+      "step": 7403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00173333589727e-06,
+      "loss": 0.9375,
+      "step": 7404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001180219319384e-06,
+      "loss": 0.8965,
+      "step": 7405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.000627045324192e-06,
+      "loss": 0.9502,
+      "step": 7406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.000073813922275e-06,
+      "loss": 0.8848,
+      "step": 7407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999520525124218e-06,
+      "loss": 0.915,
+      "step": 7408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.998967178940608e-06,
+      "loss": 0.9531,
+      "step": 7409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99841377538203e-06,
+      "loss": 0.4031,
+      "step": 7410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.997860314459071e-06,
+      "loss": 0.8877,
+      "step": 7411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99730679618232e-06,
+      "loss": 0.8765,
+      "step": 7412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996753220562368e-06,
+      "loss": 0.8721,
+      "step": 7413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996199587609804e-06,
+      "loss": 0.8882,
+      "step": 7414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995645897335218e-06,
+      "loss": 0.9341,
+      "step": 7415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995092149749208e-06,
+      "loss": 0.9434,
+      "step": 7416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.994538344862366e-06,
+      "loss": 0.8604,
+      "step": 7417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993984482685286e-06,
+      "loss": 0.8696,
+      "step": 7418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993430563228564e-06,
+      "loss": 0.9463,
+      "step": 7419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.992876586502801e-06,
+      "loss": 0.9009,
+      "step": 7420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99232255251859e-06,
+      "loss": 0.8521,
+      "step": 7421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991768461286533e-06,
+      "loss": 0.9082,
+      "step": 7422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991214312817232e-06,
+      "loss": 0.9082,
+      "step": 7423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990660107121286e-06,
+      "loss": 0.9341,
+      "step": 7424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990105844209302e-06,
+      "loss": 0.9048,
+      "step": 7425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.989551524091882e-06,
+      "loss": 0.8789,
+      "step": 7426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988997146779627e-06,
+      "loss": 0.9316,
+      "step": 7427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988442712283149e-06,
+      "loss": 0.399,
+      "step": 7428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987888220613052e-06,
+      "loss": 0.939,
+      "step": 7429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987333671779944e-06,
+      "loss": 0.829,
+      "step": 7430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.986779065794437e-06,
+      "loss": 0.9341,
+      "step": 7431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.98622440266714e-06,
+      "loss": 0.9243,
+      "step": 7432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985669682408662e-06,
+      "loss": 0.8809,
+      "step": 7433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985114905029621e-06,
+      "loss": 0.9414,
+      "step": 7434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984560070540626e-06,
+      "loss": 0.337,
+      "step": 7435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984005178952294e-06,
+      "loss": 0.8711,
+      "step": 7436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.983450230275243e-06,
+      "loss": 0.3556,
+      "step": 7437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982895224520085e-06,
+      "loss": 0.8467,
+      "step": 7438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982340161697442e-06,
+      "loss": 0.9102,
+      "step": 7439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981785041817931e-06,
+      "loss": 0.8916,
+      "step": 7440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981229864892175e-06,
+      "loss": 0.8882,
+      "step": 7441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980674630930794e-06,
+      "loss": 0.8711,
+      "step": 7442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980119339944409e-06,
+      "loss": 0.917,
+      "step": 7443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.979563991943647e-06,
+      "loss": 0.8677,
+      "step": 7444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.979008586939127e-06,
+      "loss": 0.8335,
+      "step": 7445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978453124941481e-06,
+      "loss": 0.8818,
+      "step": 7446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.977897605961333e-06,
+      "loss": 0.8809,
+      "step": 7447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.977342030009312e-06,
+      "loss": 0.9316,
+      "step": 7448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976786397096045e-06,
+      "loss": 0.918,
+      "step": 7449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976230707232164e-06,
+      "loss": 0.9458,
+      "step": 7450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975674960428298e-06,
+      "loss": 0.3617,
+      "step": 7451
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975119156695084e-06,
+      "loss": 0.8809,
+      "step": 7452
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97456329604315e-06,
+      "loss": 0.9497,
+      "step": 7453
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.974007378483135e-06,
+      "loss": 0.8877,
+      "step": 7454
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97345140402567e-06,
+      "loss": 0.9351,
+      "step": 7455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972895372681394e-06,
+      "loss": 0.8467,
+      "step": 7456
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972339284460947e-06,
+      "loss": 0.8408,
+      "step": 7457
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971783139374964e-06,
+      "loss": 0.8979,
+      "step": 7458
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971226937434087e-06,
+      "loss": 0.9302,
+      "step": 7459
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.970670678648956e-06,
+      "loss": 0.9517,
+      "step": 7460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.970114363030215e-06,
+      "loss": 0.8428,
+      "step": 7461
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969557990588506e-06,
+      "loss": 0.8931,
+      "step": 7462
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96900156133447e-06,
+      "loss": 0.8911,
+      "step": 7463
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.968445075278758e-06,
+      "loss": 0.8804,
+      "step": 7464
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967888532432015e-06,
+      "loss": 0.9507,
+      "step": 7465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967331932804886e-06,
+      "loss": 0.8906,
+      "step": 7466
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966775276408023e-06,
+      "loss": 0.8887,
+      "step": 7467
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966218563252073e-06,
+      "loss": 0.8169,
+      "step": 7468
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965661793347686e-06,
+      "loss": 0.918,
+      "step": 7469
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965104966705518e-06,
+      "loss": 0.9185,
+      "step": 7470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964548083336218e-06,
+      "loss": 0.9194,
+      "step": 7471
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963991143250443e-06,
+      "loss": 0.9067,
+      "step": 7472
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963434146458844e-06,
+      "loss": 0.9487,
+      "step": 7473
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962877092972083e-06,
+      "loss": 0.7939,
+      "step": 7474
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962319982800813e-06,
+      "loss": 0.8467,
+      "step": 7475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961762815955693e-06,
+      "loss": 0.9082,
+      "step": 7476
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961205592447386e-06,
+      "loss": 0.9082,
+      "step": 7477
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960648312286547e-06,
+      "loss": 0.8052,
+      "step": 7478
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960090975483842e-06,
+      "loss": 0.8672,
+      "step": 7479
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959533582049929e-06,
+      "loss": 0.936,
+      "step": 7480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.958976131995479e-06,
+      "loss": 0.8867,
+      "step": 7481
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.958418625331151e-06,
+      "loss": 0.9478,
+      "step": 7482
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.957861062067614e-06,
+      "loss": 0.8247,
+      "step": 7483
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.957303442215531e-06,
+      "loss": 0.8472,
+      "step": 7484
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956745765785575e-06,
+      "loss": 0.8921,
+      "step": 7485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956188032788414e-06,
+      "loss": 0.8564,
+      "step": 7486
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955630243234716e-06,
+      "loss": 0.8267,
+      "step": 7487
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955072397135154e-06,
+      "loss": 0.9082,
+      "step": 7488
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.954514494500401e-06,
+      "loss": 0.8452,
+      "step": 7489
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95395653534113e-06,
+      "loss": 0.937,
+      "step": 7490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.953398519668017e-06,
+      "loss": 0.9072,
+      "step": 7491
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.952840447491735e-06,
+      "loss": 0.8667,
+      "step": 7492
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.952282318822962e-06,
+      "loss": 0.8555,
+      "step": 7493
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951724133672375e-06,
+      "loss": 0.8691,
+      "step": 7494
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951165892050655e-06,
+      "loss": 0.9229,
+      "step": 7495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95060759396848e-06,
+      "loss": 0.8857,
+      "step": 7496
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.950049239436535e-06,
+      "loss": 0.9448,
+      "step": 7497
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.949490828465497e-06,
+      "loss": 0.873,
+      "step": 7498
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948932361066053e-06,
+      "loss": 0.8501,
+      "step": 7499
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948373837248884e-06,
+      "loss": 0.3562,
+      "step": 7500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.947815257024678e-06,
+      "loss": 0.9951,
+      "step": 7501
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.947256620404123e-06,
+      "loss": 0.9062,
+      "step": 7502
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946697927397903e-06,
+      "loss": 0.8765,
+      "step": 7503
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946139178016708e-06,
+      "loss": 0.8608,
+      "step": 7504
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945580372271229e-06,
+      "loss": 0.9209,
+      "step": 7505
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945021510172156e-06,
+      "loss": 0.9321,
+      "step": 7506
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944462591730179e-06,
+      "loss": 0.9131,
+      "step": 7507
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943903616955992e-06,
+      "loss": 0.9058,
+      "step": 7508
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943344585860293e-06,
+      "loss": 0.9111,
+      "step": 7509
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942785498453772e-06,
+      "loss": 0.9277,
+      "step": 7510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942226354747128e-06,
+      "loss": 0.8647,
+      "step": 7511
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941667154751056e-06,
+      "loss": 0.897,
+      "step": 7512
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941107898476257e-06,
+      "loss": 0.9385,
+      "step": 7513
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.94054858593343e-06,
+      "loss": 0.9292,
+      "step": 7514
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.939989217133274e-06,
+      "loss": 0.9092,
+      "step": 7515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93942979208649e-06,
+      "loss": 0.9229,
+      "step": 7516
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938870310803784e-06,
+      "loss": 0.8975,
+      "step": 7517
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938310773295858e-06,
+      "loss": 0.8838,
+      "step": 7518
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.937751179573416e-06,
+      "loss": 0.9258,
+      "step": 7519
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.937191529647165e-06,
+      "loss": 0.9448,
+      "step": 7520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.936631823527811e-06,
+      "loss": 0.9258,
+      "step": 7521
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.936072061226064e-06,
+      "loss": 0.9531,
+      "step": 7522
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93551224275263e-06,
+      "loss": 0.9243,
+      "step": 7523
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.934952368118222e-06,
+      "loss": 0.9263,
+      "step": 7524
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.934392437333549e-06,
+      "loss": 0.3748,
+      "step": 7525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933832450409325e-06,
+      "loss": 0.8887,
+      "step": 7526
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933272407356265e-06,
+      "loss": 0.8159,
+      "step": 7527
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93271230818508e-06,
+      "loss": 0.9199,
+      "step": 7528
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932152152906486e-06,
+      "loss": 0.9199,
+      "step": 7529
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931591941531203e-06,
+      "loss": 0.8364,
+      "step": 7530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931031674069947e-06,
+      "loss": 0.9165,
+      "step": 7531
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.930471350533435e-06,
+      "loss": 0.8955,
+      "step": 7532
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92991097093239e-06,
+      "loss": 0.8994,
+      "step": 7533
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929350535277527e-06,
+      "loss": 0.8901,
+      "step": 7534
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928790043579574e-06,
+      "loss": 0.9326,
+      "step": 7535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928229495849254e-06,
+      "loss": 0.9165,
+      "step": 7536
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927668892097288e-06,
+      "loss": 0.8535,
+      "step": 7537
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927108232334404e-06,
+      "loss": 0.772,
+      "step": 7538
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.926547516571327e-06,
+      "loss": 0.9082,
+      "step": 7539
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925986744818783e-06,
+      "loss": 0.9067,
+      "step": 7540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925425917087502e-06,
+      "loss": 0.8926,
+      "step": 7541
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.924865033388215e-06,
+      "loss": 0.9404,
+      "step": 7542
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.924304093731647e-06,
+      "loss": 0.8623,
+      "step": 7543
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923743098128538e-06,
+      "loss": 0.9165,
+      "step": 7544
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923182046589612e-06,
+      "loss": 0.9302,
+      "step": 7545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.922620939125609e-06,
+      "loss": 0.9077,
+      "step": 7546
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.922059775747264e-06,
+      "loss": 0.8936,
+      "step": 7547
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92149855646531e-06,
+      "loss": 0.8975,
+      "step": 7548
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920937281290484e-06,
+      "loss": 0.8794,
+      "step": 7549
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920375950233523e-06,
+      "loss": 0.8345,
+      "step": 7550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91981456330517e-06,
+      "loss": 0.8511,
+      "step": 7551
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.919253120516165e-06,
+      "loss": 0.894,
+      "step": 7552
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918691621877246e-06,
+      "loss": 0.9209,
+      "step": 7553
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918130067399157e-06,
+      "loss": 0.9556,
+      "step": 7554
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.917568457092642e-06,
+      "loss": 0.8892,
+      "step": 7555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.917006790968445e-06,
+      "loss": 0.376,
+      "step": 7556
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916445069037311e-06,
+      "loss": 0.9253,
+      "step": 7557
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.915883291309987e-06,
+      "loss": 0.9224,
+      "step": 7558
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.915321457797221e-06,
+      "loss": 0.8667,
+      "step": 7559
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91475956850976e-06,
+      "loss": 0.8792,
+      "step": 7560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.914197623458357e-06,
+      "loss": 0.9189,
+      "step": 7561
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.913635622653761e-06,
+      "loss": 0.8999,
+      "step": 7562
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.913073566106723e-06,
+      "loss": 0.9048,
+      "step": 7563
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.912511453828001e-06,
+      "loss": 0.936,
+      "step": 7564
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911949285828343e-06,
+      "loss": 0.8921,
+      "step": 7565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911387062118506e-06,
+      "loss": 0.9312,
+      "step": 7566
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910824782709246e-06,
+      "loss": 0.9263,
+      "step": 7567
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910262447611322e-06,
+      "loss": 0.8203,
+      "step": 7568
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909700056835492e-06,
+      "loss": 0.8857,
+      "step": 7569
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909137610392513e-06,
+      "loss": 0.9248,
+      "step": 7570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.908575108293148e-06,
+      "loss": 0.8667,
+      "step": 7571
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.908012550548159e-06,
+      "loss": 0.8252,
+      "step": 7572
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907449937168307e-06,
+      "loss": 0.8408,
+      "step": 7573
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.906887268164356e-06,
+      "loss": 0.9585,
+      "step": 7574
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90632454354707e-06,
+      "loss": 0.873,
+      "step": 7575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905761763327218e-06,
+      "loss": 0.9272,
+      "step": 7576
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905198927515563e-06,
+      "loss": 0.917,
+      "step": 7577
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.904636036122874e-06,
+      "loss": 0.8916,
+      "step": 7578
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90407308915992e-06,
+      "loss": 0.9219,
+      "step": 7579
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.903510086637473e-06,
+      "loss": 0.8765,
+      "step": 7580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902947028566303e-06,
+      "loss": 0.8496,
+      "step": 7581
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902383914957181e-06,
+      "loss": 0.7861,
+      "step": 7582
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901820745820881e-06,
+      "loss": 0.8652,
+      "step": 7583
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901257521168177e-06,
+      "loss": 0.9185,
+      "step": 7584
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900694241009846e-06,
+      "loss": 0.9141,
+      "step": 7585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900130905356662e-06,
+      "loss": 0.9058,
+      "step": 7586
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899567514219404e-06,
+      "loss": 0.9189,
+      "step": 7587
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899004067608851e-06,
+      "loss": 0.8457,
+      "step": 7588
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.898440565535782e-06,
+      "loss": 0.9253,
+      "step": 7589
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897877008010978e-06,
+      "loss": 0.8527,
+      "step": 7590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897313395045218e-06,
+      "loss": 0.3428,
+      "step": 7591
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89674972664929e-06,
+      "loss": 0.895,
+      "step": 7592
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.896186002833975e-06,
+      "loss": 0.8193,
+      "step": 7593
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895622223610055e-06,
+      "loss": 0.9336,
+      "step": 7594
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895058388988321e-06,
+      "loss": 0.8804,
+      "step": 7595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894494498979558e-06,
+      "loss": 0.8867,
+      "step": 7596
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893930553594554e-06,
+      "loss": 0.8896,
+      "step": 7597
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893366552844097e-06,
+      "loss": 0.9014,
+      "step": 7598
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.892802496738981e-06,
+      "loss": 0.9126,
+      "step": 7599
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.892238385289991e-06,
+      "loss": 0.9307,
+      "step": 7600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.891674218507927e-06,
+      "loss": 0.9253,
+      "step": 7601
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.891109996403576e-06,
+      "loss": 0.8901,
+      "step": 7602
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.890545718987735e-06,
+      "loss": 0.8779,
+      "step": 7603
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.889981386271202e-06,
+      "loss": 1.0024,
+      "step": 7604
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.88941699826477e-06,
+      "loss": 0.9448,
+      "step": 7605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888852554979237e-06,
+      "loss": 0.9287,
+      "step": 7606
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.8882880564254e-06,
+      "loss": 0.896,
+      "step": 7607
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.887723502614064e-06,
+      "loss": 0.8892,
+      "step": 7608
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.887158893556027e-06,
+      "loss": 0.875,
+      "step": 7609
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.886594229262088e-06,
+      "loss": 0.8418,
+      "step": 7610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.886029509743057e-06,
+      "loss": 0.8242,
+      "step": 7611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88546473500973e-06,
+      "loss": 0.8911,
+      "step": 7612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884899905072917e-06,
+      "loss": 0.9277,
+      "step": 7613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884335019943424e-06,
+      "loss": 0.8555,
+      "step": 7614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.883770079632055e-06,
+      "loss": 0.8853,
+      "step": 7615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.883205084149621e-06,
+      "loss": 0.8862,
+      "step": 7616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88264003350693e-06,
+      "loss": 0.9194,
+      "step": 7617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.882074927714793e-06,
+      "loss": 0.897,
+      "step": 7618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.881509766784021e-06,
+      "loss": 0.8511,
+      "step": 7619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880944550725426e-06,
+      "loss": 0.9194,
+      "step": 7620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880379279549822e-06,
+      "loss": 0.8994,
+      "step": 7621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.879813953268025e-06,
+      "loss": 0.8408,
+      "step": 7622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.879248571890847e-06,
+      "loss": 0.8403,
+      "step": 7623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.878683135429108e-06,
+      "loss": 0.9072,
+      "step": 7624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.878117643893623e-06,
+      "loss": 0.9448,
+      "step": 7625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877552097295213e-06,
+      "loss": 0.9224,
+      "step": 7626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876986495644698e-06,
+      "loss": 0.957,
+      "step": 7627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876420838952896e-06,
+      "loss": 0.9463,
+      "step": 7628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87585512723063e-06,
+      "loss": 0.9219,
+      "step": 7629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875289360488727e-06,
+      "loss": 0.937,
+      "step": 7630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874723538738007e-06,
+      "loss": 0.8354,
+      "step": 7631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874157661989293e-06,
+      "loss": 0.9321,
+      "step": 7632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873591730253417e-06,
+      "loss": 0.9478,
+      "step": 7633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8730257435412e-06,
+      "loss": 0.9116,
+      "step": 7634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872459701863475e-06,
+      "loss": 0.8828,
+      "step": 7635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87189360523107e-06,
+      "loss": 0.834,
+      "step": 7636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.871327453654811e-06,
+      "loss": 0.8794,
+      "step": 7637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870761247145538e-06,
+      "loss": 0.8472,
+      "step": 7638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870194985714075e-06,
+      "loss": 0.8472,
+      "step": 7639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869628669371261e-06,
+      "loss": 0.8984,
+      "step": 7640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869062298127928e-06,
+      "loss": 0.8936,
+      "step": 7641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.868495871994912e-06,
+      "loss": 0.916,
+      "step": 7642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86792939098305e-06,
+      "loss": 0.8682,
+      "step": 7643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.867362855103179e-06,
+      "loss": 0.9209,
+      "step": 7644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866796264366138e-06,
+      "loss": 0.8564,
+      "step": 7645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866229618782768e-06,
+      "loss": 0.8804,
+      "step": 7646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86566291836391e-06,
+      "loss": 0.8804,
+      "step": 7647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.865096163120403e-06,
+      "loss": 0.8521,
+      "step": 7648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86452935306309e-06,
+      "loss": 0.9097,
+      "step": 7649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.863962488202819e-06,
+      "loss": 0.9409,
+      "step": 7650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.863395568550432e-06,
+      "loss": 0.7834,
+      "step": 7651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862828594116777e-06,
+      "loss": 0.9365,
+      "step": 7652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862261564912696e-06,
+      "loss": 0.8535,
+      "step": 7653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.861694480949043e-06,
+      "loss": 0.9307,
+      "step": 7654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.861127342236665e-06,
+      "loss": 0.8989,
+      "step": 7655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.860560148786414e-06,
+      "loss": 0.8931,
+      "step": 7656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859992900609137e-06,
+      "loss": 0.8354,
+      "step": 7657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.85942559771569e-06,
+      "loss": 0.8022,
+      "step": 7658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858858240116926e-06,
+      "loss": 0.9697,
+      "step": 7659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858290827823699e-06,
+      "loss": 0.9639,
+      "step": 7660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857723360846863e-06,
+      "loss": 0.9341,
+      "step": 7661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857155839197278e-06,
+      "loss": 0.8906,
+      "step": 7662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856588262885795e-06,
+      "loss": 0.8604,
+      "step": 7663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856020631923283e-06,
+      "loss": 0.8682,
+      "step": 7664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855452946320592e-06,
+      "loss": 0.9043,
+      "step": 7665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.854885206088587e-06,
+      "loss": 0.8931,
+      "step": 7666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.85431741123813e-06,
+      "loss": 0.9092,
+      "step": 7667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.853749561780084e-06,
+      "loss": 0.9087,
+      "step": 7668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.85318165772531e-06,
+      "loss": 0.9126,
+      "step": 7669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852613699084674e-06,
+      "loss": 0.8262,
+      "step": 7670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852045685869046e-06,
+      "loss": 0.3934,
+      "step": 7671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.851477618089287e-06,
+      "loss": 0.8643,
+      "step": 7672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850909495756269e-06,
+      "loss": 0.812,
+      "step": 7673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850341318880861e-06,
+      "loss": 0.9204,
+      "step": 7674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.849773087473931e-06,
+      "loss": 0.9155,
+      "step": 7675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84920480154635e-06,
+      "loss": 0.8608,
+      "step": 7676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.848636461108993e-06,
+      "loss": 0.8979,
+      "step": 7677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.848068066172732e-06,
+      "loss": 0.9268,
+      "step": 7678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84749961674844e-06,
+      "loss": 0.8696,
+      "step": 7679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846931112846992e-06,
+      "loss": 0.8691,
+      "step": 7680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846362554479266e-06,
+      "loss": 0.9165,
+      "step": 7681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84579394165614e-06,
+      "loss": 0.9019,
+      "step": 7682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845225274388492e-06,
+      "loss": 0.8779,
+      "step": 7683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8446565526872e-06,
+      "loss": 0.9688,
+      "step": 7684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.844087776563144e-06,
+      "loss": 0.8525,
+      "step": 7685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84351894602721e-06,
+      "loss": 0.8618,
+      "step": 7686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842950061090277e-06,
+      "loss": 0.8843,
+      "step": 7687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84238112176323e-06,
+      "loss": 0.9321,
+      "step": 7688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841812128056951e-06,
+      "loss": 0.897,
+      "step": 7689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841243079982327e-06,
+      "loss": 0.9185,
+      "step": 7690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840673977550249e-06,
+      "loss": 0.9126,
+      "step": 7691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840104820771601e-06,
+      "loss": 0.877,
+      "step": 7692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839535609657272e-06,
+      "loss": 0.9526,
+      "step": 7693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.838966344218151e-06,
+      "loss": 0.8931,
+      "step": 7694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83839702446513e-06,
+      "loss": 0.875,
+      "step": 7695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.837827650409102e-06,
+      "loss": 0.917,
+      "step": 7696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83725822206096e-06,
+      "loss": 0.894,
+      "step": 7697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836688739431595e-06,
+      "loss": 0.8779,
+      "step": 7698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836119202531906e-06,
+      "loss": 0.8696,
+      "step": 7699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.835549611372788e-06,
+      "loss": 0.9097,
+      "step": 7700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.834979965965138e-06,
+      "loss": 0.9062,
+      "step": 7701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83441026631985e-06,
+      "loss": 0.9473,
+      "step": 7702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83384051244783e-06,
+      "loss": 0.9141,
+      "step": 7703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.833270704359976e-06,
+      "loss": 0.897,
+      "step": 7704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83270084206719e-06,
+      "loss": 0.8862,
+      "step": 7705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.832130925580369e-06,
+      "loss": 0.9775,
+      "step": 7706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.831560954910422e-06,
+      "loss": 0.9243,
+      "step": 7707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830990930068254e-06,
+      "loss": 0.8862,
+      "step": 7708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830420851064767e-06,
+      "loss": 1.0068,
+      "step": 7709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829850717910868e-06,
+      "loss": 0.8613,
+      "step": 7710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829280530617468e-06,
+      "loss": 0.875,
+      "step": 7711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828710289195472e-06,
+      "loss": 0.8931,
+      "step": 7712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82813999365579e-06,
+      "loss": 0.9023,
+      "step": 7713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.827569644009332e-06,
+      "loss": 0.8994,
+      "step": 7714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.826999240267013e-06,
+      "loss": 0.3008,
+      "step": 7715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.826428782439743e-06,
+      "loss": 0.8442,
+      "step": 7716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.825858270538436e-06,
+      "loss": 0.8687,
+      "step": 7717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82528770457401e-06,
+      "loss": 0.9438,
+      "step": 7718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.824717084557374e-06,
+      "loss": 0.833,
+      "step": 7719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.824146410499452e-06,
+      "loss": 0.8779,
+      "step": 7720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823575682411158e-06,
+      "loss": 0.9399,
+      "step": 7721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823004900303412e-06,
+      "loss": 0.9282,
+      "step": 7722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.822434064187133e-06,
+      "loss": 0.938,
+      "step": 7723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821863174073244e-06,
+      "loss": 0.4067,
+      "step": 7724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821292229972665e-06,
+      "loss": 0.8887,
+      "step": 7725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82072123189632e-06,
+      "loss": 0.8696,
+      "step": 7726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.820150179855132e-06,
+      "loss": 0.8284,
+      "step": 7727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81957907386003e-06,
+      "loss": 0.876,
+      "step": 7728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.819007913921933e-06,
+      "loss": 0.9214,
+      "step": 7729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818436700051775e-06,
+      "loss": 0.8394,
+      "step": 7730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81786543226048e-06,
+      "loss": 0.8706,
+      "step": 7731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817294110558981e-06,
+      "loss": 0.9331,
+      "step": 7732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816722734958204e-06,
+      "loss": 0.7915,
+      "step": 7733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816151305469084e-06,
+      "loss": 0.8711,
+      "step": 7734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815579822102551e-06,
+      "loss": 0.9043,
+      "step": 7735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815008284869541e-06,
+      "loss": 0.9648,
+      "step": 7736
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814436693780985e-06,
+      "loss": 0.894,
+      "step": 7737
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81386504884782e-06,
+      "loss": 0.8467,
+      "step": 7738
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.813293350080983e-06,
+      "loss": 0.9043,
+      "step": 7739
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812721597491411e-06,
+      "loss": 0.9199,
+      "step": 7740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812149791090043e-06,
+      "loss": 0.9468,
+      "step": 7741
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.811577930887817e-06,
+      "loss": 0.8916,
+      "step": 7742
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.811006016895675e-06,
+      "loss": 0.9355,
+      "step": 7743
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810434049124557e-06,
+      "loss": 0.855,
+      "step": 7744
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809862027585409e-06,
+      "loss": 0.915,
+      "step": 7745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809289952289173e-06,
+      "loss": 0.8887,
+      "step": 7746
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808717823246792e-06,
+      "loss": 0.957,
+      "step": 7747
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808145640469213e-06,
+      "loss": 0.9146,
+      "step": 7748
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807573403967383e-06,
+      "loss": 0.8511,
+      "step": 7749
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80700111375225e-06,
+      "loss": 0.937,
+      "step": 7750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.806428769834763e-06,
+      "loss": 0.9419,
+      "step": 7751
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80585637222587e-06,
+      "loss": 0.8945,
+      "step": 7752
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805283920936524e-06,
+      "loss": 0.9062,
+      "step": 7753
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.804711415977676e-06,
+      "loss": 0.9326,
+      "step": 7754
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.804138857360277e-06,
+      "loss": 0.9624,
+      "step": 7755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.803566245095285e-06,
+      "loss": 0.8877,
+      "step": 7756
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802993579193652e-06,
+      "loss": 0.8662,
+      "step": 7757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802420859666335e-06,
+      "loss": 0.3347,
+      "step": 7758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801848086524292e-06,
+      "loss": 0.9092,
+      "step": 7759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801275259778479e-06,
+      "loss": 0.9102,
+      "step": 7760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.800702379439856e-06,
+      "loss": 0.9326,
+      "step": 7761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.800129445519382e-06,
+      "loss": 0.9175,
+      "step": 7762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799556458028022e-06,
+      "loss": 0.3268,
+      "step": 7763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798983416976733e-06,
+      "loss": 0.9102,
+      "step": 7764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798410322376481e-06,
+      "loss": 0.8813,
+      "step": 7765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79783717423823e-06,
+      "loss": 0.9082,
+      "step": 7766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.797263972572947e-06,
+      "loss": 0.9155,
+      "step": 7767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.796690717391592e-06,
+      "loss": 0.8501,
+      "step": 7768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79611740870514e-06,
+      "loss": 0.9351,
+      "step": 7769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.795544046524554e-06,
+      "loss": 0.8794,
+      "step": 7770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.794970630860807e-06,
+      "loss": 0.9077,
+      "step": 7771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.794397161724866e-06,
+      "loss": 0.877,
+      "step": 7772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.793823639127705e-06,
+      "loss": 0.9219,
+      "step": 7773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.793250063080294e-06,
+      "loss": 0.9321,
+      "step": 7774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79267643359361e-06,
+      "loss": 0.876,
+      "step": 7775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792102750678621e-06,
+      "loss": 0.9106,
+      "step": 7776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79152901434631e-06,
+      "loss": 0.8965,
+      "step": 7777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790955224607648e-06,
+      "loss": 0.9097,
+      "step": 7778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790381381473618e-06,
+      "loss": 0.9238,
+      "step": 7779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78980748495519e-06,
+      "loss": 0.9258,
+      "step": 7780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.789233535063352e-06,
+      "loss": 0.8618,
+      "step": 7781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78865953180908e-06,
+      "loss": 0.8906,
+      "step": 7782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788085475203357e-06,
+      "loss": 0.8706,
+      "step": 7783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.787511365257168e-06,
+      "loss": 0.8428,
+      "step": 7784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78693720198149e-06,
+      "loss": 0.8599,
+      "step": 7785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.786362985387314e-06,
+      "loss": 0.8999,
+      "step": 7786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785788715485623e-06,
+      "loss": 0.8657,
+      "step": 7787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785214392287403e-06,
+      "loss": 0.876,
+      "step": 7788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784640015803644e-06,
+      "loss": 0.4237,
+      "step": 7789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784065586045332e-06,
+      "loss": 0.874,
+      "step": 7790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78349110302346e-06,
+      "loss": 0.8667,
+      "step": 7791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.782916566749015e-06,
+      "loss": 0.9131,
+      "step": 7792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.782341977232993e-06,
+      "loss": 0.8525,
+      "step": 7793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781767334486382e-06,
+      "loss": 0.8965,
+      "step": 7794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781192638520179e-06,
+      "loss": 0.9204,
+      "step": 7795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.780617889345378e-06,
+      "loss": 0.8765,
+      "step": 7796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.780043086972976e-06,
+      "loss": 0.8828,
+      "step": 7797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77946823141397e-06,
+      "loss": 0.8423,
+      "step": 7798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778893322679355e-06,
+      "loss": 0.9067,
+      "step": 7799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778318360780134e-06,
+      "loss": 0.9019,
+      "step": 7800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.777743345727302e-06,
+      "loss": 0.8896,
+      "step": 7801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.777168277531863e-06,
+      "loss": 0.9429,
+      "step": 7802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77659315620482e-06,
+      "loss": 0.9517,
+      "step": 7803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776017981757174e-06,
+      "loss": 0.8608,
+      "step": 7804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.775442754199929e-06,
+      "loss": 0.9268,
+      "step": 7805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774867473544092e-06,
+      "loss": 0.876,
+      "step": 7806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774292139800666e-06,
+      "loss": 0.873,
+      "step": 7807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77371675298066e-06,
+      "loss": 0.8594,
+      "step": 7808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.773141313095082e-06,
+      "loss": 0.9512,
+      "step": 7809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77256582015494e-06,
+      "loss": 0.9316,
+      "step": 7810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.771990274171247e-06,
+      "loss": 0.9004,
+      "step": 7811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77141467515501e-06,
+      "loss": 0.9038,
+      "step": 7812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770839023117242e-06,
+      "loss": 0.856,
+      "step": 7813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77026331806896e-06,
+      "loss": 0.9243,
+      "step": 7814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.769687560021173e-06,
+      "loss": 0.9087,
+      "step": 7815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.7691117489849e-06,
+      "loss": 0.9272,
+      "step": 7816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.768535884971156e-06,
+      "loss": 0.8784,
+      "step": 7817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767959967990957e-06,
+      "loss": 0.917,
+      "step": 7818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767383998055322e-06,
+      "loss": 0.9336,
+      "step": 7819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76680797517527e-06,
+      "loss": 0.8589,
+      "step": 7820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.766231899361823e-06,
+      "loss": 0.9326,
+      "step": 7821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765655770625997e-06,
+      "loss": 0.8813,
+      "step": 7822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76507958897882e-06,
+      "loss": 0.9087,
+      "step": 7823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764503354431316e-06,
+      "loss": 0.8652,
+      "step": 7824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.763927066994503e-06,
+      "loss": 0.9458,
+      "step": 7825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76335072667941e-06,
+      "loss": 0.9131,
+      "step": 7826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.762774333497065e-06,
+      "loss": 0.3208,
+      "step": 7827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.762197887458494e-06,
+      "loss": 0.9429,
+      "step": 7828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761621388574722e-06,
+      "loss": 0.896,
+      "step": 7829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761044836856783e-06,
+      "loss": 0.9526,
+      "step": 7830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.760468232315705e-06,
+      "loss": 0.339,
+      "step": 7831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75989157496252e-06,
+      "loss": 0.8833,
+      "step": 7832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.759314864808258e-06,
+      "loss": 0.9082,
+      "step": 7833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758738101863957e-06,
+      "loss": 0.8911,
+      "step": 7834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758161286140648e-06,
+      "loss": 0.9014,
+      "step": 7835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757584417649366e-06,
+      "loss": 0.8638,
+      "step": 7836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757007496401151e-06,
+      "loss": 0.8701,
+      "step": 7837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.756430522407036e-06,
+      "loss": 0.9604,
+      "step": 7838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755853495678063e-06,
+      "loss": 0.8799,
+      "step": 7839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755276416225271e-06,
+      "loss": 0.8882,
+      "step": 7840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.754699284059696e-06,
+      "loss": 0.8813,
+      "step": 7841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.754122099192386e-06,
+      "loss": 0.8926,
+      "step": 7842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753544861634379e-06,
+      "loss": 0.8652,
+      "step": 7843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.752967571396718e-06,
+      "loss": 0.8589,
+      "step": 7844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.752390228490451e-06,
+      "loss": 0.9233,
+      "step": 7845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.751812832926624e-06,
+      "loss": 0.9512,
+      "step": 7846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.751235384716278e-06,
+      "loss": 0.9004,
+      "step": 7847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750657883870465e-06,
+      "loss": 0.9619,
+      "step": 7848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750080330400232e-06,
+      "loss": 0.3909,
+      "step": 7849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.749502724316628e-06,
+      "loss": 0.8608,
+      "step": 7850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748925065630704e-06,
+      "loss": 0.8887,
+      "step": 7851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748347354353513e-06,
+      "loss": 0.9033,
+      "step": 7852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747769590496106e-06,
+      "loss": 0.9067,
+      "step": 7853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747191774069536e-06,
+      "loss": 0.5613,
+      "step": 7854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.746613905084859e-06,
+      "loss": 0.321,
+      "step": 7855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74603598355313e-06,
+      "loss": 0.8457,
+      "step": 7856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.745458009485404e-06,
+      "loss": 0.9116,
+      "step": 7857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744879982892741e-06,
+      "loss": 0.8965,
+      "step": 7858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744301903786199e-06,
+      "loss": 0.8853,
+      "step": 7859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743723772176837e-06,
+      "loss": 0.9297,
+      "step": 7860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743145588075714e-06,
+      "loss": 0.9121,
+      "step": 7861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.742567351493895e-06,
+      "loss": 0.8755,
+      "step": 7862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74198906244244e-06,
+      "loss": 0.3348,
+      "step": 7863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741410720932414e-06,
+      "loss": 0.9028,
+      "step": 7864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74083232697488e-06,
+      "loss": 0.8525,
+      "step": 7865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.740253880580906e-06,
+      "loss": 0.9258,
+      "step": 7866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739675381761556e-06,
+      "loss": 0.8896,
+      "step": 7867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.7390968305279e-06,
+      "loss": 0.9258,
+      "step": 7868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.738518226891004e-06,
+      "loss": 0.9019,
+      "step": 7869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.73793957086194e-06,
+      "loss": 0.9136,
+      "step": 7870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737360862451776e-06,
+      "loss": 0.8345,
+      "step": 7871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736782101671587e-06,
+      "loss": 0.9722,
+      "step": 7872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736203288532443e-06,
+      "loss": 0.917,
+      "step": 7873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735624423045418e-06,
+      "loss": 0.8862,
+      "step": 7874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735045505221588e-06,
+      "loss": 0.8887,
+      "step": 7875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734466535072026e-06,
+      "loss": 0.8896,
+      "step": 7876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.733887512607812e-06,
+      "loss": 0.9502,
+      "step": 7877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.733308437840022e-06,
+      "loss": 0.29,
+      "step": 7878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732729310779731e-06,
+      "loss": 0.9458,
+      "step": 7879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732150131438026e-06,
+      "loss": 0.8633,
+      "step": 7880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.731570899825982e-06,
+      "loss": 0.8555,
+      "step": 7881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730991615954681e-06,
+      "loss": 0.9146,
+      "step": 7882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730412279835209e-06,
+      "loss": 0.9292,
+      "step": 7883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729832891478646e-06,
+      "loss": 0.9302,
+      "step": 7884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729253450896079e-06,
+      "loss": 0.8652,
+      "step": 7885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728673958098593e-06,
+      "loss": 0.8672,
+      "step": 7886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728094413097274e-06,
+      "loss": 0.877,
+      "step": 7887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72751481590321e-06,
+      "loss": 0.814,
+      "step": 7888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72693516652749e-06,
+      "loss": 0.8882,
+      "step": 7889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726355464981201e-06,
+      "loss": 0.8706,
+      "step": 7890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725775711275438e-06,
+      "loss": 0.8823,
+      "step": 7891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72519590542129e-06,
+      "loss": 0.9312,
+      "step": 7892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72461604742985e-06,
+      "loss": 0.832,
+      "step": 7893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.724036137312209e-06,
+      "loss": 0.9395,
+      "step": 7894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.723456175079467e-06,
+      "loss": 0.9053,
+      "step": 7895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.722876160742716e-06,
+      "loss": 0.9663,
+      "step": 7896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.722296094313053e-06,
+      "loss": 0.8931,
+      "step": 7897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.721715975801576e-06,
+      "loss": 0.8701,
+      "step": 7898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.721135805219385e-06,
+      "loss": 0.8706,
+      "step": 7899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.720555582577576e-06,
+      "loss": 0.8638,
+      "step": 7900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719975307887253e-06,
+      "loss": 0.959,
+      "step": 7901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719394981159515e-06,
+      "loss": 0.8833,
+      "step": 7902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.718814602405466e-06,
+      "loss": 0.8403,
+      "step": 7903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71823417163621e-06,
+      "loss": 0.938,
+      "step": 7904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71765368886285e-06,
+      "loss": 0.8691,
+      "step": 7905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.717073154096493e-06,
+      "loss": 0.8657,
+      "step": 7906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.716492567348246e-06,
+      "loss": 0.8521,
+      "step": 7907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.715911928629214e-06,
+      "loss": 0.9263,
+      "step": 7908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.715331237950507e-06,
+      "loss": 0.8452,
+      "step": 7909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714750495323235e-06,
+      "loss": 0.8525,
+      "step": 7910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714169700758509e-06,
+      "loss": 0.9004,
+      "step": 7911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71358885426744e-06,
+      "loss": 0.855,
+      "step": 7912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.713007955861138e-06,
+      "loss": 0.8794,
+      "step": 7913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71242700555072e-06,
+      "loss": 0.9009,
+      "step": 7914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.711846003347298e-06,
+      "loss": 0.8633,
+      "step": 7915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.711264949261989e-06,
+      "loss": 0.8247,
+      "step": 7916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71068384330591e-06,
+      "loss": 0.9165,
+      "step": 7917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710102685490178e-06,
+      "loss": 0.8755,
+      "step": 7918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70952147582591e-06,
+      "loss": 0.8354,
+      "step": 7919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708940214324228e-06,
+      "loss": 0.9102,
+      "step": 7920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708358900996248e-06,
+      "loss": 0.873,
+      "step": 7921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.707777535853096e-06,
+      "loss": 0.8623,
+      "step": 7922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.707196118905893e-06,
+      "loss": 0.9429,
+      "step": 7923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706614650165763e-06,
+      "loss": 0.9468,
+      "step": 7924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706033129643828e-06,
+      "loss": 0.895,
+      "step": 7925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.705451557351215e-06,
+      "loss": 0.8745,
+      "step": 7926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.704869933299052e-06,
+      "loss": 0.8433,
+      "step": 7927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.704288257498464e-06,
+      "loss": 0.8911,
+      "step": 7928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70370652996058e-06,
+      "loss": 0.9033,
+      "step": 7929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.703124750696529e-06,
+      "loss": 0.8867,
+      "step": 7930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.702542919717444e-06,
+      "loss": 0.8662,
+      "step": 7931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701961037034454e-06,
+      "loss": 0.8745,
+      "step": 7932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701379102658689e-06,
+      "loss": 0.917,
+      "step": 7933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.700797116601286e-06,
+      "loss": 0.9053,
+      "step": 7934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70021507887338e-06,
+      "loss": 0.8652,
+      "step": 7935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699632989486104e-06,
+      "loss": 0.8916,
+      "step": 7936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699050848450593e-06,
+      "loss": 0.8813,
+      "step": 7937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.698468655777987e-06,
+      "loss": 0.9478,
+      "step": 7938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.697886411479422e-06,
+      "loss": 0.8691,
+      "step": 7939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.697304115566042e-06,
+      "loss": 0.8345,
+      "step": 7940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.696721768048981e-06,
+      "loss": 0.8472,
+      "step": 7941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.696139368939384e-06,
+      "loss": 0.9355,
+      "step": 7942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695556918248394e-06,
+      "loss": 0.9121,
+      "step": 7943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.694974415987151e-06,
+      "loss": 0.9307,
+      "step": 7944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6943918621668e-06,
+      "loss": 0.8999,
+      "step": 7945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.693809256798487e-06,
+      "loss": 0.9219,
+      "step": 7946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.693226599893357e-06,
+      "loss": 0.9175,
+      "step": 7947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69264389146256e-06,
+      "loss": 0.8827,
+      "step": 7948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.692061131517243e-06,
+      "loss": 0.8403,
+      "step": 7949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691478320068552e-06,
+      "loss": 0.874,
+      "step": 7950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.690895457127641e-06,
+      "loss": 0.8281,
+      "step": 7951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69031254270566e-06,
+      "loss": 0.8774,
+      "step": 7952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68972957681376e-06,
+      "loss": 0.9351,
+      "step": 7953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.689146559463095e-06,
+      "loss": 0.9243,
+      "step": 7954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.688563490664817e-06,
+      "loss": 0.9395,
+      "step": 7955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.687980370430085e-06,
+      "loss": 0.8672,
+      "step": 7956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.687397198770053e-06,
+      "loss": 0.9307,
+      "step": 7957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686813975695874e-06,
+      "loss": 0.9448,
+      "step": 7958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686230701218714e-06,
+      "loss": 0.9033,
+      "step": 7959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.685647375349725e-06,
+      "loss": 0.8389,
+      "step": 7960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.685063998100071e-06,
+      "loss": 0.8882,
+      "step": 7961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684480569480912e-06,
+      "loss": 0.9585,
+      "step": 7962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68389708950341e-06,
+      "loss": 0.9507,
+      "step": 7963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.683313558178724e-06,
+      "loss": 0.8613,
+      "step": 7964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682729975518024e-06,
+      "loss": 0.9307,
+      "step": 7965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682146341532473e-06,
+      "loss": 0.9219,
+      "step": 7966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681562656233233e-06,
+      "loss": 0.8555,
+      "step": 7967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680978919631477e-06,
+      "loss": 0.8784,
+      "step": 7968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680395131738367e-06,
+      "loss": 0.8491,
+      "step": 7969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679811292565074e-06,
+      "loss": 0.9248,
+      "step": 7970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67922740212277e-06,
+      "loss": 0.8477,
+      "step": 7971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.678643460422625e-06,
+      "loss": 0.9253,
+      "step": 7972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.678059467475806e-06,
+      "loss": 0.8975,
+      "step": 7973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677475423293491e-06,
+      "loss": 0.9062,
+      "step": 7974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.676891327886853e-06,
+      "loss": 0.8833,
+      "step": 7975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.676307181267064e-06,
+      "loss": 0.916,
+      "step": 7976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675722983445303e-06,
+      "loss": 0.9229,
+      "step": 7977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675138734432744e-06,
+      "loss": 0.9092,
+      "step": 7978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.674554434240565e-06,
+      "loss": 0.8882,
+      "step": 7979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.673970082879945e-06,
+      "loss": 0.8979,
+      "step": 7980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.673385680362065e-06,
+      "loss": 0.8784,
+      "step": 7981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672801226698104e-06,
+      "loss": 0.8848,
+      "step": 7982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672216721899242e-06,
+      "loss": 0.8848,
+      "step": 7983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671632165976664e-06,
+      "loss": 0.8931,
+      "step": 7984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671047558941553e-06,
+      "loss": 0.936,
+      "step": 7985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.670462900805094e-06,
+      "loss": 0.8691,
+      "step": 7986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66987819157847e-06,
+      "loss": 0.9292,
+      "step": 7987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66929343127287e-06,
+      "loss": 0.9248,
+      "step": 7988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66870861989948e-06,
+      "loss": 0.834,
+      "step": 7989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66812375746949e-06,
+      "loss": 0.9238,
+      "step": 7990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.667538843994086e-06,
+      "loss": 0.8916,
+      "step": 7991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666953879484463e-06,
+      "loss": 0.9121,
+      "step": 7992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666368863951808e-06,
+      "loss": 0.8823,
+      "step": 7993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665783797407314e-06,
+      "loss": 0.297,
+      "step": 7994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665198679862176e-06,
+      "loss": 0.895,
+      "step": 7995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.664613511327588e-06,
+      "loss": 0.832,
+      "step": 7996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.664028291814744e-06,
+      "loss": 0.8511,
+      "step": 7997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.663443021334842e-06,
+      "loss": 0.9404,
+      "step": 7998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662857699899075e-06,
+      "loss": 0.9277,
+      "step": 7999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662272327518648e-06,
+      "loss": 0.8682,
+      "step": 8000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.661686904204752e-06,
+      "loss": 0.8975,
+      "step": 8001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.661101429968596e-06,
+      "loss": 0.9028,
+      "step": 8002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.660515904821374e-06,
+      "loss": 0.9292,
+      "step": 8003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659930328774288e-06,
+      "loss": 0.7915,
+      "step": 8004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659344701838547e-06,
+      "loss": 0.8828,
+      "step": 8005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658759024025349e-06,
+      "loss": 0.8862,
+      "step": 8006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658173295345901e-06,
+      "loss": 0.8979,
+      "step": 8007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65758751581141e-06,
+      "loss": 0.8091,
+      "step": 8008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65700168543308e-06,
+      "loss": 0.9116,
+      "step": 8009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.656415804222124e-06,
+      "loss": 0.8589,
+      "step": 8010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655829872189746e-06,
+      "loss": 0.876,
+      "step": 8011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655243889347158e-06,
+      "loss": 0.9385,
+      "step": 8012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654657855705569e-06,
+      "loss": 0.8794,
+      "step": 8013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654071771276195e-06,
+      "loss": 0.8853,
+      "step": 8014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.653485636070243e-06,
+      "loss": 0.9028,
+      "step": 8015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.652899450098929e-06,
+      "loss": 0.9849,
+      "step": 8016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65231321337347e-06,
+      "loss": 0.8311,
+      "step": 8017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65172692590508e-06,
+      "loss": 0.8613,
+      "step": 8018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651140587704976e-06,
+      "loss": 0.8599,
+      "step": 8019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65055419878437e-06,
+      "loss": 0.8643,
+      "step": 8020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64996775915449e-06,
+      "loss": 0.8184,
+      "step": 8021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64938126882655e-06,
+      "loss": 0.8379,
+      "step": 8022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64879472781177e-06,
+      "loss": 0.9336,
+      "step": 8023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.648208136121374e-06,
+      "loss": 0.9189,
+      "step": 8024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647621493766583e-06,
+      "loss": 0.8955,
+      "step": 8025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64703480075862e-06,
+      "loss": 0.834,
+      "step": 8026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.646448057108711e-06,
+      "loss": 0.897,
+      "step": 8027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64586126282808e-06,
+      "loss": 0.8936,
+      "step": 8028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645274417927955e-06,
+      "loss": 0.8804,
+      "step": 8029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64468752241956e-06,
+      "loss": 0.875,
+      "step": 8030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.644100576314126e-06,
+      "loss": 0.8853,
+      "step": 8031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643513579622882e-06,
+      "loss": 0.9775,
+      "step": 8032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642926532357055e-06,
+      "loss": 0.9053,
+      "step": 8033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642339434527881e-06,
+      "loss": 0.8486,
+      "step": 8034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64175228614659e-06,
+      "loss": 0.8789,
+      "step": 8035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.641165087224414e-06,
+      "loss": 0.8628,
+      "step": 8036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64057783777259e-06,
+      "loss": 0.8677,
+      "step": 8037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.639990537802349e-06,
+      "loss": 0.8779,
+      "step": 8038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63940318732493e-06,
+      "loss": 0.8687,
+      "step": 8039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638815786351566e-06,
+      "loss": 0.8633,
+      "step": 8040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6382283348935e-06,
+      "loss": 0.8657,
+      "step": 8041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637640832961967e-06,
+      "loss": 0.9482,
+      "step": 8042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637053280568212e-06,
+      "loss": 0.8442,
+      "step": 8043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63646567772347e-06,
+      "loss": 0.873,
+      "step": 8044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635878024438985e-06,
+      "loss": 0.8442,
+      "step": 8045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635290320725999e-06,
+      "loss": 0.937,
+      "step": 8046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634702566595758e-06,
+      "loss": 0.3209,
+      "step": 8047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634114762059504e-06,
+      "loss": 0.8975,
+      "step": 8048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.633526907128483e-06,
+      "loss": 0.894,
+      "step": 8049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632939001813944e-06,
+      "loss": 0.8989,
+      "step": 8050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632351046127133e-06,
+      "loss": 0.3933,
+      "step": 8051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6317630400793e-06,
+      "loss": 0.8818,
+      "step": 8052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631174983681689e-06,
+      "loss": 0.8828,
+      "step": 8053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.630586876945556e-06,
+      "loss": 0.915,
+      "step": 8054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62999871988215e-06,
+      "loss": 0.938,
+      "step": 8055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.629410512502727e-06,
+      "loss": 0.3408,
+      "step": 8056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628822254818534e-06,
+      "loss": 0.8809,
+      "step": 8057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62823394684083e-06,
+      "loss": 0.3414,
+      "step": 8058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62764558858087e-06,
+      "loss": 0.8848,
+      "step": 8059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.627057180049908e-06,
+      "loss": 0.918,
+      "step": 8060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.626468721259204e-06,
+      "loss": 0.8608,
+      "step": 8061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6258802122200125e-06,
+      "loss": 0.8955,
+      "step": 8062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.625291652943596e-06,
+      "loss": 0.8701,
+      "step": 8063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624703043441213e-06,
+      "loss": 0.9263,
+      "step": 8064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624114383724124e-06,
+      "loss": 0.8999,
+      "step": 8065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.623525673803592e-06,
+      "loss": 0.9502,
+      "step": 8066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6229369136908814e-06,
+      "loss": 0.8481,
+      "step": 8067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.622348103397252e-06,
+      "loss": 0.8877,
+      "step": 8068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.621759242933971e-06,
+      "loss": 0.8735,
+      "step": 8069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6211703323123065e-06,
+      "loss": 0.8921,
+      "step": 8070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.620581371543523e-06,
+      "loss": 0.9082,
+      "step": 8071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6199923606388874e-06,
+      "loss": 0.8755,
+      "step": 8072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6194032996096685e-06,
+      "loss": 0.918,
+      "step": 8073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.618814188467139e-06,
+      "loss": 0.9282,
+      "step": 8074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.618225027222564e-06,
+      "loss": 0.8936,
+      "step": 8075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.617635815887221e-06,
+      "loss": 0.9487,
+      "step": 8076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6170465544723795e-06,
+      "loss": 0.8794,
+      "step": 8077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.616457242989313e-06,
+      "loss": 0.877,
+      "step": 8078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615867881449297e-06,
+      "loss": 0.9189,
+      "step": 8079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615278469863607e-06,
+      "loss": 0.8691,
+      "step": 8080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614689008243517e-06,
+      "loss": 0.9458,
+      "step": 8081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614099496600307e-06,
+      "loss": 0.8711,
+      "step": 8082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.613509934945254e-06,
+      "loss": 0.9268,
+      "step": 8083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6129203232896375e-06,
+      "loss": 0.9639,
+      "step": 8084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612330661644738e-06,
+      "loss": 0.8628,
+      "step": 8085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.611740950021836e-06,
+      "loss": 0.9062,
+      "step": 8086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.611151188432215e-06,
+      "loss": 0.9087,
+      "step": 8087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.610561376887155e-06,
+      "loss": 0.9004,
+      "step": 8088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609971515397945e-06,
+      "loss": 0.8423,
+      "step": 8089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609381603975865e-06,
+      "loss": 0.811,
+      "step": 8090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.608791642632204e-06,
+      "loss": 0.8359,
+      "step": 8091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.608201631378247e-06,
+      "loss": 0.9321,
+      "step": 8092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.607611570225284e-06,
+      "loss": 0.8804,
+      "step": 8093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6070214591846026e-06,
+      "loss": 0.8833,
+      "step": 8094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.606431298267492e-06,
+      "loss": 0.8618,
+      "step": 8095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.605841087485243e-06,
+      "loss": 0.9204,
+      "step": 8096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.605250826849149e-06,
+      "loss": 0.8711,
+      "step": 8097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.604660516370499e-06,
+      "loss": 0.937,
+      "step": 8098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60407015606059e-06,
+      "loss": 0.8906,
+      "step": 8099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.603479745930717e-06,
+      "loss": 0.8857,
+      "step": 8100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6028892859921725e-06,
+      "loss": 0.9482,
+      "step": 8101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6022987762562536e-06,
+      "loss": 0.9502,
+      "step": 8102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60170821673426e-06,
+      "loss": 0.9229,
+      "step": 8103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601117607437488e-06,
+      "loss": 0.9263,
+      "step": 8104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.600526948377237e-06,
+      "loss": 0.8926,
+      "step": 8105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.599936239564807e-06,
+      "loss": 0.8867,
+      "step": 8106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5993454810115005e-06,
+      "loss": 0.9263,
+      "step": 8107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5987546727286185e-06,
+      "loss": 0.8447,
+      "step": 8108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.598163814727465e-06,
+      "loss": 0.9712,
+      "step": 8109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597572907019343e-06,
+      "loss": 0.8726,
+      "step": 8110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596981949615558e-06,
+      "loss": 0.8936,
+      "step": 8111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596390942527416e-06,
+      "loss": 0.938,
+      "step": 8112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.595799885766225e-06,
+      "loss": 0.8774,
+      "step": 8113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.59520877934329e-06,
+      "loss": 0.9077,
+      "step": 8114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594617623269922e-06,
+      "loss": 0.9341,
+      "step": 8115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594026417557432e-06,
+      "loss": 0.8828,
+      "step": 8116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.593435162217126e-06,
+      "loss": 0.8481,
+      "step": 8117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.592843857260322e-06,
+      "loss": 0.8926,
+      "step": 8118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.592252502698326e-06,
+      "loss": 0.916,
+      "step": 8119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.591661098542456e-06,
+      "loss": 0.8604,
+      "step": 8120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.591069644804025e-06,
+      "loss": 0.8008,
+      "step": 8121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.590478141494349e-06,
+      "loss": 0.8774,
+      "step": 8122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.589886588624744e-06,
+      "loss": 0.9478,
+      "step": 8123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5892949862065275e-06,
+      "loss": 0.9351,
+      "step": 8124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.588703334251018e-06,
+      "loss": 0.9229,
+      "step": 8125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.588111632769535e-06,
+      "loss": 0.897,
+      "step": 8126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5875198817733955e-06,
+      "loss": 0.8599,
+      "step": 8127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.586928081273925e-06,
+      "loss": 0.8384,
+      "step": 8128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.586336231282445e-06,
+      "loss": 0.8225,
+      "step": 8129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.585744331810276e-06,
+      "loss": 0.8882,
+      "step": 8130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.585152382868742e-06,
+      "loss": 0.876,
+      "step": 8131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.58456038446917e-06,
+      "loss": 0.9209,
+      "step": 8132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.583968336622887e-06,
+      "loss": 0.8755,
+      "step": 8133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.583376239341216e-06,
+      "loss": 0.8877,
+      "step": 8134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582784092635488e-06,
+      "loss": 0.8687,
+      "step": 8135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582191896517028e-06,
+      "loss": 0.9429,
+      "step": 8136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.581599650997171e-06,
+      "loss": 0.897,
+      "step": 8137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.581007356087242e-06,
+      "loss": 0.8237,
+      "step": 8138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5804150117985765e-06,
+      "loss": 0.9009,
+      "step": 8139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579822618142505e-06,
+      "loss": 0.937,
+      "step": 8140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579230175130362e-06,
+      "loss": 0.897,
+      "step": 8141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.578637682773482e-06,
+      "loss": 0.9927,
+      "step": 8142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.578045141083199e-06,
+      "loss": 0.8521,
+      "step": 8143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57745255007085e-06,
+      "loss": 0.896,
+      "step": 8144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576859909747771e-06,
+      "loss": 0.8857,
+      "step": 8145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576267220125304e-06,
+      "loss": 0.8267,
+      "step": 8146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575674481214782e-06,
+      "loss": 0.8677,
+      "step": 8147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575081693027551e-06,
+      "loss": 0.8496,
+      "step": 8148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.574488855574948e-06,
+      "loss": 0.8535,
+      "step": 8149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5738959688683166e-06,
+      "loss": 0.9014,
+      "step": 8150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.573303032919e-06,
+      "loss": 0.8638,
+      "step": 8151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.572710047738341e-06,
+      "loss": 0.9106,
+      "step": 8152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5721170133376854e-06,
+      "loss": 0.9272,
+      "step": 8153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.571523929728377e-06,
+      "loss": 0.8423,
+      "step": 8154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570930796921763e-06,
+      "loss": 0.8931,
+      "step": 8155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570337614929193e-06,
+      "loss": 0.8857,
+      "step": 8156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569744383762014e-06,
+      "loss": 0.874,
+      "step": 8157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569151103431576e-06,
+      "loss": 0.8257,
+      "step": 8158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.568557773949227e-06,
+      "loss": 0.8511,
+      "step": 8159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567964395326322e-06,
+      "loss": 0.9146,
+      "step": 8160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.56737096757421e-06,
+      "loss": 0.8574,
+      "step": 8161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5667774907042465e-06,
+      "loss": 0.9175,
+      "step": 8162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.566183964727787e-06,
+      "loss": 0.8916,
+      "step": 8163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.565590389656181e-06,
+      "loss": 0.8486,
+      "step": 8164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.564996765500789e-06,
+      "loss": 0.916,
+      "step": 8165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.564403092272967e-06,
+      "loss": 0.9116,
+      "step": 8166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.563809369984074e-06,
+      "loss": 0.8848,
+      "step": 8167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.563215598645467e-06,
+      "loss": 0.8369,
+      "step": 8168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562621778268506e-06,
+      "loss": 0.894,
+      "step": 8169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562027908864553e-06,
+      "loss": 0.9175,
+      "step": 8170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.561433990444968e-06,
+      "loss": 0.8882,
+      "step": 8171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5608400230211145e-06,
+      "loss": 0.8989,
+      "step": 8172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5602460066043574e-06,
+      "loss": 0.9121,
+      "step": 8173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5596519412060576e-06,
+      "loss": 0.8628,
+      "step": 8174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.559057826837585e-06,
+      "loss": 0.8374,
+      "step": 8175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.558463663510301e-06,
+      "loss": 0.916,
+      "step": 8176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.557869451235576e-06,
+      "loss": 0.8818,
+      "step": 8177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5572751900247796e-06,
+      "loss": 0.8774,
+      "step": 8178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556680879889278e-06,
+      "loss": 0.9106,
+      "step": 8179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556086520840442e-06,
+      "loss": 0.8735,
+      "step": 8180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.555492112889644e-06,
+      "loss": 0.876,
+      "step": 8181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554897656048255e-06,
+      "loss": 0.8901,
+      "step": 8182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554303150327646e-06,
+      "loss": 0.8696,
+      "step": 8183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.553708595739193e-06,
+      "loss": 0.8525,
+      "step": 8184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.553113992294272e-06,
+      "loss": 0.9033,
+      "step": 8185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.552519340004256e-06,
+      "loss": 0.8325,
+      "step": 8186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.551924638880522e-06,
+      "loss": 0.8916,
+      "step": 8187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.551329888934451e-06,
+      "loss": 0.998,
+      "step": 8188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.550735090177416e-06,
+      "loss": 0.7725,
+      "step": 8189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.550140242620801e-06,
+      "loss": 0.9033,
+      "step": 8190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.549545346275984e-06,
+      "loss": 0.8657,
+      "step": 8191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548950401154347e-06,
+      "loss": 0.9102,
+      "step": 8192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54835540726727e-06,
+      "loss": 0.957,
+      "step": 8193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547760364626142e-06,
+      "loss": 0.8491,
+      "step": 8194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547165273242342e-06,
+      "loss": 0.3384,
+      "step": 8195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5465701331272555e-06,
+      "loss": 0.8682,
+      "step": 8196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54597494429227e-06,
+      "loss": 0.8931,
+      "step": 8197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.545379706748773e-06,
+      "loss": 0.873,
+      "step": 8198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544784420508151e-06,
+      "loss": 0.8999,
+      "step": 8199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544189085581792e-06,
+      "loss": 0.9019,
+      "step": 8200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.543593701981087e-06,
+      "loss": 0.8418,
+      "step": 8201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542998269717425e-06,
+      "loss": 0.9214,
+      "step": 8202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542402788802201e-06,
+      "loss": 0.7631,
+      "step": 8203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.541807259246805e-06,
+      "loss": 0.8779,
+      "step": 8204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54121168106263e-06,
+      "loss": 0.9229,
+      "step": 8205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540616054261071e-06,
+      "loss": 0.9395,
+      "step": 8206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540020378853523e-06,
+      "loss": 0.9014,
+      "step": 8207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.539424654851384e-06,
+      "loss": 0.8706,
+      "step": 8208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.538828882266048e-06,
+      "loss": 0.9194,
+      "step": 8209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.538233061108916e-06,
+      "loss": 0.9141,
+      "step": 8210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.537637191391387e-06,
+      "loss": 0.8633,
+      "step": 8211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.537041273124858e-06,
+      "loss": 0.8149,
+      "step": 8212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53644530632073e-06,
+      "loss": 0.9492,
+      "step": 8213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535849290990409e-06,
+      "loss": 0.8604,
+      "step": 8214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535253227145294e-06,
+      "loss": 0.873,
+      "step": 8215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5346571147967905e-06,
+      "loss": 0.8657,
+      "step": 8216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.534060953956302e-06,
+      "loss": 0.9111,
+      "step": 8217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.533464744635232e-06,
+      "loss": 0.895,
+      "step": 8218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.532868486844993e-06,
+      "loss": 0.8926,
+      "step": 8219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5322721805969864e-06,
+      "loss": 0.8496,
+      "step": 8220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.531675825902624e-06,
+      "loss": 0.8809,
+      "step": 8221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5310794227733105e-06,
+      "loss": 0.8579,
+      "step": 8222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.530482971220461e-06,
+      "loss": 0.8804,
+      "step": 8223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529886471255487e-06,
+      "loss": 0.8804,
+      "step": 8224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529289922889795e-06,
+      "loss": 0.8877,
+      "step": 8225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.528693326134801e-06,
+      "loss": 0.9404,
+      "step": 8226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.528096681001922e-06,
+      "loss": 0.9365,
+      "step": 8227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527499987502567e-06,
+      "loss": 0.8799,
+      "step": 8228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.526903245648156e-06,
+      "loss": 0.8633,
+      "step": 8229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.526306455450103e-06,
+      "loss": 0.8945,
+      "step": 8230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.525709616919826e-06,
+      "loss": 0.8667,
+      "step": 8231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.525112730068745e-06,
+      "loss": 0.8721,
+      "step": 8232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.524515794908278e-06,
+      "loss": 0.9385,
+      "step": 8233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523918811449846e-06,
+      "loss": 0.9546,
+      "step": 8234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5233217797048694e-06,
+      "loss": 0.8999,
+      "step": 8235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522724699684771e-06,
+      "loss": 0.8896,
+      "step": 8236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522127571400976e-06,
+      "loss": 0.9673,
+      "step": 8237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.521530394864902e-06,
+      "loss": 0.937,
+      "step": 8238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520933170087981e-06,
+      "loss": 0.8628,
+      "step": 8239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520335897081636e-06,
+      "loss": 0.8813,
+      "step": 8240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.519738575857294e-06,
+      "loss": 0.9775,
+      "step": 8241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.519141206426379e-06,
+      "loss": 0.8804,
+      "step": 8242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.518543788800328e-06,
+      "loss": 0.8994,
+      "step": 8243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.517946322990562e-06,
+      "loss": 0.8809,
+      "step": 8244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.517348809008516e-06,
+      "loss": 0.8296,
+      "step": 8245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516751246865621e-06,
+      "loss": 0.8584,
+      "step": 8246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516153636573309e-06,
+      "loss": 0.894,
+      "step": 8247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515555978143012e-06,
+      "loss": 0.9165,
+      "step": 8248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.514958271586165e-06,
+      "loss": 0.8687,
+      "step": 8249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5143605169142034e-06,
+      "loss": 0.8604,
+      "step": 8250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.513762714138564e-06,
+      "loss": 0.8481,
+      "step": 8251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.513164863270681e-06,
+      "loss": 0.8872,
+      "step": 8252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.512566964321995e-06,
+      "loss": 0.8691,
+      "step": 8253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.511969017303944e-06,
+      "loss": 0.8677,
+      "step": 8254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.511371022227967e-06,
+      "loss": 0.874,
+      "step": 8255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.510772979105503e-06,
+      "loss": 0.8604,
+      "step": 8256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.510174887947998e-06,
+      "loss": 0.875,
+      "step": 8257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.509576748766892e-06,
+      "loss": 0.873,
+      "step": 8258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508978561573625e-06,
+      "loss": 0.9414,
+      "step": 8259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508380326379649e-06,
+      "loss": 0.9146,
+      "step": 8260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507782043196401e-06,
+      "loss": 0.8862,
+      "step": 8261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507183712035333e-06,
+      "loss": 0.8594,
+      "step": 8262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506585332907887e-06,
+      "loss": 0.8691,
+      "step": 8263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505986905825517e-06,
+      "loss": 0.8843,
+      "step": 8264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505388430799667e-06,
+      "loss": 0.334,
+      "step": 8265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.504789907841788e-06,
+      "loss": 0.9224,
+      "step": 8266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.504191336963331e-06,
+      "loss": 0.9209,
+      "step": 8267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503592718175747e-06,
+      "loss": 0.9458,
+      "step": 8268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.502994051490491e-06,
+      "loss": 0.8599,
+      "step": 8269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.502395336919015e-06,
+      "loss": 0.8696,
+      "step": 8270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.50179657447277e-06,
+      "loss": 0.8911,
+      "step": 8271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.501197764163215e-06,
+      "loss": 0.916,
+      "step": 8272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500598906001808e-06,
+      "loss": 0.876,
+      "step": 8273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.8926,
+      "step": 8274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4994010461692545e-06,
+      "loss": 0.8677,
+      "step": 8275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498802044521029e-06,
+      "loss": 0.9639,
+      "step": 8276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498202995066781e-06,
+      "loss": 0.8579,
+      "step": 8277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497603897817975e-06,
+      "loss": 0.8623,
+      "step": 8278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497004752786068e-06,
+      "loss": 0.938,
+      "step": 8279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.496405559982527e-06,
+      "loss": 0.9189,
+      "step": 8280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.495806319418814e-06,
+      "loss": 0.9028,
+      "step": 8281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.495207031106393e-06,
+      "loss": 0.9385,
+      "step": 8282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.494607695056728e-06,
+      "loss": 0.8916,
+      "step": 8283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4940083112812886e-06,
+      "loss": 0.9258,
+      "step": 8284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.493408879791539e-06,
+      "loss": 0.9102,
+      "step": 8285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49280940059895e-06,
+      "loss": 0.8906,
+      "step": 8286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.492209873714989e-06,
+      "loss": 0.8936,
+      "step": 8287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.491610299151123e-06,
+      "loss": 0.8027,
+      "step": 8288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4910106769188284e-06,
+      "loss": 0.9185,
+      "step": 8289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.490411007029574e-06,
+      "loss": 0.8667,
+      "step": 8290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.489811289494832e-06,
+      "loss": 0.9404,
+      "step": 8291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4892115243260764e-06,
+      "loss": 0.8516,
+      "step": 8292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488611711534782e-06,
+      "loss": 0.8569,
+      "step": 8293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488011851132424e-06,
+      "loss": 0.8682,
+      "step": 8294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.487411943130478e-06,
+      "loss": 0.9209,
+      "step": 8295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4868119875404235e-06,
+      "loss": 0.8804,
+      "step": 8296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.486211984373735e-06,
+      "loss": 0.918,
+      "step": 8297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485611933641895e-06,
+      "loss": 0.8892,
+      "step": 8298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485011835356382e-06,
+      "loss": 0.917,
+      "step": 8299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.484411689528676e-06,
+      "loss": 0.8584,
+      "step": 8300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483811496170259e-06,
+      "loss": 0.8726,
+      "step": 8301
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483211255292615e-06,
+      "loss": 0.9644,
+      "step": 8302
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482610966907227e-06,
+      "loss": 0.877,
+      "step": 8303
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482010631025576e-06,
+      "loss": 0.9033,
+      "step": 8304
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.481410247659155e-06,
+      "loss": 0.9136,
+      "step": 8305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.480809816819443e-06,
+      "loss": 0.9004,
+      "step": 8306
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.48020933851793e-06,
+      "loss": 0.8442,
+      "step": 8307
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.479608812766104e-06,
+      "loss": 0.8579,
+      "step": 8308
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4790082395754545e-06,
+      "loss": 0.916,
+      "step": 8309
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4784076189574705e-06,
+      "loss": 0.9287,
+      "step": 8310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.477806950923644e-06,
+      "loss": 0.897,
+      "step": 8311
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.477206235485466e-06,
+      "loss": 0.8384,
+      "step": 8312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.476605472654428e-06,
+      "loss": 0.8638,
+      "step": 8313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.476004662442026e-06,
+      "loss": 0.8818,
+      "step": 8314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.475403804859752e-06,
+      "loss": 0.8896,
+      "step": 8315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474802899919103e-06,
+      "loss": 0.8721,
+      "step": 8316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474201947631576e-06,
+      "loss": 0.8994,
+      "step": 8317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.473600948008665e-06,
+      "loss": 0.8877,
+      "step": 8318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472999901061871e-06,
+      "loss": 0.8599,
+      "step": 8319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472398806802692e-06,
+      "loss": 0.7878,
+      "step": 8320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.471797665242626e-06,
+      "loss": 0.8774,
+      "step": 8321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4711964763931775e-06,
+      "loss": 0.8945,
+      "step": 8322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.470595240265845e-06,
+      "loss": 0.9058,
+      "step": 8323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.469993956872135e-06,
+      "loss": 0.8496,
+      "step": 8324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.469392626223546e-06,
+      "loss": 0.8965,
+      "step": 8325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468791248331585e-06,
+      "loss": 0.9155,
+      "step": 8326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468189823207757e-06,
+      "loss": 0.9644,
+      "step": 8327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4675883508635685e-06,
+      "loss": 0.8408,
+      "step": 8328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4669868313105265e-06,
+      "loss": 0.9214,
+      "step": 8329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.466385264560139e-06,
+      "loss": 0.9258,
+      "step": 8330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465783650623915e-06,
+      "loss": 0.8594,
+      "step": 8331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4651819895133645e-06,
+      "loss": 0.8892,
+      "step": 8332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.464580281239997e-06,
+      "loss": 0.8779,
+      "step": 8333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.463978525815325e-06,
+      "loss": 0.9243,
+      "step": 8334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.463376723250862e-06,
+      "loss": 0.854,
+      "step": 8335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.462774873558121e-06,
+      "loss": 0.9048,
+      "step": 8336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4621729767486125e-06,
+      "loss": 0.8833,
+      "step": 8337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.461571032833859e-06,
+      "loss": 0.8188,
+      "step": 8338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.460969041825371e-06,
+      "loss": 0.8965,
+      "step": 8339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.460367003734667e-06,
+      "loss": 0.9219,
+      "step": 8340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459764918573264e-06,
+      "loss": 0.8481,
+      "step": 8341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459162786352685e-06,
+      "loss": 0.8306,
+      "step": 8342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.458560607084444e-06,
+      "loss": 0.8379,
+      "step": 8343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457958380780066e-06,
+      "loss": 0.9111,
+      "step": 8344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4573561074510705e-06,
+      "loss": 0.915,
+      "step": 8345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.45675378710898e-06,
+      "loss": 0.9214,
+      "step": 8346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.456151419765319e-06,
+      "loss": 0.9082,
+      "step": 8347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4555490054316105e-06,
+      "loss": 0.8872,
+      "step": 8348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.454946544119379e-06,
+      "loss": 0.9673,
+      "step": 8349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.454344035840153e-06,
+      "loss": 0.9043,
+      "step": 8350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453741480605458e-06,
+      "loss": 0.9414,
+      "step": 8351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453138878426822e-06,
+      "loss": 0.9272,
+      "step": 8352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.452536229315773e-06,
+      "loss": 0.9014,
+      "step": 8353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451933533283842e-06,
+      "loss": 0.8945,
+      "step": 8354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451330790342557e-06,
+      "loss": 0.9258,
+      "step": 8355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.450728000503452e-06,
+      "loss": 0.8818,
+      "step": 8356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.450125163778059e-06,
+      "loss": 0.8745,
+      "step": 8357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.44952228017791e-06,
+      "loss": 0.8823,
+      "step": 8358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.448919349714541e-06,
+      "loss": 0.9785,
+      "step": 8359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4483163723994845e-06,
+      "loss": 0.874,
+      "step": 8360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.447713348244279e-06,
+      "loss": 0.8857,
+      "step": 8361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.447110277260458e-06,
+      "loss": 0.8633,
+      "step": 8362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.446507159459562e-06,
+      "loss": 0.8735,
+      "step": 8363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445903994853131e-06,
+      "loss": 0.8711,
+      "step": 8364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445300783452701e-06,
+      "loss": 0.8579,
+      "step": 8365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444697525269812e-06,
+      "loss": 0.8643,
+      "step": 8366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444094220316007e-06,
+      "loss": 0.8755,
+      "step": 8367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.44349086860283e-06,
+      "loss": 0.8799,
+      "step": 8368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442887470141822e-06,
+      "loss": 0.8647,
+      "step": 8369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442284024944526e-06,
+      "loss": 0.9233,
+      "step": 8370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.44168053302249e-06,
+      "loss": 0.8779,
+      "step": 8371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4410769943872575e-06,
+      "loss": 0.8931,
+      "step": 8372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.440473409050374e-06,
+      "loss": 0.8579,
+      "step": 8373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.439869777023389e-06,
+      "loss": 0.957,
+      "step": 8374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.43926609831785e-06,
+      "loss": 0.9834,
+      "step": 8375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438662372945308e-06,
+      "loss": 0.8359,
+      "step": 8376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.43805860091731e-06,
+      "loss": 0.811,
+      "step": 8377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.437454782245411e-06,
+      "loss": 0.8721,
+      "step": 8378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4368509169411605e-06,
+      "loss": 0.8916,
+      "step": 8379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.436247005016112e-06,
+      "loss": 0.9062,
+      "step": 8380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4356430464818195e-06,
+      "loss": 0.8354,
+      "step": 8381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.435039041349836e-06,
+      "loss": 0.8623,
+      "step": 8382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434434989631719e-06,
+      "loss": 0.8896,
+      "step": 8383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.433830891339025e-06,
+      "loss": 0.3719,
+      "step": 8384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4332267464833116e-06,
+      "loss": 0.8364,
+      "step": 8385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432622555076135e-06,
+      "loss": 0.8696,
+      "step": 8386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432018317129056e-06,
+      "loss": 0.8579,
+      "step": 8387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.431414032653635e-06,
+      "loss": 0.8999,
+      "step": 8388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.43080970166143e-06,
+      "loss": 0.9282,
+      "step": 8389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430205324164006e-06,
+      "loss": 0.873,
+      "step": 8390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4296009001729254e-06,
+      "loss": 0.9565,
+      "step": 8391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428996429699751e-06,
+      "loss": 0.8604,
+      "step": 8392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428391912756046e-06,
+      "loss": 0.897,
+      "step": 8393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.42778734935338e-06,
+      "loss": 0.8384,
+      "step": 8394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.427182739503311e-06,
+      "loss": 0.9336,
+      "step": 8395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.426578083217415e-06,
+      "loss": 0.8765,
+      "step": 8396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425973380507257e-06,
+      "loss": 0.9165,
+      "step": 8397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425368631384404e-06,
+      "loss": 0.8926,
+      "step": 8398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.424763835860427e-06,
+      "loss": 0.8354,
+      "step": 8399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.424158993946897e-06,
+      "loss": 0.8975,
+      "step": 8400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.423554105655386e-06,
+      "loss": 0.9077,
+      "step": 8401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422949170997467e-06,
+      "loss": 0.9067,
+      "step": 8402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422344189984711e-06,
+      "loss": 0.8838,
+      "step": 8403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.421739162628692e-06,
+      "loss": 0.8896,
+      "step": 8404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.421134088940989e-06,
+      "loss": 0.8188,
+      "step": 8405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420528968933174e-06,
+      "loss": 0.8579,
+      "step": 8406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.419923802616826e-06,
+      "loss": 0.8623,
+      "step": 8407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.419318590003524e-06,
+      "loss": 0.7582,
+      "step": 8408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.418713331104843e-06,
+      "loss": 0.8726,
+      "step": 8409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4181080259323655e-06,
+      "loss": 0.874,
+      "step": 8410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41750267449767e-06,
+      "loss": 0.8989,
+      "step": 8411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.416897276812339e-06,
+      "loss": 0.8257,
+      "step": 8412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.416291832887956e-06,
+      "loss": 0.8838,
+      "step": 8413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.415686342736101e-06,
+      "loss": 0.8296,
+      "step": 8414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41508080636836e-06,
+      "loss": 0.8804,
+      "step": 8415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.414475223796318e-06,
+      "loss": 0.8936,
+      "step": 8416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413869595031561e-06,
+      "loss": 0.8955,
+      "step": 8417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413263920085673e-06,
+      "loss": 0.8335,
+      "step": 8418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.412658198970244e-06,
+      "loss": 0.3473,
+      "step": 8419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.412052431696862e-06,
+      "loss": 0.9355,
+      "step": 8420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411446618277116e-06,
+      "loss": 0.877,
+      "step": 8421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.410840758722597e-06,
+      "loss": 0.9824,
+      "step": 8422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.410234853044895e-06,
+      "loss": 0.8389,
+      "step": 8423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.409628901255601e-06,
+      "loss": 0.8799,
+      "step": 8424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.40902290336631e-06,
+      "loss": 0.8501,
+      "step": 8425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.408416859388614e-06,
+      "loss": 0.8774,
+      "step": 8426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.40781076933411e-06,
+      "loss": 0.8779,
+      "step": 8427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4072046332143885e-06,
+      "loss": 0.8984,
+      "step": 8428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.406598451041051e-06,
+      "loss": 0.9062,
+      "step": 8429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405992222825695e-06,
+      "loss": 0.9341,
+      "step": 8430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405385948579914e-06,
+      "loss": 0.9043,
+      "step": 8431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.404779628315308e-06,
+      "loss": 0.8784,
+      "step": 8432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4041732620434816e-06,
+      "loss": 0.8594,
+      "step": 8433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4035668497760305e-06,
+      "loss": 0.9023,
+      "step": 8434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402960391524558e-06,
+      "loss": 0.327,
+      "step": 8435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402353887300666e-06,
+      "loss": 0.8628,
+      "step": 8436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4017473371159595e-06,
+      "loss": 0.9004,
+      "step": 8437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.401140740982042e-06,
+      "loss": 0.8208,
+      "step": 8438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400534098910517e-06,
+      "loss": 0.9023,
+      "step": 8439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3999274109129924e-06,
+      "loss": 0.9351,
+      "step": 8440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3993206770010766e-06,
+      "loss": 0.9229,
+      "step": 8441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3987138971863726e-06,
+      "loss": 0.9155,
+      "step": 8442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.398107071480493e-06,
+      "loss": 0.9219,
+      "step": 8443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3975001998950465e-06,
+      "loss": 0.915,
+      "step": 8444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.396893282441643e-06,
+      "loss": 0.8716,
+      "step": 8445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.396286319131895e-06,
+      "loss": 0.8765,
+      "step": 8446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.395679309977411e-06,
+      "loss": 0.8979,
+      "step": 8447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.395072254989808e-06,
+      "loss": 0.3645,
+      "step": 8448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.394465154180699e-06,
+      "loss": 0.9395,
+      "step": 8449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393858007561698e-06,
+      "loss": 0.8657,
+      "step": 8450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393250815144422e-06,
+      "loss": 0.8721,
+      "step": 8451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.392643576940486e-06,
+      "loss": 0.9634,
+      "step": 8452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.392036292961507e-06,
+      "loss": 0.9478,
+      "step": 8453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.391428963219105e-06,
+      "loss": 0.8755,
+      "step": 8454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390821587724899e-06,
+      "loss": 0.8389,
+      "step": 8455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390214166490509e-06,
+      "loss": 0.8052,
+      "step": 8456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.389606699527554e-06,
+      "loss": 0.8662,
+      "step": 8457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.388999186847658e-06,
+      "loss": 0.8745,
+      "step": 8458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3883916284624435e-06,
+      "loss": 0.9204,
+      "step": 8459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3877840243835334e-06,
+      "loss": 0.9097,
+      "step": 8460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.387176374622551e-06,
+      "loss": 0.894,
+      "step": 8461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.386568679191124e-06,
+      "loss": 0.8608,
+      "step": 8462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.385960938100877e-06,
+      "loss": 0.8931,
+      "step": 8463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.385353151363438e-06,
+      "loss": 0.8457,
+      "step": 8464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.384745318990434e-06,
+      "loss": 0.8911,
+      "step": 8465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.384137440993493e-06,
+      "loss": 0.8765,
+      "step": 8466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.383529517384247e-06,
+      "loss": 0.9385,
+      "step": 8467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382921548174325e-06,
+      "loss": 0.8701,
+      "step": 8468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3823135333753596e-06,
+      "loss": 0.9634,
+      "step": 8469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.381705472998981e-06,
+      "loss": 0.918,
+      "step": 8470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.381097367056824e-06,
+      "loss": 0.8857,
+      "step": 8471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3804892155605235e-06,
+      "loss": 0.8779,
+      "step": 8472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37988101852171e-06,
+      "loss": 0.8921,
+      "step": 8473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.379272775952026e-06,
+      "loss": 0.8223,
+      "step": 8474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3786644878631035e-06,
+      "loss": 0.9414,
+      "step": 8475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3780561542665804e-06,
+      "loss": 0.8833,
+      "step": 8476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.377447775174095e-06,
+      "loss": 0.9077,
+      "step": 8477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.376839350597288e-06,
+      "loss": 0.875,
+      "step": 8478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3762308805478e-06,
+      "loss": 0.8843,
+      "step": 8479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.375622365037269e-06,
+      "loss": 0.8608,
+      "step": 8480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3750138040773386e-06,
+      "loss": 0.9741,
+      "step": 8481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3744051976796535e-06,
+      "loss": 0.9214,
+      "step": 8482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373796545855853e-06,
+      "loss": 0.937,
+      "step": 8483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373187848617587e-06,
+      "loss": 0.9067,
+      "step": 8484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3725791059764954e-06,
+      "loss": 0.9111,
+      "step": 8485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.371970317944227e-06,
+      "loss": 0.8726,
+      "step": 8486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.371361484532429e-06,
+      "loss": 0.8574,
+      "step": 8487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370752605752749e-06,
+      "loss": 0.8657,
+      "step": 8488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370143681616836e-06,
+      "loss": 0.9419,
+      "step": 8489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3695347121363385e-06,
+      "loss": 0.894,
+      "step": 8490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.368925697322909e-06,
+      "loss": 0.9019,
+      "step": 8491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3683166371881975e-06,
+      "loss": 0.8823,
+      "step": 8492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367707531743858e-06,
+      "loss": 0.894,
+      "step": 8493
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367098381001539e-06,
+      "loss": 0.8833,
+      "step": 8494
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.366489184972902e-06,
+      "loss": 0.9072,
+      "step": 8495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365879943669596e-06,
+      "loss": 0.938,
+      "step": 8496
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365270657103278e-06,
+      "loss": 0.8442,
+      "step": 8497
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364661325285606e-06,
+      "loss": 0.8862,
+      "step": 8498
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364051948228236e-06,
+      "loss": 0.8887,
+      "step": 8499
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.363442525942827e-06,
+      "loss": 0.3132,
+      "step": 8500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362833058441038e-06,
+      "loss": 0.9121,
+      "step": 8501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362223545734531e-06,
+      "loss": 0.8901,
+      "step": 8502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.361613987834963e-06,
+      "loss": 0.9282,
+      "step": 8503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.361004384753998e-06,
+      "loss": 0.8818,
+      "step": 8504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3603947365033005e-06,
+      "loss": 0.8804,
+      "step": 8505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359785043094529e-06,
+      "loss": 0.8623,
+      "step": 8506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359175304539354e-06,
+      "loss": 0.8848,
+      "step": 8507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.358565520849438e-06,
+      "loss": 0.9468,
+      "step": 8508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357955692036445e-06,
+      "loss": 0.9219,
+      "step": 8509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357345818112044e-06,
+      "loss": 0.8896,
+      "step": 8510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356735899087904e-06,
+      "loss": 0.9448,
+      "step": 8511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356125934975693e-06,
+      "loss": 0.8501,
+      "step": 8512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.355515925787079e-06,
+      "loss": 0.874,
+      "step": 8513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.354905871533733e-06,
+      "loss": 0.895,
+      "step": 8514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.354295772227328e-06,
+      "loss": 0.9053,
+      "step": 8515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3536856278795345e-06,
+      "loss": 0.936,
+      "step": 8516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.353075438502025e-06,
+      "loss": 0.8936,
+      "step": 8517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.352465204106476e-06,
+      "loss": 0.9482,
+      "step": 8518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351854924704559e-06,
+      "loss": 0.8765,
+      "step": 8519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3512446003079516e-06,
+      "loss": 0.9229,
+      "step": 8520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3506342309283314e-06,
+      "loss": 0.8579,
+      "step": 8521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.350023816577374e-06,
+      "loss": 0.8721,
+      "step": 8522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.349413357266756e-06,
+      "loss": 0.9678,
+      "step": 8523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.34880285300816e-06,
+      "loss": 0.8853,
+      "step": 8524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.348192303813264e-06,
+      "loss": 0.9229,
+      "step": 8525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.347581709693747e-06,
+      "loss": 0.876,
+      "step": 8526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346971070661295e-06,
+      "loss": 0.874,
+      "step": 8527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346360386727587e-06,
+      "loss": 0.8735,
+      "step": 8528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3457496579043075e-06,
+      "loss": 0.8433,
+      "step": 8529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.34513888420314e-06,
+      "loss": 0.9604,
+      "step": 8530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344528065635771e-06,
+      "loss": 0.9185,
+      "step": 8531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343917202213885e-06,
+      "loss": 0.877,
+      "step": 8532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343306293949169e-06,
+      "loss": 0.8789,
+      "step": 8533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.342695340853313e-06,
+      "loss": 0.96,
+      "step": 8534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3420843429380015e-06,
+      "loss": 0.8945,
+      "step": 8535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.341473300214926e-06,
+      "loss": 0.8574,
+      "step": 8536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3408622126957775e-06,
+      "loss": 0.9136,
+      "step": 8537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340251080392247e-06,
+      "loss": 0.9258,
+      "step": 8538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3396399033160225e-06,
+      "loss": 0.9116,
+      "step": 8539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.339028681478801e-06,
+      "loss": 0.8975,
+      "step": 8540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.338417414892277e-06,
+      "loss": 0.9331,
+      "step": 8541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3378061035681415e-06,
+      "loss": 0.9434,
+      "step": 8542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.337194747518091e-06,
+      "loss": 0.9204,
+      "step": 8543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.336583346753823e-06,
+      "loss": 0.9258,
+      "step": 8544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.335971901287033e-06,
+      "loss": 0.9248,
+      "step": 8545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.33536041112942e-06,
+      "loss": 0.823,
+      "step": 8546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.334748876292681e-06,
+      "loss": 0.8916,
+      "step": 8547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.334137296788517e-06,
+      "loss": 0.9121,
+      "step": 8548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.33352567262863e-06,
+      "loss": 0.9126,
+      "step": 8549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3329140038247185e-06,
+      "loss": 0.8623,
+      "step": 8550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3323022903884854e-06,
+      "loss": 0.877,
+      "step": 8551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.331690532331634e-06,
+      "loss": 0.9233,
+      "step": 8552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.331078729665871e-06,
+      "loss": 0.853,
+      "step": 8553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.330466882402897e-06,
+      "loss": 0.8647,
+      "step": 8554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.329854990554419e-06,
+      "loss": 0.915,
+      "step": 8555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.329243054132143e-06,
+      "loss": 0.8843,
+      "step": 8556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3286310731477784e-06,
+      "loss": 0.8223,
+      "step": 8557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328019047613031e-06,
+      "loss": 0.8862,
+      "step": 8558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3274069775396105e-06,
+      "loss": 0.9209,
+      "step": 8559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3267948629392275e-06,
+      "loss": 0.8882,
+      "step": 8560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.326182703823592e-06,
+      "loss": 0.9067,
+      "step": 8561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.325570500204415e-06,
+      "loss": 0.8643,
+      "step": 8562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3249582520934116e-06,
+      "loss": 0.8926,
+      "step": 8563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.324345959502291e-06,
+      "loss": 0.9092,
+      "step": 8564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323733622442771e-06,
+      "loss": 0.8765,
+      "step": 8565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323121240926565e-06,
+      "loss": 0.9116,
+      "step": 8566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.322508814965389e-06,
+      "loss": 0.9688,
+      "step": 8567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321896344570958e-06,
+      "loss": 0.9434,
+      "step": 8568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321283829754993e-06,
+      "loss": 0.9741,
+      "step": 8569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3206712705292095e-06,
+      "loss": 0.8989,
+      "step": 8570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.320058666905328e-06,
+      "loss": 0.8232,
+      "step": 8571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319446018895067e-06,
+      "loss": 0.9424,
+      "step": 8572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.318833326510151e-06,
+      "loss": 0.8916,
+      "step": 8573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.318220589762297e-06,
+      "loss": 0.9053,
+      "step": 8574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3176078086632314e-06,
+      "loss": 0.8345,
+      "step": 8575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.316994983224677e-06,
+      "loss": 0.894,
+      "step": 8576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3163821134583555e-06,
+      "loss": 0.8643,
+      "step": 8577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.315769199375995e-06,
+      "loss": 0.9253,
+      "step": 8578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3151562409893204e-06,
+      "loss": 0.8926,
+      "step": 8579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.314543238310059e-06,
+      "loss": 0.9077,
+      "step": 8580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3139301913499384e-06,
+      "loss": 0.9292,
+      "step": 8581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.313317100120686e-06,
+      "loss": 0.8979,
+      "step": 8582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.312703964634034e-06,
+      "loss": 0.8809,
+      "step": 8583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.312090784901708e-06,
+      "loss": 0.9102,
+      "step": 8584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311477560935443e-06,
+      "loss": 0.8486,
+      "step": 8585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.310864292746971e-06,
+      "loss": 0.8589,
+      "step": 8586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3102509803480236e-06,
+      "loss": 0.8867,
+      "step": 8587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309637623750333e-06,
+      "loss": 0.813,
+      "step": 8588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309024222965635e-06,
+      "loss": 0.8608,
+      "step": 8589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.308410778005667e-06,
+      "loss": 0.4562,
+      "step": 8590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.307797288882161e-06,
+      "loss": 0.895,
+      "step": 8591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.307183755606857e-06,
+      "loss": 0.9385,
+      "step": 8592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.306570178191492e-06,
+      "loss": 0.8828,
+      "step": 8593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.305956556647806e-06,
+      "loss": 0.9189,
+      "step": 8594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.305342890987537e-06,
+      "loss": 0.8193,
+      "step": 8595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3047291812224245e-06,
+      "loss": 0.9219,
+      "step": 8596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.304115427364211e-06,
+      "loss": 0.9141,
+      "step": 8597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.30350162942464e-06,
+      "loss": 0.9258,
+      "step": 8598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.302887787415452e-06,
+      "loss": 0.916,
+      "step": 8599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.302273901348393e-06,
+      "loss": 0.9014,
+      "step": 8600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.301659971235203e-06,
+      "loss": 0.8389,
+      "step": 8601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.301045997087633e-06,
+      "loss": 0.8525,
+      "step": 8602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.300431978917428e-06,
+      "loss": 0.5098,
+      "step": 8603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2998179167363335e-06,
+      "loss": 0.8569,
+      "step": 8604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299203810556097e-06,
+      "loss": 0.9302,
+      "step": 8605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2985896603884706e-06,
+      "loss": 0.8262,
+      "step": 8606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297975466245201e-06,
+      "loss": 0.9116,
+      "step": 8607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2973612281380394e-06,
+      "loss": 0.8892,
+      "step": 8608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.296746946078737e-06,
+      "loss": 0.9233,
+      "step": 8609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2961326200790465e-06,
+      "loss": 0.9219,
+      "step": 8610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295518250150721e-06,
+      "loss": 0.874,
+      "step": 8611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.294903836305514e-06,
+      "loss": 0.8916,
+      "step": 8612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.294289378555179e-06,
+      "loss": 0.8853,
+      "step": 8613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2936748769114744e-06,
+      "loss": 0.9365,
+      "step": 8614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.293060331386154e-06,
+      "loss": 0.8965,
+      "step": 8615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.292445741990977e-06,
+      "loss": 0.8545,
+      "step": 8616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291831108737699e-06,
+      "loss": 0.8638,
+      "step": 8617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2912164316380815e-06,
+      "loss": 0.894,
+      "step": 8618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.290601710703885e-06,
+      "loss": 0.8857,
+      "step": 8619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289986945946866e-06,
+      "loss": 0.855,
+      "step": 8620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289372137378788e-06,
+      "loss": 0.9209,
+      "step": 8621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.288757285011414e-06,
+      "loss": 0.8555,
+      "step": 8622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.288142388856508e-06,
+      "loss": 0.8745,
+      "step": 8623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.28752744892583e-06,
+      "loss": 0.8555,
+      "step": 8624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286912465231149e-06,
+      "loss": 0.9526,
+      "step": 8625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286297437784226e-06,
+      "loss": 0.8589,
+      "step": 8626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.285682366596833e-06,
+      "loss": 0.9058,
+      "step": 8627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.285067251680734e-06,
+      "loss": 0.9189,
+      "step": 8628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284452093047698e-06,
+      "loss": 0.8867,
+      "step": 8629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.283836890709493e-06,
+      "loss": 0.9595,
+      "step": 8630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.283221644677889e-06,
+      "loss": 0.9287,
+      "step": 8631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.282606354964658e-06,
+      "loss": 0.3827,
+      "step": 8632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.281991021581569e-06,
+      "loss": 0.8857,
+      "step": 8633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.281375644540396e-06,
+      "loss": 0.3383,
+      "step": 8634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280760223852914e-06,
+      "loss": 0.918,
+      "step": 8635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280144759530894e-06,
+      "loss": 0.9355,
+      "step": 8636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27952925158611e-06,
+      "loss": 0.8384,
+      "step": 8637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.278913700030341e-06,
+      "loss": 0.9175,
+      "step": 8638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.278298104875362e-06,
+      "loss": 0.8994,
+      "step": 8639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.277682466132949e-06,
+      "loss": 0.9058,
+      "step": 8640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.277066783814881e-06,
+      "loss": 0.9126,
+      "step": 8641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2764510579329385e-06,
+      "loss": 0.8765,
+      "step": 8642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.275835288498899e-06,
+      "loss": 0.8701,
+      "step": 8643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.275219475524545e-06,
+      "loss": 0.896,
+      "step": 8644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.274603619021657e-06,
+      "loss": 0.8779,
+      "step": 8645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.273987719002017e-06,
+      "loss": 0.9053,
+      "step": 8646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.273371775477408e-06,
+      "loss": 0.8896,
+      "step": 8647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2727557884596155e-06,
+      "loss": 0.8571,
+      "step": 8648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.272139757960424e-06,
+      "loss": 0.9307,
+      "step": 8649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.271523683991617e-06,
+      "loss": 0.8486,
+      "step": 8650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270907566564983e-06,
+      "loss": 0.8481,
+      "step": 8651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270291405692309e-06,
+      "loss": 0.8691,
+      "step": 8652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.269675201385383e-06,
+      "loss": 0.8862,
+      "step": 8653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.269058953655993e-06,
+      "loss": 0.9038,
+      "step": 8654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.268442662515931e-06,
+      "loss": 0.9009,
+      "step": 8655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.267826327976986e-06,
+      "loss": 0.897,
+      "step": 8656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2672099500509486e-06,
+      "loss": 0.9175,
+      "step": 8657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2665935287496126e-06,
+      "loss": 0.8765,
+      "step": 8658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.26597706408477e-06,
+      "loss": 0.8682,
+      "step": 8659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.265360556068216e-06,
+      "loss": 0.8281,
+      "step": 8660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264744004711746e-06,
+      "loss": 0.9199,
+      "step": 8661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264127410027153e-06,
+      "loss": 0.9048,
+      "step": 8662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.263510772026234e-06,
+      "loss": 0.8477,
+      "step": 8663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262894090720789e-06,
+      "loss": 0.8472,
+      "step": 8664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262277366122614e-06,
+      "loss": 0.8643,
+      "step": 8665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2616605982435074e-06,
+      "loss": 0.8325,
+      "step": 8666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2610437870952685e-06,
+      "loss": 0.8062,
+      "step": 8667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2604269326897015e-06,
+      "loss": 0.8896,
+      "step": 8668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259810035038604e-06,
+      "loss": 0.8472,
+      "step": 8669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.25919309415378e-06,
+      "loss": 0.939,
+      "step": 8670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.258576110047031e-06,
+      "loss": 0.8955,
+      "step": 8671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.257959082730163e-06,
+      "loss": 0.8955,
+      "step": 8672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.257342012214978e-06,
+      "loss": 0.8843,
+      "step": 8673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256724898513286e-06,
+      "loss": 0.8394,
+      "step": 8674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256107741636888e-06,
+      "loss": 0.8628,
+      "step": 8675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.255490541597594e-06,
+      "loss": 0.9385,
+      "step": 8676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254873298407212e-06,
+      "loss": 0.9253,
+      "step": 8677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254256012077553e-06,
+      "loss": 0.8857,
+      "step": 8678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.253638682620421e-06,
+      "loss": 0.8477,
+      "step": 8679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.253021310047631e-06,
+      "loss": 0.8638,
+      "step": 8680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.252403894370994e-06,
+      "loss": 0.8857,
+      "step": 8681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251786435602319e-06,
+      "loss": 0.894,
+      "step": 8682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251168933753422e-06,
+      "loss": 0.8779,
+      "step": 8683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.250551388836117e-06,
+      "loss": 0.8867,
+      "step": 8684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2499338008622165e-06,
+      "loss": 0.897,
+      "step": 8685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249316169843536e-06,
+      "loss": 0.874,
+      "step": 8686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2486984957918935e-06,
+      "loss": 0.8892,
+      "step": 8687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.248080778719105e-06,
+      "loss": 0.8545,
+      "step": 8688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24746301863699e-06,
+      "loss": 0.9712,
+      "step": 8689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2468452155573655e-06,
+      "loss": 0.8691,
+      "step": 8690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2462273694920514e-06,
+      "loss": 0.9019,
+      "step": 8691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.245609480452866e-06,
+      "loss": 0.9087,
+      "step": 8692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244991548451634e-06,
+      "loss": 0.8472,
+      "step": 8693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244373573500177e-06,
+      "loss": 0.9058,
+      "step": 8694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243755555610317e-06,
+      "loss": 0.9043,
+      "step": 8695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243137494793875e-06,
+      "loss": 0.8462,
+      "step": 8696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2425193910626815e-06,
+      "loss": 0.8535,
+      "step": 8697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241901244428555e-06,
+      "loss": 0.9067,
+      "step": 8698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241283054903326e-06,
+      "loss": 0.9097,
+      "step": 8699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2406648224988216e-06,
+      "loss": 0.8564,
+      "step": 8700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2400465472268664e-06,
+      "loss": 0.8887,
+      "step": 8701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239428229099292e-06,
+      "loss": 0.8691,
+      "step": 8702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.238809868127926e-06,
+      "loss": 0.9546,
+      "step": 8703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2381914643246e-06,
+      "loss": 0.8921,
+      "step": 8704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.237573017701144e-06,
+      "loss": 0.8784,
+      "step": 8705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.236954528269391e-06,
+      "loss": 0.8916,
+      "step": 8706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.236335996041172e-06,
+      "loss": 0.9233,
+      "step": 8707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.235717421028321e-06,
+      "loss": 0.877,
+      "step": 8708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.235098803242673e-06,
+      "loss": 0.8926,
+      "step": 8709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.234480142696063e-06,
+      "loss": 0.9146,
+      "step": 8710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.233861439400327e-06,
+      "loss": 0.9292,
+      "step": 8711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.233242693367301e-06,
+      "loss": 0.8486,
+      "step": 8712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.232623904608823e-06,
+      "loss": 0.8979,
+      "step": 8713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.232005073136733e-06,
+      "loss": 0.9551,
+      "step": 8714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231386198962867e-06,
+      "loss": 0.8706,
+      "step": 8715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.230767282099067e-06,
+      "loss": 0.9634,
+      "step": 8716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.230148322557176e-06,
+      "loss": 0.8833,
+      "step": 8717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229529320349031e-06,
+      "loss": 0.8481,
+      "step": 8718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.228910275486477e-06,
+      "loss": 0.8359,
+      "step": 8719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.228291187981357e-06,
+      "loss": 0.8091,
+      "step": 8720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227672057845516e-06,
+      "loss": 0.8779,
+      "step": 8721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227052885090798e-06,
+      "loss": 0.8496,
+      "step": 8722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.226433669729048e-06,
+      "loss": 0.8931,
+      "step": 8723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.225814411772115e-06,
+      "loss": 0.8647,
+      "step": 8724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.225195111231843e-06,
+      "loss": 0.9619,
+      "step": 8725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2245757681200835e-06,
+      "loss": 0.9585,
+      "step": 8726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.223956382448684e-06,
+      "loss": 0.8491,
+      "step": 8727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.223336954229494e-06,
+      "loss": 0.8857,
+      "step": 8728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.222717483474363e-06,
+      "loss": 0.8794,
+      "step": 8729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.222097970195147e-06,
+      "loss": 0.8662,
+      "step": 8730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2214784144036934e-06,
+      "loss": 0.8926,
+      "step": 8731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.220858816111858e-06,
+      "loss": 0.814,
+      "step": 8732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2202391753314925e-06,
+      "loss": 0.9346,
+      "step": 8733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.219619492074454e-06,
+      "loss": 0.9355,
+      "step": 8734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.218999766352595e-06,
+      "loss": 0.342,
+      "step": 8735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.218379998177776e-06,
+      "loss": 0.8496,
+      "step": 8736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.217760187561851e-06,
+      "loss": 0.9419,
+      "step": 8737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.217140334516679e-06,
+      "loss": 0.9023,
+      "step": 8738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.216520439054119e-06,
+      "loss": 0.8911,
+      "step": 8739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.21590050118603e-06,
+      "loss": 0.9277,
+      "step": 8740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.215280520924273e-06,
+      "loss": 0.9209,
+      "step": 8741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.214660498280708e-06,
+      "loss": 0.2847,
+      "step": 8742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2140404332671986e-06,
+      "loss": 0.8755,
+      "step": 8743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.213420325895606e-06,
+      "loss": 0.9141,
+      "step": 8744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.212800176177794e-06,
+      "loss": 0.8892,
+      "step": 8745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.21217998412563e-06,
+      "loss": 0.8804,
+      "step": 8746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.211559749750976e-06,
+      "loss": 0.9331,
+      "step": 8747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.210939473065698e-06,
+      "loss": 0.8721,
+      "step": 8748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.210319154081664e-06,
+      "loss": 0.9331,
+      "step": 8749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209698792810742e-06,
+      "loss": 0.9092,
+      "step": 8750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2090783892648e-06,
+      "loss": 0.8828,
+      "step": 8751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.208457943455708e-06,
+      "loss": 0.9053,
+      "step": 8752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.207837455395335e-06,
+      "loss": 0.8608,
+      "step": 8753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.207216925095552e-06,
+      "loss": 0.9292,
+      "step": 8754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2065963525682304e-06,
+      "loss": 0.959,
+      "step": 8755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.205975737825246e-06,
+      "loss": 0.876,
+      "step": 8756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2053550808784665e-06,
+      "loss": 0.9224,
+      "step": 8757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.20473438173977e-06,
+      "loss": 0.8804,
+      "step": 8758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.204113640421031e-06,
+      "loss": 0.8442,
+      "step": 8759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203492856934126e-06,
+      "loss": 0.8574,
+      "step": 8760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202872031290929e-06,
+      "loss": 0.9175,
+      "step": 8761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202251163503318e-06,
+      "loss": 0.8955,
+      "step": 8762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.201630253583174e-06,
+      "loss": 0.9092,
+      "step": 8763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.201009301542372e-06,
+      "loss": 0.8828,
+      "step": 8764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2003883073927935e-06,
+      "loss": 0.877,
+      "step": 8765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.19976727114632e-06,
+      "loss": 0.8672,
+      "step": 8766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.199146192814833e-06,
+      "loss": 0.873,
+      "step": 8767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.198525072410214e-06,
+      "loss": 0.916,
+      "step": 8768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.197903909944345e-06,
+      "loss": 0.9336,
+      "step": 8769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.197282705429112e-06,
+      "loss": 0.8687,
+      "step": 8770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196661458876396e-06,
+      "loss": 0.917,
+      "step": 8771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196040170298086e-06,
+      "loss": 0.8779,
+      "step": 8772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.195418839706069e-06,
+      "loss": 0.917,
+      "step": 8773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.194797467112226e-06,
+      "loss": 0.8574,
+      "step": 8774
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1941760525284525e-06,
+      "loss": 0.8428,
+      "step": 8775
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193554595966632e-06,
+      "loss": 0.8535,
+      "step": 8776
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.192933097438656e-06,
+      "loss": 0.9033,
+      "step": 8777
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1923115569564126e-06,
+      "loss": 0.8862,
+      "step": 8778
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191689974531797e-06,
+      "loss": 0.8472,
+      "step": 8779
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191068350176696e-06,
+      "loss": 0.8667,
+      "step": 8780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.190446683903006e-06,
+      "loss": 0.8706,
+      "step": 8781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.189824975722619e-06,
+      "loss": 0.9194,
+      "step": 8782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1892032256474295e-06,
+      "loss": 0.8711,
+      "step": 8783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188581433689333e-06,
+      "loss": 0.9126,
+      "step": 8784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.187959599860225e-06,
+      "loss": 0.834,
+      "step": 8785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.187337724172001e-06,
+      "loss": 0.9526,
+      "step": 8786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.186715806636561e-06,
+      "loss": 0.8833,
+      "step": 8787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1860938472658006e-06,
+      "loss": 0.811,
+      "step": 8788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.185471846071621e-06,
+      "loss": 0.8662,
+      "step": 8789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.184849803065921e-06,
+      "loss": 0.8667,
+      "step": 8790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.184227718260599e-06,
+      "loss": 0.8862,
+      "step": 8791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.183605591667562e-06,
+      "loss": 0.7445,
+      "step": 8792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.182983423298708e-06,
+      "loss": 0.8936,
+      "step": 8793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.18236121316594e-06,
+      "loss": 0.8994,
+      "step": 8794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.181738961281164e-06,
+      "loss": 0.8198,
+      "step": 8795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.181116667656284e-06,
+      "loss": 0.8115,
+      "step": 8796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.180494332303205e-06,
+      "loss": 0.9722,
+      "step": 8797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1798719552338324e-06,
+      "loss": 0.8438,
+      "step": 8798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.179249536460075e-06,
+      "loss": 0.915,
+      "step": 8799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.17862707599384e-06,
+      "loss": 0.8882,
+      "step": 8800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1780045738470364e-06,
+      "loss": 0.8701,
+      "step": 8801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1773820300315735e-06,
+      "loss": 0.876,
+      "step": 8802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176759444559359e-06,
+      "loss": 0.8784,
+      "step": 8803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176136817442308e-06,
+      "loss": 0.8271,
+      "step": 8804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.175514148692331e-06,
+      "loss": 0.9453,
+      "step": 8805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174891438321338e-06,
+      "loss": 0.9658,
+      "step": 8806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174268686341246e-06,
+      "loss": 0.9238,
+      "step": 8807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173645892763968e-06,
+      "loss": 0.9062,
+      "step": 8808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173023057601418e-06,
+      "loss": 0.8799,
+      "step": 8809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172400180865514e-06,
+      "loss": 0.8687,
+      "step": 8810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.171777262568171e-06,
+      "loss": 0.8735,
+      "step": 8811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.171154302721305e-06,
+      "loss": 0.8989,
+      "step": 8812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.170531301336837e-06,
+      "loss": 0.8887,
+      "step": 8813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169908258426684e-06,
+      "loss": 0.8689,
+      "step": 8814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169285174002768e-06,
+      "loss": 0.8452,
+      "step": 8815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.168662048077007e-06,
+      "loss": 0.9038,
+      "step": 8816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1680388806613254e-06,
+      "loss": 0.897,
+      "step": 8817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.167415671767642e-06,
+      "loss": 0.9341,
+      "step": 8818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.166792421407882e-06,
+      "loss": 0.8921,
+      "step": 8819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1661691295939685e-06,
+      "loss": 0.9038,
+      "step": 8820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.165545796337826e-06,
+      "loss": 0.9185,
+      "step": 8821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164922421651382e-06,
+      "loss": 0.8638,
+      "step": 8822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164299005546558e-06,
+      "loss": 0.8525,
+      "step": 8823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.163675548035283e-06,
+      "loss": 0.9492,
+      "step": 8824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.163052049129486e-06,
+      "loss": 0.8916,
+      "step": 8825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.162428508841096e-06,
+      "loss": 0.917,
+      "step": 8826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.16180492718204e-06,
+      "loss": 0.8633,
+      "step": 8827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.161181304164248e-06,
+      "loss": 0.876,
+      "step": 8828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.160557639799655e-06,
+      "loss": 0.9478,
+      "step": 8829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.159933934100187e-06,
+      "loss": 0.876,
+      "step": 8830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.159310187077781e-06,
+      "loss": 0.8613,
+      "step": 8831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158686398744368e-06,
+      "loss": 0.9434,
+      "step": 8832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158062569111882e-06,
+      "loss": 0.4121,
+      "step": 8833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1574386981922585e-06,
+      "loss": 0.8989,
+      "step": 8834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.156814785997434e-06,
+      "loss": 0.9819,
+      "step": 8835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.156190832539342e-06,
+      "loss": 0.8975,
+      "step": 8836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1555668378299225e-06,
+      "loss": 0.8926,
+      "step": 8837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.154942801881114e-06,
+      "loss": 0.9229,
+      "step": 8838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1543187247048525e-06,
+      "loss": 0.8276,
+      "step": 8839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1536946063130785e-06,
+      "loss": 0.8521,
+      "step": 8840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.153070446717735e-06,
+      "loss": 0.8804,
+      "step": 8841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.15244624593076e-06,
+      "loss": 0.8599,
+      "step": 8842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1518220039640965e-06,
+      "loss": 0.8813,
+      "step": 8843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.151197720829685e-06,
+      "loss": 0.8882,
+      "step": 8844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.150573396539474e-06,
+      "loss": 0.8403,
+      "step": 8845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149949031105404e-06,
+      "loss": 0.9058,
+      "step": 8846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149324624539422e-06,
+      "loss": 0.8804,
+      "step": 8847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1487001768534726e-06,
+      "loss": 0.9116,
+      "step": 8848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.148075688059502e-06,
+      "loss": 0.8799,
+      "step": 8849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14745115816946e-06,
+      "loss": 0.9058,
+      "step": 8850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.146826587195293e-06,
+      "loss": 0.8735,
+      "step": 8851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1462019751489486e-06,
+      "loss": 0.8774,
+      "step": 8852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14557732204238e-06,
+      "loss": 0.8477,
+      "step": 8853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144952627887537e-06,
+      "loss": 0.8589,
+      "step": 8854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144327892696369e-06,
+      "loss": 0.2726,
+      "step": 8855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143703116480829e-06,
+      "loss": 0.856,
+      "step": 8856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143078299252871e-06,
+      "loss": 0.9038,
+      "step": 8857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1424534410244485e-06,
+      "loss": 0.8784,
+      "step": 8858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.141828541807515e-06,
+      "loss": 0.7837,
+      "step": 8859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.141203601614026e-06,
+      "loss": 0.876,
+      "step": 8860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.140578620455939e-06,
+      "loss": 0.8701,
+      "step": 8861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13995359834521e-06,
+      "loss": 0.349,
+      "step": 8862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1393285352937945e-06,
+      "loss": 0.8325,
+      "step": 8863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.138703431313656e-06,
+      "loss": 0.896,
+      "step": 8864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1380782864167475e-06,
+      "loss": 1.0127,
+      "step": 8865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1374531006150336e-06,
+      "loss": 0.8745,
+      "step": 8866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.136827873920475e-06,
+      "loss": 0.8945,
+      "step": 8867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13620260634503e-06,
+      "loss": 0.8965,
+      "step": 8868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.135577297900663e-06,
+      "loss": 0.873,
+      "step": 8869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134951948599338e-06,
+      "loss": 0.9634,
+      "step": 8870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134326558453017e-06,
+      "loss": 0.9067,
+      "step": 8871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.133701127473667e-06,
+      "loss": 0.8687,
+      "step": 8872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13307565567325e-06,
+      "loss": 0.9102,
+      "step": 8873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1324501430637375e-06,
+      "loss": 0.3376,
+      "step": 8874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.131824589657091e-06,
+      "loss": 0.8579,
+      "step": 8875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.131198995465281e-06,
+      "loss": 0.7998,
+      "step": 8876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.130573360500277e-06,
+      "loss": 0.8706,
+      "step": 8877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1299476847740465e-06,
+      "loss": 0.8179,
+      "step": 8878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.12932196829856e-06,
+      "loss": 0.9033,
+      "step": 8879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.128696211085791e-06,
+      "loss": 0.8613,
+      "step": 8880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.128070413147706e-06,
+      "loss": 0.8794,
+      "step": 8881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.127444574496282e-06,
+      "loss": 0.8813,
+      "step": 8882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1268186951434915e-06,
+      "loss": 0.9106,
+      "step": 8883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.126192775101308e-06,
+      "loss": 0.8838,
+      "step": 8884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1255668143817045e-06,
+      "loss": 0.918,
+      "step": 8885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.124940812996659e-06,
+      "loss": 0.9219,
+      "step": 8886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.124314770958148e-06,
+      "loss": 0.9072,
+      "step": 8887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123688688278148e-06,
+      "loss": 0.8633,
+      "step": 8888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123062564968636e-06,
+      "loss": 0.8809,
+      "step": 8889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.122436401041593e-06,
+      "loss": 0.9341,
+      "step": 8890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121810196508996e-06,
+      "loss": 0.8774,
+      "step": 8891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121183951382826e-06,
+      "loss": 0.8672,
+      "step": 8892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.120557665675065e-06,
+      "loss": 0.9399,
+      "step": 8893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119931339397694e-06,
+      "loss": 0.9497,
+      "step": 8894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119304972562696e-06,
+      "loss": 0.894,
+      "step": 8895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.118678565182055e-06,
+      "loss": 0.2902,
+      "step": 8896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.118052117267755e-06,
+      "loss": 0.8994,
+      "step": 8897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117425628831778e-06,
+      "loss": 0.8569,
+      "step": 8898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.116799099886114e-06,
+      "loss": 0.8721,
+      "step": 8899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1161725304427465e-06,
+      "loss": 0.8867,
+      "step": 8900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.115545920513665e-06,
+      "loss": 0.9111,
+      "step": 8901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.114919270110854e-06,
+      "loss": 0.9058,
+      "step": 8902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.114292579246307e-06,
+      "loss": 0.8501,
+      "step": 8903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1136658479320095e-06,
+      "loss": 0.8643,
+      "step": 8904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113039076179955e-06,
+      "loss": 0.8784,
+      "step": 8905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.112412264002133e-06,
+      "loss": 0.8867,
+      "step": 8906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.111785411410535e-06,
+      "loss": 0.9224,
+      "step": 8907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.111158518417155e-06,
+      "loss": 0.8579,
+      "step": 8908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.110531585033985e-06,
+      "loss": 0.9653,
+      "step": 8909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109904611273021e-06,
+      "loss": 0.8867,
+      "step": 8910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109277597146256e-06,
+      "loss": 0.3397,
+      "step": 8911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.108650542665687e-06,
+      "loss": 0.9604,
+      "step": 8912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1080234478433115e-06,
+      "loss": 0.9038,
+      "step": 8913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107396312691124e-06,
+      "loss": 0.8848,
+      "step": 8914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106769137221124e-06,
+      "loss": 0.8228,
+      "step": 8915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106141921445313e-06,
+      "loss": 0.8887,
+      "step": 8916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1055146653756855e-06,
+      "loss": 0.8872,
+      "step": 8917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.104887369024245e-06,
+      "loss": 0.8223,
+      "step": 8918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.104260032402994e-06,
+      "loss": 0.8281,
+      "step": 8919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103632655523932e-06,
+      "loss": 0.9292,
+      "step": 8920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103005238399061e-06,
+      "loss": 0.8345,
+      "step": 8921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.102377781040387e-06,
+      "loss": 0.8164,
+      "step": 8922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101750283459913e-06,
+      "loss": 0.9365,
+      "step": 8923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101122745669644e-06,
+      "loss": 0.9399,
+      "step": 8924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.100495167681586e-06,
+      "loss": 0.8916,
+      "step": 8925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.099867549507747e-06,
+      "loss": 0.9067,
+      "step": 8926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0992398911601304e-06,
+      "loss": 0.9336,
+      "step": 8927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.098612192650747e-06,
+      "loss": 0.4434,
+      "step": 8928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.097984453991605e-06,
+      "loss": 0.9541,
+      "step": 8929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0973566751947164e-06,
+      "loss": 0.9199,
+      "step": 8930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096728856272088e-06,
+      "loss": 0.8696,
+      "step": 8931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096100997235733e-06,
+      "loss": 0.8647,
+      "step": 8932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.095473098097663e-06,
+      "loss": 0.8765,
+      "step": 8933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.094845158869891e-06,
+      "loss": 0.3137,
+      "step": 8934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.094217179564428e-06,
+      "loss": 0.9395,
+      "step": 8935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.093589160193293e-06,
+      "loss": 0.9111,
+      "step": 8936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0929611007684965e-06,
+      "loss": 0.875,
+      "step": 8937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.092333001302057e-06,
+      "loss": 0.8467,
+      "step": 8938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.091704861805989e-06,
+      "loss": 0.8511,
+      "step": 8939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0910766822923116e-06,
+      "loss": 0.3368,
+      "step": 8940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.090448462773043e-06,
+      "loss": 0.9019,
+      "step": 8941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.089820203260199e-06,
+      "loss": 0.8691,
+      "step": 8942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.089191903765803e-06,
+      "loss": 0.8564,
+      "step": 8943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.088563564301874e-06,
+      "loss": 0.8213,
+      "step": 8944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0879351848804315e-06,
+      "loss": 0.9141,
+      "step": 8945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.087306765513501e-06,
+      "loss": 0.8979,
+      "step": 8946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086678306213099e-06,
+      "loss": 0.8638,
+      "step": 8947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086049806991256e-06,
+      "loss": 0.9209,
+      "step": 8948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0854212678599935e-06,
+      "loss": 0.8906,
+      "step": 8949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084792688831334e-06,
+      "loss": 0.8789,
+      "step": 8950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084164069917304e-06,
+      "loss": 0.8926,
+      "step": 8951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.083535411129934e-06,
+      "loss": 0.8911,
+      "step": 8952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.082906712481248e-06,
+      "loss": 0.9004,
+      "step": 8953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.082277973983273e-06,
+      "loss": 0.8677,
+      "step": 8954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0816491956480394e-06,
+      "loss": 0.9155,
+      "step": 8955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081020377487576e-06,
+      "loss": 0.8813,
+      "step": 8956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.080391519513914e-06,
+      "loss": 0.8901,
+      "step": 8957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079762621739084e-06,
+      "loss": 0.9043,
+      "step": 8958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079133684175119e-06,
+      "loss": 0.9131,
+      "step": 8959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.078504706834049e-06,
+      "loss": 0.8916,
+      "step": 8960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.077875689727908e-06,
+      "loss": 0.9556,
+      "step": 8961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.077246632868732e-06,
+      "loss": 0.9009,
+      "step": 8962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.076617536268554e-06,
+      "loss": 0.9062,
+      "step": 8963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075988399939408e-06,
+      "loss": 0.9473,
+      "step": 8964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075359223893337e-06,
+      "loss": 0.9629,
+      "step": 8965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.074730008142372e-06,
+      "loss": 0.916,
+      "step": 8966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.074100752698551e-06,
+      "loss": 0.856,
+      "step": 8967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.073471457573915e-06,
+      "loss": 0.9072,
+      "step": 8968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.072842122780503e-06,
+      "loss": 0.8921,
+      "step": 8969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.072212748330355e-06,
+      "loss": 0.8979,
+      "step": 8970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.071583334235511e-06,
+      "loss": 0.9141,
+      "step": 8971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.070953880508013e-06,
+      "loss": 0.9028,
+      "step": 8972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.070324387159906e-06,
+      "loss": 0.8018,
+      "step": 8973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06969485420323e-06,
+      "loss": 0.937,
+      "step": 8974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0690652816500296e-06,
+      "loss": 0.9146,
+      "step": 8975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.068435669512349e-06,
+      "loss": 0.9214,
+      "step": 8976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0678060178022355e-06,
+      "loss": 0.9199,
+      "step": 8977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0671763265317365e-06,
+      "loss": 0.9375,
+      "step": 8978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.066546595712894e-06,
+      "loss": 0.8926,
+      "step": 8979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0659168253577595e-06,
+      "loss": 0.8984,
+      "step": 8980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.065287015478381e-06,
+      "loss": 0.9214,
+      "step": 8981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.064657166086808e-06,
+      "loss": 0.8638,
+      "step": 8982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.064027277195089e-06,
+      "loss": 0.8779,
+      "step": 8983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.063397348815274e-06,
+      "loss": 0.8389,
+      "step": 8984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062767380959419e-06,
+      "loss": 0.9062,
+      "step": 8985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062137373639573e-06,
+      "loss": 0.875,
+      "step": 8986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.061507326867788e-06,
+      "loss": 0.897,
+      "step": 8987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.060877240656121e-06,
+      "loss": 0.9414,
+      "step": 8988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.060247115016624e-06,
+      "loss": 0.8262,
+      "step": 8989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.059616949961353e-06,
+      "loss": 0.8774,
+      "step": 8990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.058986745502364e-06,
+      "loss": 0.8706,
+      "step": 8991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.058356501651715e-06,
+      "loss": 0.8887,
+      "step": 8992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057726218421461e-06,
+      "loss": 0.8291,
+      "step": 8993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057095895823663e-06,
+      "loss": 0.8848,
+      "step": 8994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05646553387038e-06,
+      "loss": 0.979,
+      "step": 8995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055835132573669e-06,
+      "loss": 0.918,
+      "step": 8996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055204691945593e-06,
+      "loss": 0.832,
+      "step": 8997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.054574211998213e-06,
+      "loss": 0.8916,
+      "step": 8998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05394369274359e-06,
+      "loss": 0.8423,
+      "step": 8999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.053313134193789e-06,
+      "loss": 0.8882,
+      "step": 9000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0526825363608695e-06,
+      "loss": 0.9395,
+      "step": 9001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0520518992569e-06,
+      "loss": 0.9102,
+      "step": 9002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.051421222893944e-06,
+      "loss": 0.9292,
+      "step": 9003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.050790507284068e-06,
+      "loss": 0.959,
+      "step": 9004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.050159752439336e-06,
+      "loss": 0.8594,
+      "step": 9005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.04952895837182e-06,
+      "loss": 0.916,
+      "step": 9006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.048898125093583e-06,
+      "loss": 0.8799,
+      "step": 9007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0482672526166986e-06,
+      "loss": 0.9067,
+      "step": 9008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.047636340953231e-06,
+      "loss": 0.9121,
+      "step": 9009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.047005390115255e-06,
+      "loss": 0.8823,
+      "step": 9010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.046374400114842e-06,
+      "loss": 0.8877,
+      "step": 9011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.04574337096406e-06,
+      "loss": 0.9214,
+      "step": 9012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.045112302674982e-06,
+      "loss": 0.8779,
+      "step": 9013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.044481195259686e-06,
+      "loss": 0.8564,
+      "step": 9014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.043850048730242e-06,
+      "loss": 0.9595,
+      "step": 9015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.043218863098725e-06,
+      "loss": 0.856,
+      "step": 9016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.042587638377213e-06,
+      "loss": 0.9458,
+      "step": 9017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.04195637457778e-06,
+      "loss": 0.853,
+      "step": 9018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.041325071712504e-06,
+      "loss": 0.8867,
+      "step": 9019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.040693729793462e-06,
+      "loss": 0.9033,
+      "step": 9020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.040062348832733e-06,
+      "loss": 0.812,
+      "step": 9021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0394309288423966e-06,
+      "loss": 0.7777,
+      "step": 9022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.038799469834534e-06,
+      "loss": 0.8262,
+      "step": 9023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0381679718212245e-06,
+      "loss": 0.8516,
+      "step": 9024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037536434814551e-06,
+      "loss": 0.8857,
+      "step": 9025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.036904858826592e-06,
+      "loss": 0.9131,
+      "step": 9026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.036273243869435e-06,
+      "loss": 0.8857,
+      "step": 9027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.035641589955162e-06,
+      "loss": 0.9258,
+      "step": 9028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.035009897095857e-06,
+      "loss": 0.9604,
+      "step": 9029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.034378165303606e-06,
+      "loss": 0.8696,
+      "step": 9030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033746394590495e-06,
+      "loss": 0.9033,
+      "step": 9031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033114584968611e-06,
+      "loss": 0.9175,
+      "step": 9032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.032482736450041e-06,
+      "loss": 0.9243,
+      "step": 9033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0318508490468725e-06,
+      "loss": 0.8936,
+      "step": 9034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.031218922771197e-06,
+      "loss": 0.8142,
+      "step": 9035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.030586957635101e-06,
+      "loss": 0.9453,
+      "step": 9036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.02995495365068e-06,
+      "loss": 0.8726,
+      "step": 9037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029322910830019e-06,
+      "loss": 0.8662,
+      "step": 9038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.028690829185213e-06,
+      "loss": 0.9478,
+      "step": 9039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.028058708728356e-06,
+      "loss": 0.9116,
+      "step": 9040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.02742654947154e-06,
+      "loss": 0.875,
+      "step": 9041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.026794351426858e-06,
+      "loss": 0.8896,
+      "step": 9042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.026162114606406e-06,
+      "loss": 0.8809,
+      "step": 9043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0255298390222815e-06,
+      "loss": 0.9229,
+      "step": 9044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.024897524686578e-06,
+      "loss": 0.9277,
+      "step": 9045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.024265171611394e-06,
+      "loss": 0.9238,
+      "step": 9046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023632779808829e-06,
+      "loss": 0.8784,
+      "step": 9047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023000349290979e-06,
+      "loss": 0.9102,
+      "step": 9048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022367880069945e-06,
+      "loss": 0.7876,
+      "step": 9049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.021735372157826e-06,
+      "loss": 0.8118,
+      "step": 9050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0211028255667235e-06,
+      "loss": 0.8828,
+      "step": 9051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.020470240308739e-06,
+      "loss": 0.9189,
+      "step": 9052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0198376163959745e-06,
+      "loss": 0.8564,
+      "step": 9053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.019204953840535e-06,
+      "loss": 0.9077,
+      "step": 9054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.018572252654522e-06,
+      "loss": 0.9492,
+      "step": 9055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.01793951285004e-06,
+      "loss": 0.8784,
+      "step": 9056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.017306734439197e-06,
+      "loss": 0.9233,
+      "step": 9057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016673917434098e-06,
+      "loss": 0.8555,
+      "step": 9058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016041061846846e-06,
+      "loss": 0.8276,
+      "step": 9059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0154081676895546e-06,
+      "loss": 0.833,
+      "step": 9060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.014775234974327e-06,
+      "loss": 0.9204,
+      "step": 9061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.014142263713275e-06,
+      "loss": 0.8691,
+      "step": 9062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0135092539185076e-06,
+      "loss": 0.8936,
+      "step": 9063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.012876205602135e-06,
+      "loss": 0.8491,
+      "step": 9064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.01224311877627e-06,
+      "loss": 0.3992,
+      "step": 9065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.011609993453021e-06,
+      "loss": 0.8623,
+      "step": 9066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010976829644505e-06,
+      "loss": 0.8896,
+      "step": 9067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010343627362831e-06,
+      "loss": 0.916,
+      "step": 9068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009710386620116e-06,
+      "loss": 0.8789,
+      "step": 9069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009077107428477e-06,
+      "loss": 0.873,
+      "step": 9070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.008443789800024e-06,
+      "loss": 0.8901,
+      "step": 9071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.007810433746877e-06,
+      "loss": 0.9824,
+      "step": 9072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0071770392811555e-06,
+      "loss": 0.9072,
+      "step": 9073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.006543606414971e-06,
+      "loss": 0.9136,
+      "step": 9074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.005910135160446e-06,
+      "loss": 0.9248,
+      "step": 9075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.005276625529701e-06,
+      "loss": 0.8926,
+      "step": 9076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004643077534853e-06,
+      "loss": 0.8218,
+      "step": 9077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004009491188023e-06,
+      "loss": 0.9077,
+      "step": 9078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.003375866501334e-06,
+      "loss": 0.9136,
+      "step": 9079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002742203486908e-06,
+      "loss": 0.8267,
+      "step": 9080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002108502156868e-06,
+      "loss": 0.8403,
+      "step": 9081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.001474762523337e-06,
+      "loss": 0.8735,
+      "step": 9082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00084098459844e-06,
+      "loss": 0.8906,
+      "step": 9083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.000207168394303e-06,
+      "loss": 0.8833,
+      "step": 9084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.999573313923049e-06,
+      "loss": 0.873,
+      "step": 9085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.998939421196809e-06,
+      "loss": 0.8477,
+      "step": 9086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.998305490227706e-06,
+      "loss": 0.917,
+      "step": 9087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99767152102787e-06,
+      "loss": 0.9033,
+      "step": 9088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997037513609432e-06,
+      "loss": 0.9092,
+      "step": 9089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.996403467984517e-06,
+      "loss": 0.8853,
+      "step": 9090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9957693841652586e-06,
+      "loss": 0.9092,
+      "step": 9091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.995135262163787e-06,
+      "loss": 0.937,
+      "step": 9092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.994501101992234e-06,
+      "loss": 0.8965,
+      "step": 9093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993866903662731e-06,
+      "loss": 0.9355,
+      "step": 9094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993232667187414e-06,
+      "loss": 0.9224,
+      "step": 9095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.992598392578413e-06,
+      "loss": 0.876,
+      "step": 9096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.991964079847866e-06,
+      "loss": 0.9141,
+      "step": 9097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.991329729007907e-06,
+      "loss": 0.895,
+      "step": 9098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.990695340070673e-06,
+      "loss": 0.9146,
+      "step": 9099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9900609130483e-06,
+      "loss": 0.8604,
+      "step": 9100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.989426447952925e-06,
+      "loss": 0.8765,
+      "step": 9101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.988791944796688e-06,
+      "loss": 0.9175,
+      "step": 9102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.988157403591728e-06,
+      "loss": 0.8735,
+      "step": 9103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9875228243501814e-06,
+      "loss": 0.9448,
+      "step": 9104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986888207084192e-06,
+      "loss": 0.9355,
+      "step": 9105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986253551805902e-06,
+      "loss": 0.876,
+      "step": 9106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98561885852745e-06,
+      "loss": 0.8887,
+      "step": 9107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.984984127260978e-06,
+      "loss": 0.874,
+      "step": 9108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.984349358018636e-06,
+      "loss": 0.877,
+      "step": 9109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983714550812562e-06,
+      "loss": 0.9629,
+      "step": 9110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9830797056549035e-06,
+      "loss": 0.9253,
+      "step": 9111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.982444822557805e-06,
+      "loss": 0.9067,
+      "step": 9112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.981809901533411e-06,
+      "loss": 0.8452,
+      "step": 9113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9811749425938736e-06,
+      "loss": 0.8696,
+      "step": 9114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.980539945751337e-06,
+      "loss": 0.9346,
+      "step": 9115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979904911017948e-06,
+      "loss": 0.9233,
+      "step": 9116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9792698384058595e-06,
+      "loss": 0.9155,
+      "step": 9117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.978634727927221e-06,
+      "loss": 0.9565,
+      "step": 9118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.977999579594183e-06,
+      "loss": 0.8086,
+      "step": 9119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.977364393418893e-06,
+      "loss": 0.8657,
+      "step": 9120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.976729169413508e-06,
+      "loss": 0.8926,
+      "step": 9121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.97609390759018e-06,
+      "loss": 0.918,
+      "step": 9122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.975458607961061e-06,
+      "loss": 0.8696,
+      "step": 9123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974823270538304e-06,
+      "loss": 0.9072,
+      "step": 9124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974187895334068e-06,
+      "loss": 0.8745,
+      "step": 9125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.973552482360506e-06,
+      "loss": 0.8765,
+      "step": 9126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.972917031629774e-06,
+      "loss": 0.8701,
+      "step": 9127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.972281543154033e-06,
+      "loss": 0.9224,
+      "step": 9128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971646016945436e-06,
+      "loss": 0.8857,
+      "step": 9129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971010453016144e-06,
+      "loss": 0.834,
+      "step": 9130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.970374851378317e-06,
+      "loss": 0.9072,
+      "step": 9131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969739212044113e-06,
+      "loss": 0.8545,
+      "step": 9132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9691035350256944e-06,
+      "loss": 0.8442,
+      "step": 9133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.968467820335222e-06,
+      "loss": 0.9326,
+      "step": 9134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967832067984859e-06,
+      "loss": 0.8611,
+      "step": 9135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967196277986767e-06,
+      "loss": 0.7734,
+      "step": 9136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.966560450353109e-06,
+      "loss": 0.8813,
+      "step": 9137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.965924585096053e-06,
+      "loss": 0.8447,
+      "step": 9138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.965288682227761e-06,
+      "loss": 0.9146,
+      "step": 9139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9646527417604e-06,
+      "loss": 0.8643,
+      "step": 9140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9640167637061355e-06,
+      "loss": 0.8784,
+      "step": 9141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.963380748077135e-06,
+      "loss": 0.9194,
+      "step": 9142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962744694885568e-06,
+      "loss": 0.9409,
+      "step": 9143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962108604143602e-06,
+      "loss": 0.9292,
+      "step": 9144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.961472475863406e-06,
+      "loss": 0.9009,
+      "step": 9145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9608363100571494e-06,
+      "loss": 0.8584,
+      "step": 9146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.960200106737006e-06,
+      "loss": 0.8804,
+      "step": 9147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.959563865915146e-06,
+      "loss": 0.9121,
+      "step": 9148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9589275876037395e-06,
+      "loss": 0.895,
+      "step": 9149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.958291271814963e-06,
+      "loss": 0.8486,
+      "step": 9150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.957654918560988e-06,
+      "loss": 0.8892,
+      "step": 9151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.95701852785399e-06,
+      "loss": 0.8911,
+      "step": 9152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.956382099706143e-06,
+      "loss": 0.8848,
+      "step": 9153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955745634129623e-06,
+      "loss": 0.8945,
+      "step": 9154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955109131136609e-06,
+      "loss": 0.8789,
+      "step": 9155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.954472590739275e-06,
+      "loss": 0.8623,
+      "step": 9156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953836012949799e-06,
+      "loss": 0.896,
+      "step": 9157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953199397780365e-06,
+      "loss": 0.8862,
+      "step": 9158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.952562745243147e-06,
+      "loss": 0.8818,
+      "step": 9159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9519260553503265e-06,
+      "loss": 0.8975,
+      "step": 9160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.951289328114084e-06,
+      "loss": 0.9551,
+      "step": 9161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.950652563546602e-06,
+      "loss": 0.8975,
+      "step": 9162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.950015761660064e-06,
+      "loss": 0.8677,
+      "step": 9163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.949378922466651e-06,
+      "loss": 0.873,
+      "step": 9164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.948742045978548e-06,
+      "loss": 0.9336,
+      "step": 9165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.948105132207937e-06,
+      "loss": 0.8979,
+      "step": 9166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947468181167007e-06,
+      "loss": 0.9253,
+      "step": 9167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.946831192867941e-06,
+      "loss": 0.8936,
+      "step": 9168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9461941673229264e-06,
+      "loss": 0.8457,
+      "step": 9169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9455571045441495e-06,
+      "loss": 0.9082,
+      "step": 9170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9449200045438014e-06,
+      "loss": 0.8979,
+      "step": 9171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.944282867334067e-06,
+      "loss": 0.8481,
+      "step": 9172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.943645692927139e-06,
+      "loss": 0.8799,
+      "step": 9173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.943008481335205e-06,
+      "loss": 0.9028,
+      "step": 9174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.942371232570458e-06,
+      "loss": 0.8867,
+      "step": 9175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.941733946645088e-06,
+      "loss": 0.8901,
+      "step": 9176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.941096623571286e-06,
+      "loss": 0.8809,
+      "step": 9177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9404592633612486e-06,
+      "loss": 0.8477,
+      "step": 9178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939821866027168e-06,
+      "loss": 0.9048,
+      "step": 9179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939184431581239e-06,
+      "loss": 0.9165,
+      "step": 9180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.938546960035655e-06,
+      "loss": 0.3353,
+      "step": 9181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9379094514026115e-06,
+      "loss": 0.7986,
+      "step": 9182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.937271905694308e-06,
+      "loss": 0.9326,
+      "step": 9183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.93663432292294e-06,
+      "loss": 0.9058,
+      "step": 9184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935996703100707e-06,
+      "loss": 0.4331,
+      "step": 9185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935359046239803e-06,
+      "loss": 0.9297,
+      "step": 9186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.934721352352434e-06,
+      "loss": 0.8838,
+      "step": 9187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.934083621450796e-06,
+      "loss": 0.8564,
+      "step": 9188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933445853547089e-06,
+      "loss": 0.8794,
+      "step": 9189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.932808048653518e-06,
+      "loss": 0.874,
+      "step": 9190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9321702067822815e-06,
+      "loss": 0.9541,
+      "step": 9191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.931532327945585e-06,
+      "loss": 0.9238,
+      "step": 9192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.930894412155631e-06,
+      "loss": 0.8784,
+      "step": 9193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.930256459424627e-06,
+      "loss": 0.3682,
+      "step": 9194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9296184697647705e-06,
+      "loss": 0.8774,
+      "step": 9195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.928980443188275e-06,
+      "loss": 0.8677,
+      "step": 9196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.928342379707344e-06,
+      "loss": 0.873,
+      "step": 9197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927704279334184e-06,
+      "loss": 0.9951,
+      "step": 9198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927066142081003e-06,
+      "loss": 0.8584,
+      "step": 9199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92642796796001e-06,
+      "loss": 0.9087,
+      "step": 9200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.925789756983416e-06,
+      "loss": 0.9038,
+      "step": 9201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.925151509163427e-06,
+      "loss": 0.8667,
+      "step": 9202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.924513224512257e-06,
+      "loss": 0.8823,
+      "step": 9203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.923874903042116e-06,
+      "loss": 0.9282,
+      "step": 9204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.923236544765216e-06,
+      "loss": 0.918,
+      "step": 9205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.922598149693771e-06,
+      "loss": 0.9287,
+      "step": 9206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.921959717839994e-06,
+      "loss": 0.8916,
+      "step": 9207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.921321249216099e-06,
+      "loss": 0.8716,
+      "step": 9208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.920682743834301e-06,
+      "loss": 0.8135,
+      "step": 9209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.920044201706815e-06,
+      "loss": 0.8408,
+      "step": 9210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.919405622845857e-06,
+      "loss": 0.834,
+      "step": 9211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918767007263646e-06,
+      "loss": 0.8408,
+      "step": 9212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918128354972397e-06,
+      "loss": 0.8691,
+      "step": 9213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.917489665984333e-06,
+      "loss": 0.8257,
+      "step": 9214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916850940311668e-06,
+      "loss": 0.8096,
+      "step": 9215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916212177966623e-06,
+      "loss": 0.8716,
+      "step": 9216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.915573378961422e-06,
+      "loss": 0.9023,
+      "step": 9217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914934543308283e-06,
+      "loss": 0.8555,
+      "step": 9218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914295671019427e-06,
+      "loss": 0.8965,
+      "step": 9219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.913656762107079e-06,
+      "loss": 0.9048,
+      "step": 9220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9130178165834625e-06,
+      "loss": 0.8931,
+      "step": 9221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9123788344608e-06,
+      "loss": 0.817,
+      "step": 9222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911739815751317e-06,
+      "loss": 0.8613,
+      "step": 9223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911100760467239e-06,
+      "loss": 0.8271,
+      "step": 9224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910461668620792e-06,
+      "loss": 0.9023,
+      "step": 9225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9098225402242015e-06,
+      "loss": 0.8979,
+      "step": 9226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.909183375289697e-06,
+      "loss": 0.7562,
+      "step": 9227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.908544173829505e-06,
+      "loss": 0.8545,
+      "step": 9228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.907904935855858e-06,
+      "loss": 0.8965,
+      "step": 9229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.907265661380982e-06,
+      "loss": 0.8896,
+      "step": 9230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906626350417108e-06,
+      "loss": 0.8862,
+      "step": 9231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.905987002976466e-06,
+      "loss": 0.9434,
+      "step": 9232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.905347619071292e-06,
+      "loss": 0.9097,
+      "step": 9233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904708198713815e-06,
+      "loss": 0.8276,
+      "step": 9234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904068741916267e-06,
+      "loss": 0.8892,
+      "step": 9235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.903429248690885e-06,
+      "loss": 0.9038,
+      "step": 9236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.902789719049903e-06,
+      "loss": 0.8843,
+      "step": 9237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.902150153005554e-06,
+      "loss": 0.8994,
+      "step": 9238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.901510550570074e-06,
+      "loss": 0.8672,
+      "step": 9239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.900870911755702e-06,
+      "loss": 0.918,
+      "step": 9240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9002312365746736e-06,
+      "loss": 0.9453,
+      "step": 9241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.899591525039227e-06,
+      "loss": 0.9438,
+      "step": 9242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.898951777161603e-06,
+      "loss": 0.8145,
+      "step": 9243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.898311992954037e-06,
+      "loss": 0.9297,
+      "step": 9244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.897672172428772e-06,
+      "loss": 0.8711,
+      "step": 9245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.897032315598049e-06,
+      "loss": 0.832,
+      "step": 9246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.896392422474108e-06,
+      "loss": 0.9009,
+      "step": 9247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.895752493069189e-06,
+      "loss": 0.347,
+      "step": 9248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.89511252739554e-06,
+      "loss": 0.9194,
+      "step": 9249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.894472525465402e-06,
+      "loss": 0.3842,
+      "step": 9250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893832487291019e-06,
+      "loss": 0.9277,
+      "step": 9251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8931924128846365e-06,
+      "loss": 0.873,
+      "step": 9252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.892552302258498e-06,
+      "loss": 0.915,
+      "step": 9253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.891912155424853e-06,
+      "loss": 0.9492,
+      "step": 9254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.891271972395947e-06,
+      "loss": 0.8672,
+      "step": 9255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.89063175318403e-06,
+      "loss": 0.9253,
+      "step": 9256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.889991497801345e-06,
+      "loss": 0.8765,
+      "step": 9257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8893512062601466e-06,
+      "loss": 0.8892,
+      "step": 9258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888710878572682e-06,
+      "loss": 0.8154,
+      "step": 9259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888070514751202e-06,
+      "loss": 0.9111,
+      "step": 9260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.887430114807956e-06,
+      "loss": 0.8418,
+      "step": 9261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8867896787552e-06,
+      "loss": 0.9282,
+      "step": 9262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.886149206605184e-06,
+      "loss": 0.8936,
+      "step": 9263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8855086983701606e-06,
+      "loss": 0.8623,
+      "step": 9264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884868154062383e-06,
+      "loss": 0.9014,
+      "step": 9265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884227573694111e-06,
+      "loss": 0.8369,
+      "step": 9266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8835869572775935e-06,
+      "loss": 0.8901,
+      "step": 9267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8829463048250914e-06,
+      "loss": 0.8867,
+      "step": 9268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8823056163488575e-06,
+      "loss": 0.8828,
+      "step": 9269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8816648918611525e-06,
+      "loss": 0.8926,
+      "step": 9270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881024131374232e-06,
+      "loss": 0.8247,
+      "step": 9271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8803833349003555e-06,
+      "loss": 0.9268,
+      "step": 9272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.879742502451784e-06,
+      "loss": 0.8623,
+      "step": 9273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.879101634040776e-06,
+      "loss": 0.8877,
+      "step": 9274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.878460729679593e-06,
+      "loss": 0.9351,
+      "step": 9275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877819789380496e-06,
+      "loss": 0.916,
+      "step": 9276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877178813155745e-06,
+      "loss": 0.877,
+      "step": 9277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.876537801017607e-06,
+      "loss": 0.8521,
+      "step": 9278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.875896752978345e-06,
+      "loss": 0.856,
+      "step": 9279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.875255669050221e-06,
+      "loss": 0.8384,
+      "step": 9280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.874614549245499e-06,
+      "loss": 0.9365,
+      "step": 9281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87397339357645e-06,
+      "loss": 0.9097,
+      "step": 9282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.873332202055336e-06,
+      "loss": 0.8818,
+      "step": 9283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.872690974694425e-06,
+      "loss": 0.8784,
+      "step": 9284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.872049711505984e-06,
+      "loss": 0.8628,
+      "step": 9285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.871408412502282e-06,
+      "loss": 0.8652,
+      "step": 9286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.870767077695588e-06,
+      "loss": 0.9292,
+      "step": 9287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.870125707098172e-06,
+      "loss": 0.8779,
+      "step": 9288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.869484300722304e-06,
+      "loss": 0.9146,
+      "step": 9289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868842858580256e-06,
+      "loss": 0.8896,
+      "step": 9290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8682013806842985e-06,
+      "loss": 0.8687,
+      "step": 9291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8675598670467066e-06,
+      "loss": 0.8564,
+      "step": 9292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.86691831767975e-06,
+      "loss": 0.8096,
+      "step": 9293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8662767325957036e-06,
+      "loss": 0.8623,
+      "step": 9294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.865635111806843e-06,
+      "loss": 0.7974,
+      "step": 9295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864993455325444e-06,
+      "loss": 0.8833,
+      "step": 9296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864351763163779e-06,
+      "loss": 0.9746,
+      "step": 9297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.863710035334129e-06,
+      "loss": 0.9663,
+      "step": 9298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.863068271848767e-06,
+      "loss": 0.8955,
+      "step": 9299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8624264727199755e-06,
+      "loss": 0.3788,
+      "step": 9300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.86178463796003e-06,
+      "loss": 0.9141,
+      "step": 9301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8611427675812095e-06,
+      "loss": 0.8721,
+      "step": 9302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.860500861595796e-06,
+      "loss": 0.9277,
+      "step": 9303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859858920016069e-06,
+      "loss": 0.2848,
+      "step": 9304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859216942854311e-06,
+      "loss": 0.8804,
+      "step": 9305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.858574930122801e-06,
+      "loss": 0.855,
+      "step": 9306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.857932881833825e-06,
+      "loss": 0.8735,
+      "step": 9307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8572907979996665e-06,
+      "loss": 0.9238,
+      "step": 9308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856648678632607e-06,
+      "loss": 0.8633,
+      "step": 9309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856006523744932e-06,
+      "loss": 0.8716,
+      "step": 9310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855364333348928e-06,
+      "loss": 0.8862,
+      "step": 9311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8547221074568816e-06,
+      "loss": 0.9194,
+      "step": 9312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.854079846081077e-06,
+      "loss": 0.8628,
+      "step": 9313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.853437549233805e-06,
+      "loss": 0.7661,
+      "step": 9314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.852795216927351e-06,
+      "loss": 0.8823,
+      "step": 9315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.852152849174005e-06,
+      "loss": 0.8706,
+      "step": 9316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.851510445986056e-06,
+      "loss": 0.9399,
+      "step": 9317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850868007375795e-06,
+      "loss": 0.8706,
+      "step": 9318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850225533355513e-06,
+      "loss": 0.9253,
+      "step": 9319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.849583023937499e-06,
+      "loss": 0.812,
+      "step": 9320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848940479134049e-06,
+      "loss": 0.9668,
+      "step": 9321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848297898957452e-06,
+      "loss": 0.8662,
+      "step": 9322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.847655283420006e-06,
+      "loss": 0.9082,
+      "step": 9323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8470126325340025e-06,
+      "loss": 0.8975,
+      "step": 9324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.846369946311737e-06,
+      "loss": 0.3048,
+      "step": 9325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8457272247655035e-06,
+      "loss": 0.9048,
+      "step": 9326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8450844679076005e-06,
+      "loss": 0.7988,
+      "step": 9327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.844441675750324e-06,
+      "loss": 0.877,
+      "step": 9328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.843798848305972e-06,
+      "loss": 0.9443,
+      "step": 9329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.843155985586842e-06,
+      "loss": 0.8511,
+      "step": 9330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.842513087605234e-06,
+      "loss": 0.8789,
+      "step": 9331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841870154373447e-06,
+      "loss": 0.8823,
+      "step": 9332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841227185903781e-06,
+      "loss": 0.8315,
+      "step": 9333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8405841822085384e-06,
+      "loss": 0.7286,
+      "step": 9334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839941143300019e-06,
+      "loss": 0.8779,
+      "step": 9335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839298069190528e-06,
+      "loss": 0.9336,
+      "step": 9336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.838654959892365e-06,
+      "loss": 0.8662,
+      "step": 9337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.838011815417838e-06,
+      "loss": 0.9263,
+      "step": 9338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.837368635779245e-06,
+      "loss": 0.9307,
+      "step": 9339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.836725420988895e-06,
+      "loss": 0.9263,
+      "step": 9340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.836082171059095e-06,
+      "loss": 0.8965,
+      "step": 9341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.83543888600215e-06,
+      "loss": 0.8408,
+      "step": 9342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.834795565830364e-06,
+      "loss": 0.9243,
+      "step": 9343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.834152210556051e-06,
+      "loss": 0.8501,
+      "step": 9344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833508820191514e-06,
+      "loss": 0.8843,
+      "step": 9345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.832865394749065e-06,
+      "loss": 0.8579,
+      "step": 9346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.832221934241012e-06,
+      "loss": 0.8442,
+      "step": 9347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.831578438679667e-06,
+      "loss": 0.8833,
+      "step": 9348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830934908077339e-06,
+      "loss": 0.8745,
+      "step": 9349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830291342446342e-06,
+      "loss": 0.8877,
+      "step": 9350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.829647741798987e-06,
+      "loss": 0.8516,
+      "step": 9351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8290041061475875e-06,
+      "loss": 0.7627,
+      "step": 9352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8283604355044575e-06,
+      "loss": 0.8779,
+      "step": 9353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.827716729881913e-06,
+      "loss": 0.8682,
+      "step": 9354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.827072989292264e-06,
+      "loss": 0.8911,
+      "step": 9355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8264292137478305e-06,
+      "loss": 0.9438,
+      "step": 9356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82578540326093e-06,
+      "loss": 0.9102,
+      "step": 9357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.825141557843876e-06,
+      "loss": 0.8516,
+      "step": 9358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.824497677508988e-06,
+      "loss": 0.8901,
+      "step": 9359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.823853762268584e-06,
+      "loss": 0.876,
+      "step": 9360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8232098121349855e-06,
+      "loss": 0.8293,
+      "step": 9361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.822565827120509e-06,
+      "loss": 0.8252,
+      "step": 9362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.821921807237475e-06,
+      "loss": 0.8403,
+      "step": 9363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.821277752498208e-06,
+      "loss": 0.8848,
+      "step": 9364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.820633662915026e-06,
+      "loss": 0.8413,
+      "step": 9365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819989538500253e-06,
+      "loss": 0.8726,
+      "step": 9366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819345379266214e-06,
+      "loss": 0.9409,
+      "step": 9367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8187011852252294e-06,
+      "loss": 0.8179,
+      "step": 9368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.818056956389625e-06,
+      "loss": 0.8789,
+      "step": 9369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8174126927717275e-06,
+      "loss": 0.8906,
+      "step": 9370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.81676839438386e-06,
+      "loss": 0.8872,
+      "step": 9371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.816124061238351e-06,
+      "loss": 0.8457,
+      "step": 9372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815479693347527e-06,
+      "loss": 0.8877,
+      "step": 9373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8148352907237145e-06,
+      "loss": 0.8057,
+      "step": 9374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.814190853379244e-06,
+      "loss": 0.8774,
+      "step": 9375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.813546381326443e-06,
+      "loss": 0.9204,
+      "step": 9376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.812901874577643e-06,
+      "loss": 0.8931,
+      "step": 9377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.812257333145172e-06,
+      "loss": 0.9097,
+      "step": 9378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811612757041364e-06,
+      "loss": 0.3669,
+      "step": 9379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8109681462785484e-06,
+      "loss": 0.9404,
+      "step": 9380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.810323500869058e-06,
+      "loss": 0.8896,
+      "step": 9381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809678820825227e-06,
+      "loss": 0.8926,
+      "step": 9382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809034106159388e-06,
+      "loss": 0.8457,
+      "step": 9383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.808389356883876e-06,
+      "loss": 0.9048,
+      "step": 9384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.807744573011026e-06,
+      "loss": 0.9209,
+      "step": 9385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.807099754553174e-06,
+      "loss": 0.8716,
+      "step": 9386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.806454901522656e-06,
+      "loss": 0.9209,
+      "step": 9387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.805810013931808e-06,
+      "loss": 0.8862,
+      "step": 9388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80516509179297e-06,
+      "loss": 0.9468,
+      "step": 9389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80452013511848e-06,
+      "loss": 0.894,
+      "step": 9390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.803875143920675e-06,
+      "loss": 0.9175,
+      "step": 9391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.803230118211895e-06,
+      "loss": 0.9048,
+      "step": 9392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.802585058004484e-06,
+      "loss": 0.7906,
+      "step": 9393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.801939963310778e-06,
+      "loss": 0.8291,
+      "step": 9394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8012948341431216e-06,
+      "loss": 0.8706,
+      "step": 9395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.800649670513857e-06,
+      "loss": 0.853,
+      "step": 9396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.800004472435326e-06,
+      "loss": 0.8735,
+      "step": 9397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.799359239919874e-06,
+      "loss": 0.9102,
+      "step": 9398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7987139729798436e-06,
+      "loss": 0.8201,
+      "step": 9399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.798068671627581e-06,
+      "loss": 0.8579,
+      "step": 9400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7974233358754305e-06,
+      "loss": 0.8682,
+      "step": 9401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796777965735739e-06,
+      "loss": 0.8833,
+      "step": 9402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796132561220855e-06,
+      "loss": 0.8184,
+      "step": 9403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795487122343124e-06,
+      "loss": 0.8848,
+      "step": 9404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.794841649114894e-06,
+      "loss": 0.8599,
+      "step": 9405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.794196141548517e-06,
+      "loss": 0.9668,
+      "step": 9406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.793550599656339e-06,
+      "loss": 0.8975,
+      "step": 9407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792905023450712e-06,
+      "loss": 0.8672,
+      "step": 9408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792259412943987e-06,
+      "loss": 0.8608,
+      "step": 9409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.791613768148515e-06,
+      "loss": 0.8882,
+      "step": 9410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7909680890766485e-06,
+      "loss": 0.8604,
+      "step": 9411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.79032237574074e-06,
+      "loss": 0.9136,
+      "step": 9412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7896766281531435e-06,
+      "loss": 0.8306,
+      "step": 9413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.789030846326212e-06,
+      "loss": 0.8716,
+      "step": 9414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.788385030272302e-06,
+      "loss": 0.9048,
+      "step": 9415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78773918000377e-06,
+      "loss": 0.9199,
+      "step": 9416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.787093295532967e-06,
+      "loss": 0.8667,
+      "step": 9417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.786447376872255e-06,
+      "loss": 0.8936,
+      "step": 9418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78580142403399e-06,
+      "loss": 0.9546,
+      "step": 9419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.785155437030529e-06,
+      "loss": 0.875,
+      "step": 9420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7845094158742294e-06,
+      "loss": 0.9062,
+      "step": 9421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783863360577456e-06,
+      "loss": 0.9033,
+      "step": 9422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783217271152564e-06,
+      "loss": 0.8374,
+      "step": 9423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.782571147611916e-06,
+      "loss": 0.8257,
+      "step": 9424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781924989967873e-06,
+      "loss": 0.8589,
+      "step": 9425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781278798232796e-06,
+      "loss": 0.8442,
+      "step": 9426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78063257241905e-06,
+      "loss": 0.8857,
+      "step": 9427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7799863125389965e-06,
+      "loss": 0.8628,
+      "step": 9428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.779340018605001e-06,
+      "loss": 0.8477,
+      "step": 9429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778693690629424e-06,
+      "loss": 0.8818,
+      "step": 9430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778047328624636e-06,
+      "loss": 0.8809,
+      "step": 9431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.777400932603e-06,
+      "loss": 0.8179,
+      "step": 9432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776754502576886e-06,
+      "loss": 0.8975,
+      "step": 9433
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776108038558655e-06,
+      "loss": 0.9028,
+      "step": 9434
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7754615405606795e-06,
+      "loss": 0.9106,
+      "step": 9435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774815008595328e-06,
+      "loss": 0.8672,
+      "step": 9436
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7741684426749675e-06,
+      "loss": 0.918,
+      "step": 9437
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.773521842811967e-06,
+      "loss": 0.8975,
+      "step": 9438
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7728752090187035e-06,
+      "loss": 0.8896,
+      "step": 9439
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772228541307541e-06,
+      "loss": 0.8799,
+      "step": 9440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.771581839690854e-06,
+      "loss": 0.8364,
+      "step": 9441
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7709351041810144e-06,
+      "loss": 0.9512,
+      "step": 9442
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770288334790396e-06,
+      "loss": 0.894,
+      "step": 9443
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.769641531531372e-06,
+      "loss": 0.8535,
+      "step": 9444
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768994694416316e-06,
+      "loss": 0.9346,
+      "step": 9445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768347823457606e-06,
+      "loss": 0.8794,
+      "step": 9446
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767700918667614e-06,
+      "loss": 0.9058,
+      "step": 9447
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767053980058718e-06,
+      "loss": 0.8848,
+      "step": 9448
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.766407007643297e-06,
+      "loss": 0.8345,
+      "step": 9449
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.765760001433723e-06,
+      "loss": 0.7798,
+      "step": 9450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76511296144238e-06,
+      "loss": 0.9028,
+      "step": 9451
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.764465887681644e-06,
+      "loss": 0.9312,
+      "step": 9452
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.763818780163895e-06,
+      "loss": 0.8765,
+      "step": 9453
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.763171638901512e-06,
+      "loss": 0.8335,
+      "step": 9454
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76252446390688e-06,
+      "loss": 0.8896,
+      "step": 9455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.761877255192375e-06,
+      "loss": 0.8496,
+      "step": 9456
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.761230012770383e-06,
+      "loss": 0.8584,
+      "step": 9457
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.760582736653285e-06,
+      "loss": 0.7837,
+      "step": 9458
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.759935426853466e-06,
+      "loss": 0.8726,
+      "step": 9459
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.759288083383308e-06,
+      "loss": 0.875,
+      "step": 9460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.758640706255196e-06,
+      "loss": 0.8633,
+      "step": 9461
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757993295481519e-06,
+      "loss": 0.4087,
+      "step": 9462
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757345851074656e-06,
+      "loss": 0.8711,
+      "step": 9463
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756698373046999e-06,
+      "loss": 0.8921,
+      "step": 9464
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756050861410936e-06,
+      "loss": 0.9165,
+      "step": 9465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.755403316178851e-06,
+      "loss": 0.8267,
+      "step": 9466
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.754755737363133e-06,
+      "loss": 0.874,
+      "step": 9467
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7541081249761755e-06,
+      "loss": 0.8809,
+      "step": 9468
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7534604790303646e-06,
+      "loss": 0.8423,
+      "step": 9469
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752812799538091e-06,
+      "loss": 0.3374,
+      "step": 9470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752165086511746e-06,
+      "loss": 0.9175,
+      "step": 9471
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.751517339963723e-06,
+      "loss": 0.8677,
+      "step": 9472
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750869559906412e-06,
+      "loss": 0.917,
+      "step": 9473
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750221746352208e-06,
+      "loss": 0.8643,
+      "step": 9474
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.749573899313504e-06,
+      "loss": 0.9858,
+      "step": 9475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.748926018802695e-06,
+      "loss": 0.8906,
+      "step": 9476
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.748278104832175e-06,
+      "loss": 0.8721,
+      "step": 9477
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.74763015741434e-06,
+      "loss": 0.8267,
+      "step": 9478
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.746982176561584e-06,
+      "loss": 0.8799,
+      "step": 9479
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7463341622863074e-06,
+      "loss": 0.9209,
+      "step": 9480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.745686114600908e-06,
+      "loss": 0.894,
+      "step": 9481
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.745038033517781e-06,
+      "loss": 0.3387,
+      "step": 9482
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.744389919049326e-06,
+      "loss": 0.8589,
+      "step": 9483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743741771207944e-06,
+      "loss": 0.8398,
+      "step": 9484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743093590006033e-06,
+      "loss": 0.8296,
+      "step": 9485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.742445375455997e-06,
+      "loss": 0.9019,
+      "step": 9486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.741797127570232e-06,
+      "loss": 0.8242,
+      "step": 9487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.741148846361146e-06,
+      "loss": 0.9053,
+      "step": 9488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740500531841139e-06,
+      "loss": 0.9263,
+      "step": 9489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739852184022613e-06,
+      "loss": 0.3585,
+      "step": 9490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739203802917973e-06,
+      "loss": 0.9331,
+      "step": 9491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.738555388539622e-06,
+      "loss": 0.9038,
+      "step": 9492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.73790694089997e-06,
+      "loss": 0.9634,
+      "step": 9493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7372584600114175e-06,
+      "loss": 0.7944,
+      "step": 9494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.736609945886375e-06,
+      "loss": 0.8813,
+      "step": 9495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735961398537245e-06,
+      "loss": 0.8823,
+      "step": 9496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735312817976439e-06,
+      "loss": 0.8628,
+      "step": 9497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7346642042163655e-06,
+      "loss": 0.8091,
+      "step": 9498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7340155572694325e-06,
+      "loss": 0.8779,
+      "step": 9499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.733366877148047e-06,
+      "loss": 0.8779,
+      "step": 9500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732718163864624e-06,
+      "loss": 0.814,
+      "step": 9501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732069417431571e-06,
+      "loss": 0.9146,
+      "step": 9502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.731420637861301e-06,
+      "loss": 0.8809,
+      "step": 9503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.730771825166225e-06,
+      "loss": 0.873,
+      "step": 9504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.730122979358757e-06,
+      "loss": 0.9697,
+      "step": 9505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7294741004513095e-06,
+      "loss": 0.8711,
+      "step": 9506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7288251884562985e-06,
+      "loss": 0.8789,
+      "step": 9507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7281762433861375e-06,
+      "loss": 0.8955,
+      "step": 9508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.72752726525324e-06,
+      "loss": 0.8047,
+      "step": 9509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.726878254070025e-06,
+      "loss": 0.8257,
+      "step": 9510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.726229209848909e-06,
+      "loss": 0.8809,
+      "step": 9511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725580132602305e-06,
+      "loss": 0.8813,
+      "step": 9512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.724931022342636e-06,
+      "loss": 0.9238,
+      "step": 9513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.724281879082317e-06,
+      "loss": 0.8081,
+      "step": 9514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.72363270283377e-06,
+      "loss": 0.8823,
+      "step": 9515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.722983493609412e-06,
+      "loss": 0.8984,
+      "step": 9516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.722334251421665e-06,
+      "loss": 0.8682,
+      "step": 9517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.721684976282951e-06,
+      "loss": 0.9258,
+      "step": 9518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.721035668205689e-06,
+      "loss": 0.8271,
+      "step": 9519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.720386327202302e-06,
+      "loss": 0.9336,
+      "step": 9520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719736953285215e-06,
+      "loss": 0.8809,
+      "step": 9521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719087546466849e-06,
+      "loss": 0.8784,
+      "step": 9522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718438106759631e-06,
+      "loss": 0.8584,
+      "step": 9523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7177886341759825e-06,
+      "loss": 0.9189,
+      "step": 9524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.717139128728331e-06,
+      "loss": 0.9097,
+      "step": 9525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7164895904291015e-06,
+      "loss": 0.8159,
+      "step": 9526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.715840019290723e-06,
+      "loss": 0.9097,
+      "step": 9527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.71519041532562e-06,
+      "loss": 0.9561,
+      "step": 9528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.71454077854622e-06,
+      "loss": 0.7915,
+      "step": 9529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713891108964956e-06,
+      "loss": 0.8892,
+      "step": 9530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713241406594253e-06,
+      "loss": 0.8984,
+      "step": 9531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.712591671446542e-06,
+      "loss": 0.8838,
+      "step": 9532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7119419035342535e-06,
+      "loss": 0.8701,
+      "step": 9533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.71129210286982e-06,
+      "loss": 0.8315,
+      "step": 9534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7106422694656704e-06,
+      "loss": 0.9053,
+      "step": 9535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.709992403334239e-06,
+      "loss": 0.8789,
+      "step": 9536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.709342504487958e-06,
+      "loss": 0.9126,
+      "step": 9537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708692572939262e-06,
+      "loss": 0.8267,
+      "step": 9538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708042608700585e-06,
+      "loss": 0.3052,
+      "step": 9539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.70739261178436e-06,
+      "loss": 0.8984,
+      "step": 9540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706742582203024e-06,
+      "loss": 0.873,
+      "step": 9541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706092519969013e-06,
+      "loss": 0.9082,
+      "step": 9542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.705442425094765e-06,
+      "loss": 0.8867,
+      "step": 9543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.704792297592715e-06,
+      "loss": 0.8594,
+      "step": 9544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7041421374753004e-06,
+      "loss": 0.8643,
+      "step": 9545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.703491944754963e-06,
+      "loss": 0.8638,
+      "step": 9546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702841719444141e-06,
+      "loss": 0.9326,
+      "step": 9547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702191461555272e-06,
+      "loss": 0.8921,
+      "step": 9548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.701541171100798e-06,
+      "loss": 0.9282,
+      "step": 9549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700890848093161e-06,
+      "loss": 0.8896,
+      "step": 9550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700240492544802e-06,
+      "loss": 0.8994,
+      "step": 9551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.699590104468162e-06,
+      "loss": 0.9185,
+      "step": 9552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.698939683875686e-06,
+      "loss": 0.8921,
+      "step": 9553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.698289230779817e-06,
+      "loss": 0.8882,
+      "step": 9554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6976387451929985e-06,
+      "loss": 0.8403,
+      "step": 9555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.696988227127674e-06,
+      "loss": 0.8398,
+      "step": 9556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.696337676596292e-06,
+      "loss": 0.8838,
+      "step": 9557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6956870936112975e-06,
+      "loss": 0.8301,
+      "step": 9558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.695036478185137e-06,
+      "loss": 0.8491,
+      "step": 9559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.694385830330257e-06,
+      "loss": 0.8989,
+      "step": 9560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.693735150059107e-06,
+      "loss": 0.7601,
+      "step": 9561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.693084437384132e-06,
+      "loss": 0.8638,
+      "step": 9562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.692433692317787e-06,
+      "loss": 0.9043,
+      "step": 9563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6917829148725165e-06,
+      "loss": 0.8804,
+      "step": 9564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691132105060774e-06,
+      "loss": 0.8623,
+      "step": 9565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.690481262895008e-06,
+      "loss": 0.875,
+      "step": 9566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.689830388387672e-06,
+      "loss": 0.8896,
+      "step": 9567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6891794815512175e-06,
+      "loss": 0.8806,
+      "step": 9568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6885285423980995e-06,
+      "loss": 0.8843,
+      "step": 9569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687877570940769e-06,
+      "loss": 0.8594,
+      "step": 9570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687226567191681e-06,
+      "loss": 0.8843,
+      "step": 9571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.686575531163288e-06,
+      "loss": 0.9336,
+      "step": 9572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.685924462868049e-06,
+      "loss": 0.9307,
+      "step": 9573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.685273362318418e-06,
+      "loss": 0.8125,
+      "step": 9574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.684622229526852e-06,
+      "loss": 0.9473,
+      "step": 9575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68397106450581e-06,
+      "loss": 0.8774,
+      "step": 9576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.683319867267746e-06,
+      "loss": 0.8589,
+      "step": 9577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682668637825119e-06,
+      "loss": 0.8979,
+      "step": 9578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682017376190392e-06,
+      "loss": 0.8755,
+      "step": 9579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6813660823760205e-06,
+      "loss": 0.9111,
+      "step": 9580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.680714756394467e-06,
+      "loss": 0.8711,
+      "step": 9581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.680063398258192e-06,
+      "loss": 0.917,
+      "step": 9582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6794120079796554e-06,
+      "loss": 0.8999,
+      "step": 9583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.678760585571322e-06,
+      "loss": 0.8398,
+      "step": 9584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.678109131045652e-06,
+      "loss": 0.8716,
+      "step": 9585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.677457644415111e-06,
+      "loss": 0.9424,
+      "step": 9586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.676806125692161e-06,
+      "loss": 0.8921,
+      "step": 9587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.676154574889269e-06,
+      "loss": 0.9067,
+      "step": 9588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.675502992018899e-06,
+      "loss": 0.8535,
+      "step": 9589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6748513770935145e-06,
+      "loss": 0.9521,
+      "step": 9590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.674199730125584e-06,
+      "loss": 0.8862,
+      "step": 9591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.673548051127576e-06,
+      "loss": 0.9141,
+      "step": 9592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6728963401119565e-06,
+      "loss": 0.8218,
+      "step": 9593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.672244597091192e-06,
+      "loss": 0.915,
+      "step": 9594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6715928220777545e-06,
+      "loss": 0.9082,
+      "step": 9595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670941015084112e-06,
+      "loss": 0.8965,
+      "step": 9596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670289176122735e-06,
+      "loss": 0.9038,
+      "step": 9597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.669637305206094e-06,
+      "loss": 0.8418,
+      "step": 9598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6689854023466595e-06,
+      "loss": 0.9014,
+      "step": 9599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668333467556905e-06,
+      "loss": 0.9141,
+      "step": 9600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667681500849302e-06,
+      "loss": 0.8442,
+      "step": 9601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667029502236324e-06,
+      "loss": 0.9072,
+      "step": 9602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.666377471730445e-06,
+      "loss": 0.895,
+      "step": 9603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.665725409344139e-06,
+      "loss": 0.8496,
+      "step": 9604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.665073315089882e-06,
+      "loss": 0.8545,
+      "step": 9605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.664421188980148e-06,
+      "loss": 0.8447,
+      "step": 9606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663769031027413e-06,
+      "loss": 0.8794,
+      "step": 9607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663116841244155e-06,
+      "loss": 0.8975,
+      "step": 9608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.662464619642852e-06,
+      "loss": 0.9229,
+      "step": 9609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6618123662359805e-06,
+      "loss": 0.8799,
+      "step": 9610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661160081036019e-06,
+      "loss": 0.9028,
+      "step": 9611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.660507764055448e-06,
+      "loss": 0.8613,
+      "step": 9612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.659855415306747e-06,
+      "loss": 0.8608,
+      "step": 9613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.659203034802397e-06,
+      "loss": 0.8643,
+      "step": 9614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.658550622554878e-06,
+      "loss": 0.8711,
+      "step": 9615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657898178576671e-06,
+      "loss": 0.8706,
+      "step": 9616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657245702880261e-06,
+      "loss": 0.8794,
+      "step": 9617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.656593195478127e-06,
+      "loss": 0.856,
+      "step": 9618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.655940656382757e-06,
+      "loss": 0.9141,
+      "step": 9619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65528808560663e-06,
+      "loss": 0.9004,
+      "step": 9620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.654635483162236e-06,
+      "loss": 0.877,
+      "step": 9621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653982849062058e-06,
+      "loss": 0.8975,
+      "step": 9622
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65333018331858e-06,
+      "loss": 0.895,
+      "step": 9623
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.652677485944291e-06,
+      "loss": 0.8511,
+      "step": 9624
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.652024756951678e-06,
+      "loss": 0.8765,
+      "step": 9625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.651371996353229e-06,
+      "loss": 0.3287,
+      "step": 9626
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650719204161428e-06,
+      "loss": 0.8462,
+      "step": 9627
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650066380388771e-06,
+      "loss": 0.8843,
+      "step": 9628
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649413525047742e-06,
+      "loss": 0.8652,
+      "step": 9629
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.648760638150833e-06,
+      "loss": 0.8853,
+      "step": 9630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.648107719710536e-06,
+      "loss": 0.8872,
+      "step": 9631
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.647454769739341e-06,
+      "loss": 0.8105,
+      "step": 9632
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.646801788249739e-06,
+      "loss": 0.8916,
+      "step": 9633
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.646148775254225e-06,
+      "loss": 0.8999,
+      "step": 9634
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6454957307652914e-06,
+      "loss": 0.8789,
+      "step": 9635
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644842654795429e-06,
+      "loss": 0.7686,
+      "step": 9636
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644189547357137e-06,
+      "loss": 0.9185,
+      "step": 9637
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6435364084629085e-06,
+      "loss": 0.9028,
+      "step": 9638
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.642883238125237e-06,
+      "loss": 0.9692,
+      "step": 9639
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.642230036356619e-06,
+      "loss": 0.8486,
+      "step": 9640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.641576803169556e-06,
+      "loss": 0.835,
+      "step": 9641
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6409235385765405e-06,
+      "loss": 0.9009,
+      "step": 9642
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640270242590071e-06,
+      "loss": 0.9116,
+      "step": 9643
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.639616915222649e-06,
+      "loss": 0.9277,
+      "step": 9644
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638963556486769e-06,
+      "loss": 0.8545,
+      "step": 9645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638310166394935e-06,
+      "loss": 0.8516,
+      "step": 9646
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637656744959647e-06,
+      "loss": 0.9209,
+      "step": 9647
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637003292193403e-06,
+      "loss": 0.8677,
+      "step": 9648
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636349808108706e-06,
+      "loss": 0.8271,
+      "step": 9649
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.63569629271806e-06,
+      "loss": 0.8262,
+      "step": 9650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6350427460339676e-06,
+      "loss": 0.9033,
+      "step": 9651
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.634389168068927e-06,
+      "loss": 0.8354,
+      "step": 9652
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.633735558835449e-06,
+      "loss": 0.8145,
+      "step": 9653
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.633081918346036e-06,
+      "loss": 0.8062,
+      "step": 9654
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.632428246613192e-06,
+      "loss": 0.9082,
+      "step": 9655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6317745436494204e-06,
+      "loss": 0.8657,
+      "step": 9656
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.631120809467233e-06,
+      "loss": 0.8813,
+      "step": 9657
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.630467044079134e-06,
+      "loss": 0.8438,
+      "step": 9658
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.629813247497632e-06,
+      "loss": 0.3185,
+      "step": 9659
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.629159419735232e-06,
+      "loss": 0.855,
+      "step": 9660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628505560804447e-06,
+      "loss": 0.875,
+      "step": 9661
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.627851670717783e-06,
+      "loss": 0.8477,
+      "step": 9662
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.627197749487753e-06,
+      "loss": 0.8613,
+      "step": 9663
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.626543797126864e-06,
+      "loss": 0.855,
+      "step": 9664
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.62588981364763e-06,
+      "loss": 0.8877,
+      "step": 9665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6252357990625615e-06,
+      "loss": 0.8418,
+      "step": 9666
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.624581753384172e-06,
+      "loss": 0.8726,
+      "step": 9667
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623927676624974e-06,
+      "loss": 0.7388,
+      "step": 9668
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623273568797478e-06,
+      "loss": 0.8477,
+      "step": 9669
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622619429914202e-06,
+      "loss": 0.8452,
+      "step": 9670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621965259987661e-06,
+      "loss": 0.8857,
+      "step": 9671
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621311059030368e-06,
+      "loss": 0.8252,
+      "step": 9672
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.620656827054836e-06,
+      "loss": 0.8369,
+      "step": 9673
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.62000256407359e-06,
+      "loss": 0.9272,
+      "step": 9674
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6193482700991396e-06,
+      "loss": 0.877,
+      "step": 9675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618693945144006e-06,
+      "loss": 0.9214,
+      "step": 9676
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618039589220705e-06,
+      "loss": 0.9014,
+      "step": 9677
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.617385202341758e-06,
+      "loss": 0.875,
+      "step": 9678
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.616730784519683e-06,
+      "loss": 0.8584,
+      "step": 9679
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.616076335767001e-06,
+      "loss": 0.8643,
+      "step": 9680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.615421856096231e-06,
+      "loss": 0.8882,
+      "step": 9681
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614767345519897e-06,
+      "loss": 0.8857,
+      "step": 9682
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614112804050517e-06,
+      "loss": 0.894,
+      "step": 9683
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.613458231700618e-06,
+      "loss": 0.9116,
+      "step": 9684
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6128036284827195e-06,
+      "loss": 0.8599,
+      "step": 9685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.612148994409345e-06,
+      "loss": 0.8574,
+      "step": 9686
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6114943294930225e-06,
+      "loss": 0.8579,
+      "step": 9687
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.610839633746271e-06,
+      "loss": 0.8799,
+      "step": 9688
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.610184907181619e-06,
+      "loss": 0.8975,
+      "step": 9689
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.609530149811595e-06,
+      "loss": 0.8311,
+      "step": 9690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.608875361648721e-06,
+      "loss": 0.8521,
+      "step": 9691
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.608220542705527e-06,
+      "loss": 0.8931,
+      "step": 9692
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.60756569299454e-06,
+      "loss": 0.8423,
+      "step": 9693
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.606910812528287e-06,
+      "loss": 0.8936,
+      "step": 9694
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.606255901319298e-06,
+      "loss": 0.9243,
+      "step": 9695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.605600959380103e-06,
+      "loss": 0.8623,
+      "step": 9696
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.604945986723232e-06,
+      "loss": 0.8901,
+      "step": 9697
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.604290983361214e-06,
+      "loss": 0.8599,
+      "step": 9698
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6036359493065815e-06,
+      "loss": 0.3729,
+      "step": 9699
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.602980884571868e-06,
+      "loss": 0.8638,
+      "step": 9700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.602325789169604e-06,
+      "loss": 0.9214,
+      "step": 9701
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601670663112321e-06,
+      "loss": 0.8877,
+      "step": 9702
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601015506412555e-06,
+      "loss": 0.8306,
+      "step": 9703
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.600360319082839e-06,
+      "loss": 0.8667,
+      "step": 9704
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599705101135709e-06,
+      "loss": 0.8633,
+      "step": 9705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599049852583699e-06,
+      "loss": 0.8936,
+      "step": 9706
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.598394573439347e-06,
+      "loss": 0.8511,
+      "step": 9707
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597739263715186e-06,
+      "loss": 0.8726,
+      "step": 9708
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597083923423756e-06,
+      "loss": 0.8931,
+      "step": 9709
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.596428552577594e-06,
+      "loss": 0.8589,
+      "step": 9710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5957731511892375e-06,
+      "loss": 0.8433,
+      "step": 9711
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.595117719271227e-06,
+      "loss": 0.8408,
+      "step": 9712
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.594462256836101e-06,
+      "loss": 0.835,
+      "step": 9713
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.593806763896398e-06,
+      "loss": 0.3301,
+      "step": 9714
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.593151240464661e-06,
+      "loss": 0.9121,
+      "step": 9715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592495686553431e-06,
+      "loss": 0.8887,
+      "step": 9716
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.591840102175249e-06,
+      "loss": 0.9219,
+      "step": 9717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.591184487342656e-06,
+      "loss": 0.8574,
+      "step": 9718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.590528842068196e-06,
+      "loss": 0.915,
+      "step": 9719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.589873166364415e-06,
+      "loss": 0.8433,
+      "step": 9720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5892174602438555e-06,
+      "loss": 0.9175,
+      "step": 9721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.588561723719058e-06,
+      "loss": 0.9126,
+      "step": 9722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.587905956802575e-06,
+      "loss": 0.9297,
+      "step": 9723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.587250159506947e-06,
+      "loss": 0.8965,
+      "step": 9724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5865943318447236e-06,
+      "loss": 0.8135,
+      "step": 9725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.585938473828449e-06,
+      "loss": 0.8657,
+      "step": 9726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5852825854706735e-06,
+      "loss": 0.9092,
+      "step": 9727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584626666783943e-06,
+      "loss": 0.8735,
+      "step": 9728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.583970717780808e-06,
+      "loss": 0.8706,
+      "step": 9729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5833147384738176e-06,
+      "loss": 0.8408,
+      "step": 9730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.582658728875519e-06,
+      "loss": 0.9072,
+      "step": 9731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.582002688998467e-06,
+      "loss": 0.873,
+      "step": 9732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.581346618855211e-06,
+      "loss": 0.8955,
+      "step": 9733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.580690518458302e-06,
+      "loss": 0.8589,
+      "step": 9734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.58003438782029e-06,
+      "loss": 0.8599,
+      "step": 9735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.579378226953733e-06,
+      "loss": 0.8867,
+      "step": 9736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578722035871181e-06,
+      "loss": 0.8638,
+      "step": 9737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578065814585189e-06,
+      "loss": 0.8975,
+      "step": 9738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.577409563108311e-06,
+      "loss": 0.8623,
+      "step": 9739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.576753281453102e-06,
+      "loss": 0.8569,
+      "step": 9740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.576096969632119e-06,
+      "loss": 0.9189,
+      "step": 9741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.575440627657917e-06,
+      "loss": 0.9126,
+      "step": 9742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.574784255543052e-06,
+      "loss": 0.9277,
+      "step": 9743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.574127853300084e-06,
+      "loss": 0.8975,
+      "step": 9744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5734714209415685e-06,
+      "loss": 0.8955,
+      "step": 9745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.572814958480067e-06,
+      "loss": 0.8921,
+      "step": 9746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5721584659281354e-06,
+      "loss": 0.8896,
+      "step": 9747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.571501943298335e-06,
+      "loss": 0.8975,
+      "step": 9748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.570845390603228e-06,
+      "loss": 0.8931,
+      "step": 9749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.570188807855371e-06,
+      "loss": 0.8945,
+      "step": 9750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.569532195067328e-06,
+      "loss": 0.79,
+      "step": 9751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.568875552251662e-06,
+      "loss": 0.9165,
+      "step": 9752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5682188794209335e-06,
+      "loss": 0.4141,
+      "step": 9753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5675621765877075e-06,
+      "loss": 0.8882,
+      "step": 9754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5669054437645465e-06,
+      "loss": 0.8774,
+      "step": 9755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5662486809640144e-06,
+      "loss": 0.8813,
+      "step": 9756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.565591888198678e-06,
+      "loss": 0.8452,
+      "step": 9757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.564935065481102e-06,
+      "loss": 0.8021,
+      "step": 9758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5642782128238515e-06,
+      "loss": 0.8516,
+      "step": 9759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.563621330239494e-06,
+      "loss": 0.8037,
+      "step": 9760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562964417740596e-06,
+      "loss": 0.9448,
+      "step": 9761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562307475339726e-06,
+      "loss": 0.8989,
+      "step": 9762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.561650503049452e-06,
+      "loss": 0.8423,
+      "step": 9763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5609935008823424e-06,
+      "loss": 0.938,
+      "step": 9764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.560336468850969e-06,
+      "loss": 0.8789,
+      "step": 9765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5596794069678996e-06,
+      "loss": 0.9209,
+      "step": 9766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.559022315245705e-06,
+      "loss": 0.8491,
+      "step": 9767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5583651936969565e-06,
+      "loss": 0.9746,
+      "step": 9768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5577080423342266e-06,
+      "loss": 0.8633,
+      "step": 9769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5570508611700866e-06,
+      "loss": 0.8564,
+      "step": 9770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.556393650217111e-06,
+      "loss": 0.8525,
+      "step": 9771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5557364094878715e-06,
+      "loss": 0.8623,
+      "step": 9772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.555079138994944e-06,
+      "loss": 0.8169,
+      "step": 9773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.554421838750901e-06,
+      "loss": 0.8799,
+      "step": 9774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5537645087683184e-06,
+      "loss": 0.9126,
+      "step": 9775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553107149059774e-06,
+      "loss": 0.9014,
+      "step": 9776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.552449759637841e-06,
+      "loss": 0.8955,
+      "step": 9777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551792340515099e-06,
+      "loss": 0.9155,
+      "step": 9778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551134891704124e-06,
+      "loss": 0.896,
+      "step": 9779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.550477413217492e-06,
+      "loss": 0.8804,
+      "step": 9780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5498199050677856e-06,
+      "loss": 0.9365,
+      "step": 9781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.549162367267583e-06,
+      "loss": 0.834,
+      "step": 9782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.548504799829462e-06,
+      "loss": 0.8262,
+      "step": 9783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.547847202766003e-06,
+      "loss": 0.8848,
+      "step": 9784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.547189576089791e-06,
+      "loss": 0.9365,
+      "step": 9785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5465319198134025e-06,
+      "loss": 0.8857,
+      "step": 9786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545874233949421e-06,
+      "loss": 0.8545,
+      "step": 9787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.54521651851043e-06,
+      "loss": 0.8926,
+      "step": 9788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.544558773509012e-06,
+      "loss": 0.8735,
+      "step": 9789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.543900998957751e-06,
+      "loss": 0.8843,
+      "step": 9790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5432431948692295e-06,
+      "loss": 0.8877,
+      "step": 9791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.542585361256036e-06,
+      "loss": 0.8345,
+      "step": 9792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541927498130751e-06,
+      "loss": 0.9429,
+      "step": 9793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541269605505965e-06,
+      "loss": 0.9448,
+      "step": 9794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.540611683394265e-06,
+      "loss": 0.8809,
+      "step": 9795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.539953731808233e-06,
+      "loss": 0.875,
+      "step": 9796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.539295750760458e-06,
+      "loss": 0.8086,
+      "step": 9797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.538637740263533e-06,
+      "loss": 0.875,
+      "step": 9798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537979700330042e-06,
+      "loss": 0.8325,
+      "step": 9799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537321630972575e-06,
+      "loss": 0.8027,
+      "step": 9800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.536663532203723e-06,
+      "loss": 0.8911,
+      "step": 9801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.536005404036076e-06,
+      "loss": 0.9058,
+      "step": 9802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5353472464822264e-06,
+      "loss": 0.9336,
+      "step": 9803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.534689059554764e-06,
+      "loss": 0.9258,
+      "step": 9804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5340308432662815e-06,
+      "loss": 0.8335,
+      "step": 9805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5333725976293705e-06,
+      "loss": 0.8174,
+      "step": 9806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.532714322656627e-06,
+      "loss": 0.9614,
+      "step": 9807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.532056018360644e-06,
+      "loss": 0.8262,
+      "step": 9808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.531397684754013e-06,
+      "loss": 0.9048,
+      "step": 9809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.530739321849331e-06,
+      "loss": 0.8418,
+      "step": 9810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.530080929659195e-06,
+      "loss": 0.8774,
+      "step": 9811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5294225081961995e-06,
+      "loss": 0.9819,
+      "step": 9812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5287640574729404e-06,
+      "loss": 0.8379,
+      "step": 9813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.528105577502017e-06,
+      "loss": 0.9678,
+      "step": 9814
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.527447068296026e-06,
+      "loss": 0.8965,
+      "step": 9815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526788529867565e-06,
+      "loss": 0.9248,
+      "step": 9816
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526129962229233e-06,
+      "loss": 0.8423,
+      "step": 9817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5254713653936304e-06,
+      "loss": 0.8525,
+      "step": 9818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.524812739373357e-06,
+      "loss": 0.8335,
+      "step": 9819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5241540841810135e-06,
+      "loss": 0.9092,
+      "step": 9820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5234953998292e-06,
+      "loss": 0.8457,
+      "step": 9821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522836686330518e-06,
+      "loss": 0.873,
+      "step": 9822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522177943697572e-06,
+      "loss": 0.9106,
+      "step": 9823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.521519171942962e-06,
+      "loss": 0.8931,
+      "step": 9824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.520860371079294e-06,
+      "loss": 0.9053,
+      "step": 9825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5202015411191685e-06,
+      "loss": 0.8843,
+      "step": 9826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.519542682075194e-06,
+      "loss": 0.2805,
+      "step": 9827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.518883793959973e-06,
+      "loss": 0.3665,
+      "step": 9828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.518224876786112e-06,
+      "loss": 0.8574,
+      "step": 9829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.517565930566214e-06,
+      "loss": 0.9077,
+      "step": 9830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.516906955312892e-06,
+      "loss": 0.8794,
+      "step": 9831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.516247951038748e-06,
+      "loss": 0.9092,
+      "step": 9832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5155889177563905e-06,
+      "loss": 0.8955,
+      "step": 9833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.51492985547843e-06,
+      "loss": 0.9209,
+      "step": 9834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.514270764217474e-06,
+      "loss": 0.8428,
+      "step": 9835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.513611643986131e-06,
+      "loss": 0.9233,
+      "step": 9836
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.512952494797014e-06,
+      "loss": 0.854,
+      "step": 9837
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.51229331666273e-06,
+      "loss": 0.9189,
+      "step": 9838
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5116341095958915e-06,
+      "loss": 0.877,
+      "step": 9839
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.510974873609111e-06,
+      "loss": 0.9399,
+      "step": 9840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.510315608715002e-06,
+      "loss": 0.9409,
+      "step": 9841
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.509656314926172e-06,
+      "loss": 0.876,
+      "step": 9842
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50899699225524e-06,
+      "loss": 0.9038,
+      "step": 9843
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.508337640714818e-06,
+      "loss": 0.8276,
+      "step": 9844
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5076782603175195e-06,
+      "loss": 0.9155,
+      "step": 9845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.507018851075959e-06,
+      "loss": 0.8721,
+      "step": 9846
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5063594130027555e-06,
+      "loss": 0.8696,
+      "step": 9847
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505699946110521e-06,
+      "loss": 0.8276,
+      "step": 9848
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5050404504118735e-06,
+      "loss": 0.8765,
+      "step": 9849
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5043809259194315e-06,
+      "loss": 0.8896,
+      "step": 9850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50372137264581e-06,
+      "loss": 0.9131,
+      "step": 9851
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5030617906036305e-06,
+      "loss": 0.9268,
+      "step": 9852
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5024021798055116e-06,
+      "loss": 0.8491,
+      "step": 9853
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50174254026407e-06,
+      "loss": 0.8901,
+      "step": 9854
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.501082871991927e-06,
+      "loss": 0.8818,
+      "step": 9855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.500423175001705e-06,
+      "loss": 0.8687,
+      "step": 9856
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499763449306023e-06,
+      "loss": 0.9058,
+      "step": 9857
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499103694917499e-06,
+      "loss": 0.936,
+      "step": 9858
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.498443911848762e-06,
+      "loss": 0.8433,
+      "step": 9859
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497784100112433e-06,
+      "loss": 0.9614,
+      "step": 9860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4971242597211325e-06,
+      "loss": 0.8413,
+      "step": 9861
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.496464390687484e-06,
+      "loss": 0.9014,
+      "step": 9862
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.495804493024117e-06,
+      "loss": 0.9131,
+      "step": 9863
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.495144566743652e-06,
+      "loss": 0.8828,
+      "step": 9864
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.494484611858713e-06,
+      "loss": 0.8486,
+      "step": 9865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.493824628381931e-06,
+      "loss": 0.8462,
+      "step": 9866
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4931646163259284e-06,
+      "loss": 0.8896,
+      "step": 9867
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.492504575703334e-06,
+      "loss": 0.936,
+      "step": 9868
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.491844506526776e-06,
+      "loss": 0.8506,
+      "step": 9869
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4911844088088826e-06,
+      "loss": 0.8691,
+      "step": 9870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.490524282562278e-06,
+      "loss": 0.8628,
+      "step": 9871
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4898641277995986e-06,
+      "loss": 0.8857,
+      "step": 9872
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.48920394453347e-06,
+      "loss": 0.8301,
+      "step": 9873
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.488543732776522e-06,
+      "loss": 0.9097,
+      "step": 9874
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.487883492541387e-06,
+      "loss": 0.8945,
+      "step": 9875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4872232238406975e-06,
+      "loss": 0.8765,
+      "step": 9876
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.486562926687083e-06,
+      "loss": 0.8794,
+      "step": 9877
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.485902601093178e-06,
+      "loss": 0.9004,
+      "step": 9878
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.485242247071612e-06,
+      "loss": 0.8955,
+      "step": 9879
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.484581864635026e-06,
+      "loss": 0.9009,
+      "step": 9880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.483921453796046e-06,
+      "loss": 0.8735,
+      "step": 9881
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.483261014567311e-06,
+      "loss": 0.8994,
+      "step": 9882
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.482600546961456e-06,
+      "loss": 0.9053,
+      "step": 9883
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.481940050991116e-06,
+      "loss": 0.8315,
+      "step": 9884
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.481279526668928e-06,
+      "loss": 0.8784,
+      "step": 9885
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4806189740075274e-06,
+      "loss": 0.9106,
+      "step": 9886
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4799583930195546e-06,
+      "loss": 0.7856,
+      "step": 9887
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.479297783717643e-06,
+      "loss": 0.8887,
+      "step": 9888
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478637146114435e-06,
+      "loss": 0.8931,
+      "step": 9889
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.47797648022257e-06,
+      "loss": 0.8687,
+      "step": 9890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.477315786054683e-06,
+      "loss": 0.8062,
+      "step": 9891
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.476655063623418e-06,
+      "loss": 0.7964,
+      "step": 9892
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.475994312941416e-06,
+      "loss": 0.3051,
+      "step": 9893
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.475333534021316e-06,
+      "loss": 0.8906,
+      "step": 9894
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.47467272687576e-06,
+      "loss": 0.8359,
+      "step": 9895
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4740118915173925e-06,
+      "loss": 0.9185,
+      "step": 9896
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.473351027958854e-06,
+      "loss": 0.8242,
+      "step": 9897
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472690136212788e-06,
+      "loss": 0.8545,
+      "step": 9898
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472029216291839e-06,
+      "loss": 0.8862,
+      "step": 9899
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.471368268208652e-06,
+      "loss": 0.8545,
+      "step": 9900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470707291975872e-06,
+      "loss": 0.8931,
+      "step": 9901
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470046287606143e-06,
+      "loss": 0.353,
+      "step": 9902
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.469385255112113e-06,
+      "loss": 0.8613,
+      "step": 9903
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4687241945064265e-06,
+      "loss": 0.8589,
+      "step": 9904
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468063105801732e-06,
+      "loss": 0.9014,
+      "step": 9905
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.467401989010677e-06,
+      "loss": 0.8579,
+      "step": 9906
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466740844145911e-06,
+      "loss": 0.8716,
+      "step": 9907
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466079671220078e-06,
+      "loss": 0.8667,
+      "step": 9908
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.465418470245835e-06,
+      "loss": 0.8877,
+      "step": 9909
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.464757241235824e-06,
+      "loss": 0.8716,
+      "step": 9910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4640959842027e-06,
+      "loss": 0.8936,
+      "step": 9911
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.463434699159112e-06,
+      "loss": 0.8701,
+      "step": 9912
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462773386117713e-06,
+      "loss": 0.854,
+      "step": 9913
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462112045091154e-06,
+      "loss": 0.8511,
+      "step": 9914
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.461450676092088e-06,
+      "loss": 0.9282,
+      "step": 9915
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.460789279133167e-06,
+      "loss": 0.8423,
+      "step": 9916
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.460127854227045e-06,
+      "loss": 0.8789,
+      "step": 9917
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.459466401386376e-06,
+      "loss": 0.856,
+      "step": 9918
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.458804920623817e-06,
+      "loss": 0.8818,
+      "step": 9919
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.458143411952018e-06,
+      "loss": 0.9292,
+      "step": 9920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.45748187538364e-06,
+      "loss": 0.8667,
+      "step": 9921
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.456820310931336e-06,
+      "loss": 0.8818,
+      "step": 9922
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4561587186077644e-06,
+      "loss": 0.8481,
+      "step": 9923
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.45549709842558e-06,
+      "loss": 0.8521,
+      "step": 9924
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.454835450397446e-06,
+      "loss": 0.8081,
+      "step": 9925
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.454173774536016e-06,
+      "loss": 0.8535,
+      "step": 9926
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.453512070853949e-06,
+      "loss": 0.8794,
+      "step": 9927
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4528503393639066e-06,
+      "loss": 0.916,
+      "step": 9928
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4521885800785465e-06,
+      "loss": 0.9028,
+      "step": 9929
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.451526793010532e-06,
+      "loss": 0.8594,
+      "step": 9930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.450864978172522e-06,
+      "loss": 0.8574,
+      "step": 9931
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.45020313557718e-06,
+      "loss": 0.8096,
+      "step": 9932
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.449541265237163e-06,
+      "loss": 0.8015,
+      "step": 9933
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4488793671651405e-06,
+      "loss": 0.8516,
+      "step": 9934
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.448217441373773e-06,
+      "loss": 0.8289,
+      "step": 9935
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447555487875723e-06,
+      "loss": 0.9355,
+      "step": 9936
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.446893506683654e-06,
+      "loss": 0.8896,
+      "step": 9937
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.446231497810233e-06,
+      "loss": 0.8882,
+      "step": 9938
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.445569461268126e-06,
+      "loss": 0.8848,
+      "step": 9939
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.444907397069996e-06,
+      "loss": 0.8472,
+      "step": 9940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4442453052285096e-06,
+      "loss": 0.8516,
+      "step": 9941
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.443583185756337e-06,
+      "loss": 0.8032,
+      "step": 9942
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4429210386661415e-06,
+      "loss": 0.8936,
+      "step": 9943
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.442258863970594e-06,
+      "loss": 0.8979,
+      "step": 9944
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.441596661682361e-06,
+      "loss": 0.875,
+      "step": 9945
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.440934431814112e-06,
+      "loss": 0.8613,
+      "step": 9946
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.440272174378517e-06,
+      "loss": 0.896,
+      "step": 9947
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.439609889388245e-06,
+      "loss": 0.9116,
+      "step": 9948
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4389475768559675e-06,
+      "loss": 0.9053,
+      "step": 9949
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.438285236794356e-06,
+      "loss": 0.8818,
+      "step": 9950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.437622869216081e-06,
+      "loss": 0.8975,
+      "step": 9951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436960474133817e-06,
+      "loss": 0.856,
+      "step": 9952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436298051560231e-06,
+      "loss": 0.9277,
+      "step": 9953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.435635601508003e-06,
+      "loss": 0.9048,
+      "step": 9954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434973123989803e-06,
+      "loss": 0.8843,
+      "step": 9955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434310619018306e-06,
+      "loss": 0.8535,
+      "step": 9956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.433648086606185e-06,
+      "loss": 0.9629,
+      "step": 9957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432985526766119e-06,
+      "loss": 0.8945,
+      "step": 9958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432322939510781e-06,
+      "loss": 0.8457,
+      "step": 9959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.431660324852847e-06,
+      "loss": 0.8706,
+      "step": 9960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430997682804997e-06,
+      "loss": 0.8677,
+      "step": 9961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430335013379904e-06,
+      "loss": 0.8457,
+      "step": 9962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.429672316590251e-06,
+      "loss": 0.4109,
+      "step": 9963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.429009592448711e-06,
+      "loss": 0.832,
+      "step": 9964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.428346840967969e-06,
+      "loss": 0.896,
+      "step": 9965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4276840621606985e-06,
+      "loss": 0.8394,
+      "step": 9966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.427021256039583e-06,
+      "loss": 0.8716,
+      "step": 9967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4263584226173025e-06,
+      "loss": 0.8843,
+      "step": 9968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4256955619065375e-06,
+      "loss": 0.9097,
+      "step": 9969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.425032673919967e-06,
+      "loss": 0.9336,
+      "step": 9970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4243697586702805e-06,
+      "loss": 0.8354,
+      "step": 9971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.423706816170153e-06,
+      "loss": 0.9116,
+      "step": 9972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.423043846432272e-06,
+      "loss": 0.8428,
+      "step": 9973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422380849469318e-06,
+      "loss": 0.8794,
+      "step": 9974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.421717825293979e-06,
+      "loss": 0.8687,
+      "step": 9975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.421054773918936e-06,
+      "loss": 0.8755,
+      "step": 9976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.420391695356878e-06,
+      "loss": 0.8809,
+      "step": 9977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.419728589620486e-06,
+      "loss": 0.8711,
+      "step": 9978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.419065456722451e-06,
+      "loss": 0.8774,
+      "step": 9979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.418402296675456e-06,
+      "loss": 0.9077,
+      "step": 9980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.417739109492193e-06,
+      "loss": 0.8657,
+      "step": 9981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.417075895185343e-06,
+      "loss": 0.8823,
+      "step": 9982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4164126537676e-06,
+      "loss": 0.833,
+      "step": 9983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.41574938525165e-06,
+      "loss": 0.8584,
+      "step": 9984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.415086089650185e-06,
+      "loss": 0.4242,
+      "step": 9985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.414422766975891e-06,
+      "loss": 0.8555,
+      "step": 9986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413759417241463e-06,
+      "loss": 0.8486,
+      "step": 9987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413096040459588e-06,
+      "loss": 0.877,
+      "step": 9988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.412432636642959e-06,
+      "loss": 0.8696,
+      "step": 9989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.411769205804267e-06,
+      "loss": 0.9067,
+      "step": 9990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.411105747956208e-06,
+      "loss": 0.9297,
+      "step": 9991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4104422631114706e-06,
+      "loss": 0.8994,
+      "step": 9992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40977875128275e-06,
+      "loss": 0.9116,
+      "step": 9993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.409115212482742e-06,
+      "loss": 0.8564,
+      "step": 9994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.408451646724138e-06,
+      "loss": 0.8594,
+      "step": 9995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.407788054019635e-06,
+      "loss": 0.8755,
+      "step": 9996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4071244343819275e-06,
+      "loss": 0.8789,
+      "step": 9997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.406460787823714e-06,
+      "loss": 0.8545,
+      "step": 9998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405797114357687e-06,
+      "loss": 0.7764,
+      "step": 9999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405133413996547e-06,
+      "loss": 0.8398,
+      "step": 10000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.404469686752992e-06,
+      "loss": 0.9233,
+      "step": 10001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.403805932639717e-06,
+      "loss": 0.875,
+      "step": 10002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.403142151669422e-06,
+      "loss": 0.8901,
+      "step": 10003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.402478343854808e-06,
+      "loss": 0.936,
+      "step": 10004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.401814509208573e-06,
+      "loss": 0.4233,
+      "step": 10005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.401150647743416e-06,
+      "loss": 0.8701,
+      "step": 10006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4004867594720395e-06,
+      "loss": 0.894,
+      "step": 10007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.399822844407145e-06,
+      "loss": 0.8584,
+      "step": 10008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3991589025614345e-06,
+      "loss": 0.9233,
+      "step": 10009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.398494933947608e-06,
+      "loss": 0.8711,
+      "step": 10010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.397830938578371e-06,
+      "loss": 0.8081,
+      "step": 10011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.397166916466424e-06,
+      "loss": 0.8994,
+      "step": 10012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.396502867624473e-06,
+      "loss": 0.9414,
+      "step": 10013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.395838792065223e-06,
+      "loss": 0.8423,
+      "step": 10014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.395174689801375e-06,
+      "loss": 0.8853,
+      "step": 10015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.394510560845637e-06,
+      "loss": 0.8032,
+      "step": 10016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.393846405210715e-06,
+      "loss": 0.9375,
+      "step": 10017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.393182222909315e-06,
+      "loss": 0.8887,
+      "step": 10018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.392518013954143e-06,
+      "loss": 0.8989,
+      "step": 10019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.391853778357908e-06,
+      "loss": 0.8765,
+      "step": 10020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.391189516133316e-06,
+      "loss": 0.8228,
+      "step": 10021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.390525227293077e-06,
+      "loss": 0.8442,
+      "step": 10022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.389860911849899e-06,
+      "loss": 0.9473,
+      "step": 10023
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.389196569816492e-06,
+      "loss": 0.8813,
+      "step": 10024
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.388532201205565e-06,
+      "loss": 0.8613,
+      "step": 10025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387867806029828e-06,
+      "loss": 0.9224,
+      "step": 10026
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387203384301994e-06,
+      "loss": 0.8706,
+      "step": 10027
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.386538936034774e-06,
+      "loss": 0.8853,
+      "step": 10028
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.385874461240878e-06,
+      "loss": 0.9102,
+      "step": 10029
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.385209959933021e-06,
+      "loss": 0.8726,
+      "step": 10030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384545432123914e-06,
+      "loss": 0.8545,
+      "step": 10031
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38388087782627e-06,
+      "loss": 0.8496,
+      "step": 10032
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.383216297052805e-06,
+      "loss": 0.937,
+      "step": 10033
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.382551689816233e-06,
+      "loss": 0.8584,
+      "step": 10034
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.381887056129268e-06,
+      "loss": 0.8633,
+      "step": 10035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.381222396004627e-06,
+      "loss": 0.9189,
+      "step": 10036
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3805577094550246e-06,
+      "loss": 0.8804,
+      "step": 10037
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.379892996493176e-06,
+      "loss": 0.874,
+      "step": 10038
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.379228257131801e-06,
+      "loss": 0.9033,
+      "step": 10039
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.378563491383616e-06,
+      "loss": 0.9346,
+      "step": 10040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.37789869926134e-06,
+      "loss": 0.8374,
+      "step": 10041
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.377233880777689e-06,
+      "loss": 0.9248,
+      "step": 10042
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.376569035945386e-06,
+      "loss": 0.8823,
+      "step": 10043
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375904164777146e-06,
+      "loss": 0.8564,
+      "step": 10044
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375239267285691e-06,
+      "loss": 0.7981,
+      "step": 10045
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.374574343483744e-06,
+      "loss": 0.9058,
+      "step": 10046
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3739093933840225e-06,
+      "loss": 0.8462,
+      "step": 10047
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.373244416999248e-06,
+      "loss": 0.8862,
+      "step": 10048
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.372579414342145e-06,
+      "loss": 0.8848,
+      "step": 10049
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371914385425435e-06,
+      "loss": 0.8379,
+      "step": 10050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371249330261842e-06,
+      "loss": 0.9082,
+      "step": 10051
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.370584248864085e-06,
+      "loss": 0.8677,
+      "step": 10052
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369919141244895e-06,
+      "loss": 0.8887,
+      "step": 10053
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369254007416992e-06,
+      "loss": 0.8955,
+      "step": 10054
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.368588847393102e-06,
+      "loss": 0.8672,
+      "step": 10055
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.367923661185951e-06,
+      "loss": 0.8652,
+      "step": 10056
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.367258448808264e-06,
+      "loss": 0.8755,
+      "step": 10057
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.366593210272769e-06,
+      "loss": 0.9204,
+      "step": 10058
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.365927945592192e-06,
+      "loss": 0.8687,
+      "step": 10059
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.36526265477926e-06,
+      "loss": 0.9199,
+      "step": 10060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.364597337846705e-06,
+      "loss": 0.9062,
+      "step": 10061
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.363931994807251e-06,
+      "loss": 0.853,
+      "step": 10062
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3632666256736284e-06,
+      "loss": 0.8604,
+      "step": 10063
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.362601230458567e-06,
+      "loss": 0.8921,
+      "step": 10064
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361935809174796e-06,
+      "loss": 0.894,
+      "step": 10065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361270361835049e-06,
+      "loss": 0.8667,
+      "step": 10066
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.360604888452054e-06,
+      "loss": 0.8647,
+      "step": 10067
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.359939389038543e-06,
+      "loss": 0.8677,
+      "step": 10068
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.359273863607249e-06,
+      "loss": 0.8198,
+      "step": 10069
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.358608312170903e-06,
+      "loss": 0.8589,
+      "step": 10070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.35794273474224e-06,
+      "loss": 0.853,
+      "step": 10071
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357277131333992e-06,
+      "loss": 0.8672,
+      "step": 10072
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.356611501958895e-06,
+      "loss": 0.9282,
+      "step": 10073
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.355945846629681e-06,
+      "loss": 0.9033,
+      "step": 10074
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.355280165359086e-06,
+      "loss": 0.8403,
+      "step": 10075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.354614458159847e-06,
+      "loss": 0.938,
+      "step": 10076
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353948725044698e-06,
+      "loss": 0.9062,
+      "step": 10077
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353282966026376e-06,
+      "loss": 0.854,
+      "step": 10078
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.352617181117619e-06,
+      "loss": 0.9321,
+      "step": 10079
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351951370331164e-06,
+      "loss": 0.8491,
+      "step": 10080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351285533679748e-06,
+      "loss": 0.8369,
+      "step": 10081
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.350619671176111e-06,
+      "loss": 0.9844,
+      "step": 10082
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.349953782832991e-06,
+      "loss": 0.3839,
+      "step": 10083
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3492878686631285e-06,
+      "loss": 0.2958,
+      "step": 10084
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.348621928679262e-06,
+      "loss": 0.8545,
+      "step": 10085
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.347955962894135e-06,
+      "loss": 0.8867,
+      "step": 10086
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.347289971320484e-06,
+      "loss": 0.835,
+      "step": 10087
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.346623953971054e-06,
+      "loss": 0.9507,
+      "step": 10088
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.345957910858585e-06,
+      "loss": 0.9067,
+      "step": 10089
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.34529184199582e-06,
+      "loss": 0.8936,
+      "step": 10090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.344625747395503e-06,
+      "loss": 0.8281,
+      "step": 10091
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.343959627070377e-06,
+      "loss": 0.9429,
+      "step": 10092
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.343293481033184e-06,
+      "loss": 0.8857,
+      "step": 10093
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.34262730929667e-06,
+      "loss": 0.958,
+      "step": 10094
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.341961111873581e-06,
+      "loss": 0.8237,
+      "step": 10095
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.341294888776661e-06,
+      "loss": 0.8809,
+      "step": 10096
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3406286400186555e-06,
+      "loss": 0.8804,
+      "step": 10097
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3399623656123125e-06,
+      "loss": 0.8599,
+      "step": 10098
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.339296065570378e-06,
+      "loss": 0.8828,
+      "step": 10099
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.338629739905597e-06,
+      "loss": 0.8813,
+      "step": 10100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337963388630723e-06,
+      "loss": 0.8916,
+      "step": 10101
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337297011758498e-06,
+      "loss": 0.8721,
+      "step": 10102
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.336630609301675e-06,
+      "loss": 0.8691,
+      "step": 10103
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3359641812730025e-06,
+      "loss": 0.8916,
+      "step": 10104
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.33529772768523e-06,
+      "loss": 0.8628,
+      "step": 10105
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.334631248551108e-06,
+      "loss": 0.8599,
+      "step": 10106
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333964743883386e-06,
+      "loss": 0.853,
+      "step": 10107
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333298213694817e-06,
+      "loss": 0.9189,
+      "step": 10108
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.332631657998154e-06,
+      "loss": 0.8965,
+      "step": 10109
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331965076806144e-06,
+      "loss": 0.3879,
+      "step": 10110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331298470131545e-06,
+      "loss": 0.8647,
+      "step": 10111
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.330631837987109e-06,
+      "loss": 0.8848,
+      "step": 10112
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329965180385589e-06,
+      "loss": 0.8589,
+      "step": 10113
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329298497339738e-06,
+      "loss": 0.9087,
+      "step": 10114
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.328631788862313e-06,
+      "loss": 0.8867,
+      "step": 10115
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327965054966067e-06,
+      "loss": 0.895,
+      "step": 10116
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3272982956637586e-06,
+      "loss": 0.8726,
+      "step": 10117
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.326631510968142e-06,
+      "loss": 0.9048,
+      "step": 10118
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3259647008919735e-06,
+      "loss": 0.9473,
+      "step": 10119
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3252978654480106e-06,
+      "loss": 0.8657,
+      "step": 10120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.324631004649011e-06,
+      "loss": 0.8867,
+      "step": 10121
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323964118507734e-06,
+      "loss": 0.9019,
+      "step": 10122
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3232972070369355e-06,
+      "loss": 0.8809,
+      "step": 10123
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.322630270249377e-06,
+      "loss": 0.8892,
+      "step": 10124
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321963308157819e-06,
+      "loss": 0.8643,
+      "step": 10125
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321296320775017e-06,
+      "loss": 0.8784,
+      "step": 10126
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.320629308113735e-06,
+      "loss": 0.8325,
+      "step": 10127
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3199622701867345e-06,
+      "loss": 0.875,
+      "step": 10128
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3192952070067744e-06,
+      "loss": 0.9121,
+      "step": 10129
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.318628118586618e-06,
+      "loss": 0.9414,
+      "step": 10130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317961004939029e-06,
+      "loss": 0.8647,
+      "step": 10131
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317293866076769e-06,
+      "loss": 0.9062,
+      "step": 10132
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.316626702012601e-06,
+      "loss": 0.8701,
+      "step": 10133
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3159595127592895e-06,
+      "loss": 0.8677,
+      "step": 10134
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3152922983295984e-06,
+      "loss": 0.9287,
+      "step": 10135
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.314625058736293e-06,
+      "loss": 0.8501,
+      "step": 10136
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.31395779399214e-06,
+      "loss": 0.8501,
+      "step": 10137
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.313290504109904e-06,
+      "loss": 0.8647,
+      "step": 10138
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.312623189102348e-06,
+      "loss": 0.8789,
+      "step": 10139
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3119558489822455e-06,
+      "loss": 0.9609,
+      "step": 10140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.311288483762359e-06,
+      "loss": 0.8691,
+      "step": 10141
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.310621093455457e-06,
+      "loss": 0.8647,
+      "step": 10142
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.309953678074307e-06,
+      "loss": 0.8594,
+      "step": 10143
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.309286237631682e-06,
+      "loss": 0.8911,
+      "step": 10144
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308618772140346e-06,
+      "loss": 0.9175,
+      "step": 10145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.307951281613072e-06,
+      "loss": 0.877,
+      "step": 10146
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3072837660626274e-06,
+      "loss": 0.8647,
+      "step": 10147
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306616225501785e-06,
+      "loss": 0.8818,
+      "step": 10148
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.305948659943315e-06,
+      "loss": 0.8848,
+      "step": 10149
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.305281069399989e-06,
+      "loss": 0.9053,
+      "step": 10150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.30461345388458e-06,
+      "loss": 0.9268,
+      "step": 10151
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.30394581340986e-06,
+      "loss": 0.9204,
+      "step": 10152
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.303278147988601e-06,
+      "loss": 0.9038,
+      "step": 10153
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302610457633579e-06,
+      "loss": 0.8257,
+      "step": 10154
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.301942742357563e-06,
+      "loss": 0.8872,
+      "step": 10155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3012750021733325e-06,
+      "loss": 0.8984,
+      "step": 10156
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3006072370936625e-06,
+      "loss": 0.8701,
+      "step": 10157
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.299939447131325e-06,
+      "loss": 0.8247,
+      "step": 10158
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.299271632299095e-06,
+      "loss": 0.9346,
+      "step": 10159
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.298603792609754e-06,
+      "loss": 0.9639,
+      "step": 10160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.297935928076077e-06,
+      "loss": 0.7781,
+      "step": 10161
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2972680387108395e-06,
+      "loss": 0.957,
+      "step": 10162
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.296600124526819e-06,
+      "loss": 0.9106,
+      "step": 10163
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2959321855367964e-06,
+      "loss": 0.8862,
+      "step": 10164
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.295264221753549e-06,
+      "loss": 0.9258,
+      "step": 10165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.294596233189855e-06,
+      "loss": 0.8857,
+      "step": 10166
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293928219858495e-06,
+      "loss": 0.8403,
+      "step": 10167
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293260181772249e-06,
+      "loss": 0.8584,
+      "step": 10168
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.292592118943899e-06,
+      "loss": 0.8579,
+      "step": 10169
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.291924031386223e-06,
+      "loss": 0.8877,
+      "step": 10170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.291255919112007e-06,
+      "loss": 0.8574,
+      "step": 10171
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.290587782134028e-06,
+      "loss": 0.9028,
+      "step": 10172
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289919620465073e-06,
+      "loss": 0.8965,
+      "step": 10173
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289251434117922e-06,
+      "loss": 0.8677,
+      "step": 10174
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.28858322310536e-06,
+      "loss": 0.8696,
+      "step": 10175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287914987440169e-06,
+      "loss": 0.8872,
+      "step": 10176
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287246727135137e-06,
+      "loss": 0.8618,
+      "step": 10177
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.286578442203045e-06,
+      "loss": 0.8604,
+      "step": 10178
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.285910132656681e-06,
+      "loss": 0.8462,
+      "step": 10179
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2852417985088286e-06,
+      "loss": 0.877,
+      "step": 10180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.284573439772275e-06,
+      "loss": 0.8408,
+      "step": 10181
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283905056459809e-06,
+      "loss": 0.8623,
+      "step": 10182
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283236648584216e-06,
+      "loss": 0.8765,
+      "step": 10183
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.282568216158282e-06,
+      "loss": 0.8223,
+      "step": 10184
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2818997591947985e-06,
+      "loss": 0.8848,
+      "step": 10185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.281231277706552e-06,
+      "loss": 0.8789,
+      "step": 10186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.280562771706333e-06,
+      "loss": 0.9453,
+      "step": 10187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279894241206929e-06,
+      "loss": 0.8638,
+      "step": 10188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279225686221131e-06,
+      "loss": 0.8247,
+      "step": 10189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.278557106761732e-06,
+      "loss": 0.9404,
+      "step": 10190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.277888502841519e-06,
+      "loss": 0.8311,
+      "step": 10191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2772198744732844e-06,
+      "loss": 0.8711,
+      "step": 10192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2765512216698245e-06,
+      "loss": 0.8696,
+      "step": 10193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2758825444439254e-06,
+      "loss": 0.916,
+      "step": 10194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.275213842808383e-06,
+      "loss": 0.3152,
+      "step": 10195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.274545116775991e-06,
+      "loss": 0.8623,
+      "step": 10196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.273876366359541e-06,
+      "loss": 0.8813,
+      "step": 10197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.273207591571831e-06,
+      "loss": 0.8877,
+      "step": 10198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.272538792425651e-06,
+      "loss": 0.8442,
+      "step": 10199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.271869968933802e-06,
+      "loss": 0.8955,
+      "step": 10200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.271201121109072e-06,
+      "loss": 0.8579,
+      "step": 10201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270532248964263e-06,
+      "loss": 0.8442,
+      "step": 10202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.269863352512171e-06,
+      "loss": 0.9414,
+      "step": 10203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2691944317655914e-06,
+      "loss": 0.8726,
+      "step": 10204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.268525486737322e-06,
+      "loss": 0.8418,
+      "step": 10205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267856517440163e-06,
+      "loss": 0.8647,
+      "step": 10206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267187523886908e-06,
+      "loss": 0.8872,
+      "step": 10207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.266518506090361e-06,
+      "loss": 0.9253,
+      "step": 10208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.265849464063318e-06,
+      "loss": 0.8755,
+      "step": 10209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26518039781858e-06,
+      "loss": 0.8784,
+      "step": 10210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.264511307368948e-06,
+      "loss": 0.8965,
+      "step": 10211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263842192727222e-06,
+      "loss": 0.8491,
+      "step": 10212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263173053906203e-06,
+      "loss": 0.792,
+      "step": 10213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.262503890918693e-06,
+      "loss": 0.9043,
+      "step": 10214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.261834703777494e-06,
+      "loss": 0.8477,
+      "step": 10215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.261165492495409e-06,
+      "loss": 0.915,
+      "step": 10216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26049625708524e-06,
+      "loss": 0.9258,
+      "step": 10217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.259826997559792e-06,
+      "loss": 0.8774,
+      "step": 10218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.259157713931868e-06,
+      "loss": 0.9116,
+      "step": 10219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.258488406214274e-06,
+      "loss": 0.9409,
+      "step": 10220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2578190744198115e-06,
+      "loss": 0.9004,
+      "step": 10221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.25714971856129e-06,
+      "loss": 0.8037,
+      "step": 10222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.256480338651513e-06,
+      "loss": 0.8618,
+      "step": 10223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255810934703287e-06,
+      "loss": 0.8267,
+      "step": 10224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255141506729417e-06,
+      "loss": 0.8481,
+      "step": 10225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.254472054742715e-06,
+      "loss": 0.8276,
+      "step": 10226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2538025787559845e-06,
+      "loss": 0.7998,
+      "step": 10227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2531330787820344e-06,
+      "loss": 0.812,
+      "step": 10228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.252463554833674e-06,
+      "loss": 0.8389,
+      "step": 10229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251794006923713e-06,
+      "loss": 0.8672,
+      "step": 10230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251124435064959e-06,
+      "loss": 0.8398,
+      "step": 10231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.250454839270224e-06,
+      "loss": 0.9048,
+      "step": 10232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.249785219552318e-06,
+      "loss": 0.875,
+      "step": 10233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.24911557592405e-06,
+      "loss": 0.8657,
+      "step": 10234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.248445908398233e-06,
+      "loss": 0.895,
+      "step": 10235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.247776216987678e-06,
+      "loss": 0.8682,
+      "step": 10236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2471065017051995e-06,
+      "loss": 0.8813,
+      "step": 10237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.246436762563606e-06,
+      "loss": 0.9209,
+      "step": 10238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245766999575715e-06,
+      "loss": 0.8701,
+      "step": 10239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245097212754337e-06,
+      "loss": 0.894,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.244427402112288e-06,
+      "loss": 0.3325,
+      "step": 10241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.243757567662382e-06,
+      "loss": 0.8496,
+      "step": 10242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.243087709417433e-06,
+      "loss": 0.8838,
+      "step": 10243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242417827390257e-06,
+      "loss": 0.8452,
+      "step": 10244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.241747921593671e-06,
+      "loss": 0.854,
+      "step": 10245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.241077992040491e-06,
+      "loss": 0.8623,
+      "step": 10246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.240408038743532e-06,
+      "loss": 0.8701,
+      "step": 10247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.239738061715614e-06,
+      "loss": 0.8853,
+      "step": 10248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.239068060969553e-06,
+      "loss": 0.9038,
+      "step": 10249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.238398036518165e-06,
+      "loss": 0.9082,
+      "step": 10250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.237727988374273e-06,
+      "loss": 0.8286,
+      "step": 10251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.237057916550696e-06,
+      "loss": 0.874,
+      "step": 10252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23638782106025e-06,
+      "loss": 0.8257,
+      "step": 10253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.235717701915755e-06,
+      "loss": 0.8911,
+      "step": 10254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.235047559130036e-06,
+      "loss": 0.8511,
+      "step": 10255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.234377392715909e-06,
+      "loss": 0.9365,
+      "step": 10256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.233707202686199e-06,
+      "loss": 0.8809,
+      "step": 10257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2330369890537244e-06,
+      "loss": 0.8789,
+      "step": 10258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23236675183131e-06,
+      "loss": 0.8965,
+      "step": 10259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.231696491031778e-06,
+      "loss": 0.8301,
+      "step": 10260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23102620666795e-06,
+      "loss": 0.9292,
+      "step": 10261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230355898752653e-06,
+      "loss": 0.8345,
+      "step": 10262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.229685567298707e-06,
+      "loss": 0.8853,
+      "step": 10263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.229015212318938e-06,
+      "loss": 0.8574,
+      "step": 10264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2283448338261735e-06,
+      "loss": 0.8535,
+      "step": 10265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2276744318332346e-06,
+      "loss": 0.8506,
+      "step": 10266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22700400635295e-06,
+      "loss": 0.894,
+      "step": 10267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.226333557398146e-06,
+      "loss": 0.8896,
+      "step": 10268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2256630849816475e-06,
+      "loss": 0.9062,
+      "step": 10269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.224992589116283e-06,
+      "loss": 0.8721,
+      "step": 10270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.224322069814879e-06,
+      "loss": 0.8672,
+      "step": 10271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.223651527090266e-06,
+      "loss": 0.8677,
+      "step": 10272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.222980960955272e-06,
+      "loss": 0.2837,
+      "step": 10273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.222310371422723e-06,
+      "loss": 0.8608,
+      "step": 10274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.221639758505451e-06,
+      "loss": 0.8945,
+      "step": 10275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.220969122216285e-06,
+      "loss": 0.8379,
+      "step": 10276
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.220298462568056e-06,
+      "loss": 0.8613,
+      "step": 10277
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.219627779573595e-06,
+      "loss": 0.9033,
+      "step": 10278
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2189570732457324e-06,
+      "loss": 0.894,
+      "step": 10279
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.218286343597299e-06,
+      "loss": 0.8701,
+      "step": 10280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.21761559064113e-06,
+      "loss": 0.8716,
+      "step": 10281
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.216944814390056e-06,
+      "loss": 0.8618,
+      "step": 10282
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.216274014856908e-06,
+      "loss": 0.8296,
+      "step": 10283
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.215603192054523e-06,
+      "loss": 0.7737,
+      "step": 10284
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214932345995733e-06,
+      "loss": 0.8726,
+      "step": 10285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214261476693374e-06,
+      "loss": 0.9097,
+      "step": 10286
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213590584160276e-06,
+      "loss": 0.3163,
+      "step": 10287
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.212919668409282e-06,
+      "loss": 0.8555,
+      "step": 10288
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.212248729453221e-06,
+      "loss": 0.8564,
+      "step": 10289
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.211577767304931e-06,
+      "loss": 0.8926,
+      "step": 10290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.210906781977251e-06,
+      "loss": 0.8594,
+      "step": 10291
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.210235773483016e-06,
+      "loss": 0.874,
+      "step": 10292
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2095647418350615e-06,
+      "loss": 0.9561,
+      "step": 10293
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.208893687046229e-06,
+      "loss": 0.9067,
+      "step": 10294
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.208222609129355e-06,
+      "loss": 0.9189,
+      "step": 10295
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.207551508097278e-06,
+      "loss": 0.8589,
+      "step": 10296
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206880383962839e-06,
+      "loss": 0.8647,
+      "step": 10297
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206209236738876e-06,
+      "loss": 0.8906,
+      "step": 10298
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.20553806643823e-06,
+      "loss": 0.9258,
+      "step": 10299
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204866873073739e-06,
+      "loss": 0.9043,
+      "step": 10300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204195656658248e-06,
+      "loss": 0.8477,
+      "step": 10301
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203524417204596e-06,
+      "loss": 0.9199,
+      "step": 10302
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.202853154725626e-06,
+      "loss": 0.8975,
+      "step": 10303
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.20218186923418e-06,
+      "loss": 0.812,
+      "step": 10304
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.201510560743098e-06,
+      "loss": 0.8267,
+      "step": 10305
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.200839229265228e-06,
+      "loss": 0.8589,
+      "step": 10306
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.20016787481341e-06,
+      "loss": 0.7778,
+      "step": 10307
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.19949649740049e-06,
+      "loss": 0.8384,
+      "step": 10308
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.198825097039311e-06,
+      "loss": 0.9072,
+      "step": 10309
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.198153673742719e-06,
+      "loss": 0.812,
+      "step": 10310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.197482227523562e-06,
+      "loss": 0.8628,
+      "step": 10311
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196810758394678e-06,
+      "loss": 0.8589,
+      "step": 10312
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196139266368922e-06,
+      "loss": 0.8833,
+      "step": 10313
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1954677514591366e-06,
+      "loss": 0.9111,
+      "step": 10314
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.194796213678168e-06,
+      "loss": 0.8779,
+      "step": 10315
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1941246530388645e-06,
+      "loss": 0.9116,
+      "step": 10316
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193453069554076e-06,
+      "loss": 0.9053,
+      "step": 10317
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1927814632366504e-06,
+      "loss": 0.8647,
+      "step": 10318
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1921098340994335e-06,
+      "loss": 0.8345,
+      "step": 10319
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1914381821552784e-06,
+      "loss": 0.8682,
+      "step": 10320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.190766507417033e-06,
+      "loss": 0.8613,
+      "step": 10321
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.190094809897548e-06,
+      "loss": 0.8623,
+      "step": 10322
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.189423089609674e-06,
+      "loss": 0.9082,
+      "step": 10323
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.188751346566262e-06,
+      "loss": 0.8735,
+      "step": 10324
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.188079580780163e-06,
+      "loss": 0.8477,
+      "step": 10325
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.187407792264231e-06,
+      "loss": 0.9219,
+      "step": 10326
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.186735981031317e-06,
+      "loss": 0.8218,
+      "step": 10327
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.186064147094271e-06,
+      "loss": 0.9199,
+      "step": 10328
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.185392290465949e-06,
+      "loss": 0.8228,
+      "step": 10329
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1847204111592065e-06,
+      "loss": 0.8784,
+      "step": 10330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184048509186895e-06,
+      "loss": 0.8901,
+      "step": 10331
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.18337658456187e-06,
+      "loss": 0.8813,
+      "step": 10332
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.182704637296985e-06,
+      "loss": 0.9097,
+      "step": 10333
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.182032667405098e-06,
+      "loss": 0.8926,
+      "step": 10334
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.181360674899063e-06,
+      "loss": 0.7858,
+      "step": 10335
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.180688659791736e-06,
+      "loss": 0.8525,
+      "step": 10336
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.180016622095974e-06,
+      "loss": 0.9238,
+      "step": 10337
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.179344561824636e-06,
+      "loss": 0.9185,
+      "step": 10338
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178672478990578e-06,
+      "loss": 0.8706,
+      "step": 10339
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178000373606657e-06,
+      "loss": 0.8853,
+      "step": 10340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.177328245685734e-06,
+      "loss": 0.8604,
+      "step": 10341
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.176656095240665e-06,
+      "loss": 0.9453,
+      "step": 10342
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.175983922284312e-06,
+      "loss": 0.8354,
+      "step": 10343
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.175311726829534e-06,
+      "loss": 0.8286,
+      "step": 10344
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174639508889187e-06,
+      "loss": 0.7649,
+      "step": 10345
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1739672684761375e-06,
+      "loss": 0.9414,
+      "step": 10346
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.173295005603246e-06,
+      "loss": 0.8906,
+      "step": 10347
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.17262272028337e-06,
+      "loss": 0.855,
+      "step": 10348
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.171950412529372e-06,
+      "loss": 0.916,
+      "step": 10349
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1712780823541195e-06,
+      "loss": 0.9185,
+      "step": 10350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.17060572977047e-06,
+      "loss": 0.8896,
+      "step": 10351
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1699333547912885e-06,
+      "loss": 0.8687,
+      "step": 10352
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.169260957429438e-06,
+      "loss": 0.8442,
+      "step": 10353
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.168588537697783e-06,
+      "loss": 0.8423,
+      "step": 10354
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.167916095609188e-06,
+      "loss": 0.8975,
+      "step": 10355
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.167243631176516e-06,
+      "loss": 0.8232,
+      "step": 10356
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.166571144412636e-06,
+      "loss": 0.8345,
+      "step": 10357
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1658986353304105e-06,
+      "loss": 0.9019,
+      "step": 10358
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.165226103942707e-06,
+      "loss": 0.9185,
+      "step": 10359
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164553550262394e-06,
+      "loss": 0.9072,
+      "step": 10360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.163880974302334e-06,
+      "loss": 0.9155,
+      "step": 10361
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.163208376075398e-06,
+      "loss": 0.8701,
+      "step": 10362
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.162535755594453e-06,
+      "loss": 0.7817,
+      "step": 10363
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.161863112872367e-06,
+      "loss": 0.7764,
+      "step": 10364
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.161190447922007e-06,
+      "loss": 0.896,
+      "step": 10365
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.160517760756246e-06,
+      "loss": 0.8486,
+      "step": 10366
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.159845051387951e-06,
+      "loss": 0.8511,
+      "step": 10367
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.159172319829993e-06,
+      "loss": 0.9048,
+      "step": 10368
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.15849956609524e-06,
+      "loss": 0.9072,
+      "step": 10369
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157826790196566e-06,
+      "loss": 0.8936,
+      "step": 10370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157153992146842e-06,
+      "loss": 0.8628,
+      "step": 10371
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.156481171958937e-06,
+      "loss": 0.8979,
+      "step": 10372
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155808329645726e-06,
+      "loss": 0.8628,
+      "step": 10373
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155135465220079e-06,
+      "loss": 0.8677,
+      "step": 10374
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.154462578694871e-06,
+      "loss": 0.8789,
+      "step": 10375
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153789670082976e-06,
+      "loss": 0.8901,
+      "step": 10376
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153116739397265e-06,
+      "loss": 0.8447,
+      "step": 10377
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.152443786650613e-06,
+      "loss": 0.8916,
+      "step": 10378
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151770811855896e-06,
+      "loss": 0.8403,
+      "step": 10379
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151097815025989e-06,
+      "loss": 0.856,
+      "step": 10380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1504247961737664e-06,
+      "loss": 0.895,
+      "step": 10381
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.149751755312103e-06,
+      "loss": 0.8433,
+      "step": 10382
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14907869245388e-06,
+      "loss": 0.8721,
+      "step": 10383
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1484056076119694e-06,
+      "loss": 0.9248,
+      "step": 10384
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1477325007992495e-06,
+      "loss": 0.8638,
+      "step": 10385
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.147059372028599e-06,
+      "loss": 0.8115,
+      "step": 10386
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.146386221312894e-06,
+      "loss": 0.8301,
+      "step": 10387
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.145713048665015e-06,
+      "loss": 0.8872,
+      "step": 10388
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1450398540978395e-06,
+      "loss": 0.8662,
+      "step": 10389
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.144366637624247e-06,
+      "loss": 0.9121,
+      "step": 10390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.143693399257118e-06,
+      "loss": 0.8398,
+      "step": 10391
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.143020139009332e-06,
+      "loss": 0.8521,
+      "step": 10392
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.142346856893771e-06,
+      "loss": 0.832,
+      "step": 10393
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1416735529233114e-06,
+      "loss": 0.8999,
+      "step": 10394
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14100022711084e-06,
+      "loss": 0.8618,
+      "step": 10395
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.140326879469236e-06,
+      "loss": 0.9551,
+      "step": 10396
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1396535100113816e-06,
+      "loss": 0.8257,
+      "step": 10397
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138980118750159e-06,
+      "loss": 0.9131,
+      "step": 10398
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138306705698453e-06,
+      "loss": 0.9199,
+      "step": 10399
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.137633270869146e-06,
+      "loss": 0.9199,
+      "step": 10400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1369598142751205e-06,
+      "loss": 0.9199,
+      "step": 10401
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.136286335929262e-06,
+      "loss": 0.9028,
+      "step": 10402
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1356128358444565e-06,
+      "loss": 0.9136,
+      "step": 10403
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134939314033588e-06,
+      "loss": 0.8862,
+      "step": 10404
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134265770509541e-06,
+      "loss": 0.8618,
+      "step": 10405
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.133592205285205e-06,
+      "loss": 0.8716,
+      "step": 10406
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.13291861837346e-06,
+      "loss": 0.8867,
+      "step": 10407
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.132245009787199e-06,
+      "loss": 0.8867,
+      "step": 10408
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.131571379539307e-06,
+      "loss": 0.9082,
+      "step": 10409
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.130897727642671e-06,
+      "loss": 0.4111,
+      "step": 10410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1302240541101765e-06,
+      "loss": 0.9146,
+      "step": 10411
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.129550358954719e-06,
+      "loss": 0.8813,
+      "step": 10412
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128876642189181e-06,
+      "loss": 0.9263,
+      "step": 10413
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128202903826454e-06,
+      "loss": 0.2708,
+      "step": 10414
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.127529143879428e-06,
+      "loss": 0.9233,
+      "step": 10415
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.126855362360993e-06,
+      "loss": 0.9292,
+      "step": 10416
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.126181559284039e-06,
+      "loss": 0.8369,
+      "step": 10417
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.125507734661458e-06,
+      "loss": 0.896,
+      "step": 10418
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.124833888506139e-06,
+      "loss": 0.9521,
+      "step": 10419
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.124160020830976e-06,
+      "loss": 0.9341,
+      "step": 10420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.123486131648859e-06,
+      "loss": 0.9302,
+      "step": 10421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.122812220972685e-06,
+      "loss": 0.9204,
+      "step": 10422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1221382888153406e-06,
+      "loss": 0.8491,
+      "step": 10423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.121464335189723e-06,
+      "loss": 0.8745,
+      "step": 10424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.120790360108728e-06,
+      "loss": 0.3422,
+      "step": 10425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.120116363585244e-06,
+      "loss": 0.8379,
+      "step": 10426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.11944234563217e-06,
+      "loss": 0.8789,
+      "step": 10427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.118768306262401e-06,
+      "loss": 0.9219,
+      "step": 10428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.118094245488831e-06,
+      "loss": 0.9175,
+      "step": 10429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117420163324356e-06,
+      "loss": 0.8311,
+      "step": 10430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.116746059781872e-06,
+      "loss": 0.9229,
+      "step": 10431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.116071934874276e-06,
+      "loss": 0.8564,
+      "step": 10432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.115397788614466e-06,
+      "loss": 0.9004,
+      "step": 10433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.114723621015338e-06,
+      "loss": 0.8794,
+      "step": 10434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.114049432089791e-06,
+      "loss": 0.8862,
+      "step": 10435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113375221850722e-06,
+      "loss": 0.8882,
+      "step": 10436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.112700990311031e-06,
+      "loss": 0.8643,
+      "step": 10437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1120267374836185e-06,
+      "loss": 0.8076,
+      "step": 10438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.11135246338138e-06,
+      "loss": 0.8252,
+      "step": 10439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.110678168017217e-06,
+      "loss": 0.8857,
+      "step": 10440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.110003851404032e-06,
+      "loss": 0.9282,
+      "step": 10441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1093295135547235e-06,
+      "loss": 0.8916,
+      "step": 10442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.108655154482194e-06,
+      "loss": 0.895,
+      "step": 10443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.107980774199342e-06,
+      "loss": 0.7954,
+      "step": 10444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.107306372719075e-06,
+      "loss": 0.8579,
+      "step": 10445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.10663195005429e-06,
+      "loss": 0.9023,
+      "step": 10446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105957506217894e-06,
+      "loss": 0.8755,
+      "step": 10447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105283041222787e-06,
+      "loss": 0.9282,
+      "step": 10448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104608555081873e-06,
+      "loss": 0.9805,
+      "step": 10449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.103934047808058e-06,
+      "loss": 0.8931,
+      "step": 10450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.103259519414244e-06,
+      "loss": 0.9038,
+      "step": 10451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.102584969913337e-06,
+      "loss": 0.9644,
+      "step": 10452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.101910399318243e-06,
+      "loss": 0.8477,
+      "step": 10453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.101235807641866e-06,
+      "loss": 0.8594,
+      "step": 10454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.100561194897113e-06,
+      "loss": 0.8633,
+      "step": 10455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.099886561096888e-06,
+      "loss": 0.8936,
+      "step": 10456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.099211906254102e-06,
+      "loss": 0.8452,
+      "step": 10457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.098537230381661e-06,
+      "loss": 0.8438,
+      "step": 10458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.09786253349247e-06,
+      "loss": 0.8735,
+      "step": 10459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.097187815599438e-06,
+      "loss": 0.9365,
+      "step": 10460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0965130767154764e-06,
+      "loss": 0.9277,
+      "step": 10461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0958383168534905e-06,
+      "loss": 0.8911,
+      "step": 10462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.095163536026391e-06,
+      "loss": 0.8267,
+      "step": 10463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.094488734247087e-06,
+      "loss": 0.8506,
+      "step": 10464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.09381391152849e-06,
+      "loss": 0.9287,
+      "step": 10465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0931390678835076e-06,
+      "loss": 0.936,
+      "step": 10466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.092464203325053e-06,
+      "loss": 0.8765,
+      "step": 10467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.091789317866038e-06,
+      "loss": 0.874,
+      "step": 10468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.09111441151937e-06,
+      "loss": 0.3738,
+      "step": 10469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.090439484297966e-06,
+      "loss": 0.8394,
+      "step": 10470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.089764536214736e-06,
+      "loss": 0.8984,
+      "step": 10471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.089089567282594e-06,
+      "loss": 0.8564,
+      "step": 10472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0884145775144495e-06,
+      "loss": 0.8657,
+      "step": 10473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.087739566923222e-06,
+      "loss": 0.9307,
+      "step": 10474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.087064535521821e-06,
+      "loss": 0.8711,
+      "step": 10475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.086389483323163e-06,
+      "loss": 0.8745,
+      "step": 10476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.085714410340162e-06,
+      "loss": 0.8994,
+      "step": 10477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0850393165857335e-06,
+      "loss": 0.8774,
+      "step": 10478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.084364202072792e-06,
+      "loss": 0.4066,
+      "step": 10479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083689066814257e-06,
+      "loss": 0.8594,
+      "step": 10480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.08301391082304e-06,
+      "loss": 0.4421,
+      "step": 10481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.082338734112062e-06,
+      "loss": 0.8315,
+      "step": 10482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081663536694238e-06,
+      "loss": 0.8677,
+      "step": 10483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.080988318582486e-06,
+      "loss": 0.896,
+      "step": 10484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.080313079789723e-06,
+      "loss": 0.8955,
+      "step": 10485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07963782032887e-06,
+      "loss": 0.8926,
+      "step": 10486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.078962540212844e-06,
+      "loss": 0.8799,
+      "step": 10487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0782872394545635e-06,
+      "loss": 0.8774,
+      "step": 10488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.077611918066949e-06,
+      "loss": 0.8853,
+      "step": 10489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.076936576062922e-06,
+      "loss": 0.9155,
+      "step": 10490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.076261213455399e-06,
+      "loss": 0.917,
+      "step": 10491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0755858302573044e-06,
+      "loss": 0.8125,
+      "step": 10492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.074910426481556e-06,
+      "loss": 0.8809,
+      "step": 10493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07423500214108e-06,
+      "loss": 0.8945,
+      "step": 10494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.073559557248795e-06,
+      "loss": 0.8984,
+      "step": 10495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072884091817624e-06,
+      "loss": 0.8638,
+      "step": 10496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072208605860488e-06,
+      "loss": 0.3103,
+      "step": 10497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.071533099390313e-06,
+      "loss": 0.8491,
+      "step": 10498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.070857572420021e-06,
+      "loss": 0.854,
+      "step": 10499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.070182024962537e-06,
+      "loss": 0.8701,
+      "step": 10500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.069506457030785e-06,
+      "loss": 0.8677,
+      "step": 10501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.068830868637686e-06,
+      "loss": 0.8618,
+      "step": 10502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0681552597961705e-06,
+      "loss": 0.8164,
+      "step": 10503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.067479630519162e-06,
+      "loss": 0.9409,
+      "step": 10504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.066803980819585e-06,
+      "loss": 0.8838,
+      "step": 10505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0661283107103665e-06,
+      "loss": 0.814,
+      "step": 10506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.065452620204435e-06,
+      "loss": 0.8828,
+      "step": 10507
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064776909314716e-06,
+      "loss": 0.8955,
+      "step": 10508
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064101178054135e-06,
+      "loss": 0.9575,
+      "step": 10509
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.063425426435623e-06,
+      "loss": 0.8472,
+      "step": 10510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.062749654472105e-06,
+      "loss": 0.8403,
+      "step": 10511
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.062073862176514e-06,
+      "loss": 0.9136,
+      "step": 10512
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.061398049561774e-06,
+      "loss": 0.8638,
+      "step": 10513
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.060722216640819e-06,
+      "loss": 0.8599,
+      "step": 10514
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.060046363426575e-06,
+      "loss": 0.8701,
+      "step": 10515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.059370489931974e-06,
+      "loss": 0.8828,
+      "step": 10516
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058694596169947e-06,
+      "loss": 0.8115,
+      "step": 10517
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058018682153422e-06,
+      "loss": 0.4023,
+      "step": 10518
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.057342747895334e-06,
+      "loss": 0.8447,
+      "step": 10519
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.056666793408614e-06,
+      "loss": 0.8486,
+      "step": 10520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.055990818706191e-06,
+      "loss": 0.917,
+      "step": 10521
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.055314823801e-06,
+      "loss": 0.8672,
+      "step": 10522
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.054638808705975e-06,
+      "loss": 0.8667,
+      "step": 10523
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053962773434046e-06,
+      "loss": 0.8442,
+      "step": 10524
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053286717998149e-06,
+      "loss": 0.9355,
+      "step": 10525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.052610642411217e-06,
+      "loss": 0.9087,
+      "step": 10526
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.051934546686186e-06,
+      "loss": 0.8804,
+      "step": 10527
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.051258430835988e-06,
+      "loss": 0.8818,
+      "step": 10528
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.050582294873561e-06,
+      "loss": 0.8979,
+      "step": 10529
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.049906138811837e-06,
+      "loss": 0.9331,
+      "step": 10530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.049229962663755e-06,
+      "loss": 0.8589,
+      "step": 10531
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.048553766442251e-06,
+      "loss": 0.917,
+      "step": 10532
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.047877550160262e-06,
+      "loss": 0.8804,
+      "step": 10533
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.047201313830724e-06,
+      "loss": 0.9141,
+      "step": 10534
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.046525057466572e-06,
+      "loss": 0.8706,
+      "step": 10535
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04584878108075e-06,
+      "loss": 0.9316,
+      "step": 10536
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.045172484686191e-06,
+      "loss": 0.8555,
+      "step": 10537
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.044496168295835e-06,
+      "loss": 0.8242,
+      "step": 10538
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.043819831922621e-06,
+      "loss": 0.8979,
+      "step": 10539
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04314347557949e-06,
+      "loss": 0.8564,
+      "step": 10540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.042467099279381e-06,
+      "loss": 0.8716,
+      "step": 10541
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041790703035233e-06,
+      "loss": 0.3265,
+      "step": 10542
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041114286859986e-06,
+      "loss": 0.8271,
+      "step": 10543
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.040437850766584e-06,
+      "loss": 0.915,
+      "step": 10544
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039761394767965e-06,
+      "loss": 0.9077,
+      "step": 10545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039084918877074e-06,
+      "loss": 0.8848,
+      "step": 10546
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.038408423106848e-06,
+      "loss": 0.8154,
+      "step": 10547
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.037731907470234e-06,
+      "loss": 0.8965,
+      "step": 10548
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.037055371980175e-06,
+      "loss": 0.8843,
+      "step": 10549
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0363788166496105e-06,
+      "loss": 0.8608,
+      "step": 10550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.035702241491487e-06,
+      "loss": 0.8911,
+      "step": 10551
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.035025646518747e-06,
+      "loss": 0.9463,
+      "step": 10552
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.034349031744337e-06,
+      "loss": 0.9048,
+      "step": 10553
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.033672397181198e-06,
+      "loss": 0.8765,
+      "step": 10554
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0329957428422765e-06,
+      "loss": 0.8442,
+      "step": 10555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.032319068740522e-06,
+      "loss": 0.8867,
+      "step": 10556
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0316423748888744e-06,
+      "loss": 0.8643,
+      "step": 10557
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0309656613002836e-06,
+      "loss": 0.8838,
+      "step": 10558
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0302889279876945e-06,
+      "loss": 0.8779,
+      "step": 10559
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.029612174964055e-06,
+      "loss": 0.8618,
+      "step": 10560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.028935402242311e-06,
+      "loss": 0.8726,
+      "step": 10561
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.028258609835412e-06,
+      "loss": 0.9102,
+      "step": 10562
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.027581797756307e-06,
+      "loss": 0.8882,
+      "step": 10563
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.026904966017941e-06,
+      "loss": 0.9199,
+      "step": 10564
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.026228114633266e-06,
+      "loss": 0.8086,
+      "step": 10565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0255512436152295e-06,
+      "loss": 0.8091,
+      "step": 10566
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.024874352976781e-06,
+      "loss": 0.8921,
+      "step": 10567
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.024197442730872e-06,
+      "loss": 0.9072,
+      "step": 10568
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.023520512890453e-06,
+      "loss": 0.335,
+      "step": 10569
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.022843563468473e-06,
+      "loss": 0.8203,
+      "step": 10570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.022166594477883e-06,
+      "loss": 0.8721,
+      "step": 10571
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.021489605931636e-06,
+      "loss": 0.8477,
+      "step": 10572
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0208125978426825e-06,
+      "loss": 0.8955,
+      "step": 10573
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020135570223975e-06,
+      "loss": 0.9185,
+      "step": 10574
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.019458523088468e-06,
+      "loss": 0.8354,
+      "step": 10575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.018781456449112e-06,
+      "loss": 0.9502,
+      "step": 10576
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.018104370318861e-06,
+      "loss": 0.8857,
+      "step": 10577
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01742726471067e-06,
+      "loss": 0.834,
+      "step": 10578
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0167501396374925e-06,
+      "loss": 0.8374,
+      "step": 10579
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.016072995112281e-06,
+      "loss": 0.8813,
+      "step": 10580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.015395831147992e-06,
+      "loss": 0.8784,
+      "step": 10581
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0147186477575835e-06,
+      "loss": 0.8682,
+      "step": 10582
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.014041444954006e-06,
+      "loss": 0.8511,
+      "step": 10583
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.013364222750218e-06,
+      "loss": 0.835,
+      "step": 10584
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.012686981159176e-06,
+      "loss": 0.8813,
+      "step": 10585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.012009720193836e-06,
+      "loss": 0.8931,
+      "step": 10586
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.011332439867155e-06,
+      "loss": 0.875,
+      "step": 10587
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.010655140192091e-06,
+      "loss": 0.7617,
+      "step": 10588
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0099778211816004e-06,
+      "loss": 0.8965,
+      "step": 10589
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.009300482848645e-06,
+      "loss": 0.8882,
+      "step": 10590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.008623125206179e-06,
+      "loss": 0.8604,
+      "step": 10591
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0079457482671636e-06,
+      "loss": 0.832,
+      "step": 10592
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.007268352044559e-06,
+      "loss": 0.8779,
+      "step": 10593
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.006590936551322e-06,
+      "loss": 0.876,
+      "step": 10594
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.005913501800416e-06,
+      "loss": 0.8745,
+      "step": 10595
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.005236047804797e-06,
+      "loss": 0.8418,
+      "step": 10596
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.004558574577429e-06,
+      "loss": 0.8545,
+      "step": 10597
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.003881082131274e-06,
+      "loss": 1.0049,
+      "step": 10598
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0032035704792915e-06,
+      "loss": 0.8784,
+      "step": 10599
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.002526039634442e-06,
+      "loss": 0.8545,
+      "step": 10600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.001848489609692e-06,
+      "loss": 0.8706,
+      "step": 10601
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.001170920418001e-06,
+      "loss": 0.937,
+      "step": 10602
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.000493332072333e-06,
+      "loss": 0.8701,
+      "step": 10603
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.999815724585649e-06,
+      "loss": 0.8696,
+      "step": 10604
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.999138097970917e-06,
+      "loss": 0.8838,
+      "step": 10605
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.998460452241097e-06,
+      "loss": 0.9243,
+      "step": 10606
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997782787409157e-06,
+      "loss": 0.8662,
+      "step": 10607
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997105103488059e-06,
+      "loss": 0.8525,
+      "step": 10608
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.996427400490768e-06,
+      "loss": 0.8848,
+      "step": 10609
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.995749678430251e-06,
+      "loss": 0.8076,
+      "step": 10610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.995071937319474e-06,
+      "loss": 0.9165,
+      "step": 10611
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.994394177171404e-06,
+      "loss": 0.8853,
+      "step": 10612
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.993716397999005e-06,
+      "loss": 0.8018,
+      "step": 10613
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9930385998152455e-06,
+      "loss": 0.8062,
+      "step": 10614
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.992360782633094e-06,
+      "loss": 0.8315,
+      "step": 10615
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.991682946465516e-06,
+      "loss": 0.9019,
+      "step": 10616
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.991005091325479e-06,
+      "loss": 0.8486,
+      "step": 10617
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.990327217225956e-06,
+      "loss": 0.79,
+      "step": 10618
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9896493241799115e-06,
+      "loss": 0.9258,
+      "step": 10619
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9889714122003155e-06,
+      "loss": 0.8403,
+      "step": 10620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.988293481300138e-06,
+      "loss": 0.8209,
+      "step": 10621
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.987615531492349e-06,
+      "loss": 0.8511,
+      "step": 10622
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.986937562789918e-06,
+      "loss": 0.854,
+      "step": 10623
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.986259575205817e-06,
+      "loss": 0.8384,
+      "step": 10624
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.985581568753017e-06,
+      "loss": 0.8726,
+      "step": 10625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.984903543444487e-06,
+      "loss": 0.8628,
+      "step": 10626
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.984225499293202e-06,
+      "loss": 0.855,
+      "step": 10627
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.983547436312132e-06,
+      "loss": 0.8477,
+      "step": 10628
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.982869354514247e-06,
+      "loss": 0.8516,
+      "step": 10629
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9821912539125234e-06,
+      "loss": 0.8677,
+      "step": 10630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.981513134519935e-06,
+      "loss": 0.8472,
+      "step": 10631
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.980834996349453e-06,
+      "loss": 0.9082,
+      "step": 10632
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.980156839414051e-06,
+      "loss": 0.3876,
+      "step": 10633
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9794786637267054e-06,
+      "loss": 0.8618,
+      "step": 10634
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9788004693003885e-06,
+      "loss": 0.8545,
+      "step": 10635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9781222561480764e-06,
+      "loss": 0.876,
+      "step": 10636
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.977444024282745e-06,
+      "loss": 0.8364,
+      "step": 10637
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.976765773717368e-06,
+      "loss": 0.896,
+      "step": 10638
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9760875044649235e-06,
+      "loss": 0.8999,
+      "step": 10639
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.975409216538386e-06,
+      "loss": 0.8955,
+      "step": 10640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.974730909950734e-06,
+      "loss": 0.9639,
+      "step": 10641
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.974052584714943e-06,
+      "loss": 0.8325,
+      "step": 10642
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.973374240843992e-06,
+      "loss": 0.7708,
+      "step": 10643
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.972695878350858e-06,
+      "loss": 0.813,
+      "step": 10644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.972017497248519e-06,
+      "loss": 0.8867,
+      "step": 10645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.971339097549952e-06,
+      "loss": 0.854,
+      "step": 10646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.970660679268139e-06,
+      "loss": 0.8589,
+      "step": 10647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.969982242416057e-06,
+      "loss": 0.9204,
+      "step": 10648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9693037870066865e-06,
+      "loss": 0.8911,
+      "step": 10649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.968625313053007e-06,
+      "loss": 0.853,
+      "step": 10650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.967946820567999e-06,
+      "loss": 0.873,
+      "step": 10651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.967268309564642e-06,
+      "loss": 0.8965,
+      "step": 10652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.966589780055919e-06,
+      "loss": 0.8184,
+      "step": 10653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.965911232054811e-06,
+      "loss": 0.8418,
+      "step": 10654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.965232665574298e-06,
+      "loss": 0.8003,
+      "step": 10655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9645540806273625e-06,
+      "loss": 0.9097,
+      "step": 10656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.96387547722699e-06,
+      "loss": 0.9014,
+      "step": 10657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.963196855386158e-06,
+      "loss": 0.8828,
+      "step": 10658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.962518215117854e-06,
+      "loss": 0.9058,
+      "step": 10659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.961839556435059e-06,
+      "loss": 0.813,
+      "step": 10660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.961160879350759e-06,
+      "loss": 0.8975,
+      "step": 10661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.960482183877934e-06,
+      "loss": 0.8765,
+      "step": 10662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.959803470029573e-06,
+      "loss": 0.8491,
+      "step": 10663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.95912473781866e-06,
+      "loss": 0.8374,
+      "step": 10664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.958445987258179e-06,
+      "loss": 0.7925,
+      "step": 10665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.957767218361113e-06,
+      "loss": 0.8438,
+      "step": 10666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.957088431140455e-06,
+      "loss": 0.8828,
+      "step": 10667
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.956409625609185e-06,
+      "loss": 0.855,
+      "step": 10668
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9557308017802925e-06,
+      "loss": 0.8433,
+      "step": 10669
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.955051959666763e-06,
+      "loss": 0.8979,
+      "step": 10670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.954373099281585e-06,
+      "loss": 0.874,
+      "step": 10671
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9536942206377456e-06,
+      "loss": 0.9282,
+      "step": 10672
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9530153237482335e-06,
+      "loss": 0.897,
+      "step": 10673
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9523364086260375e-06,
+      "loss": 0.8662,
+      "step": 10674
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9516574752841416e-06,
+      "loss": 0.9448,
+      "step": 10675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.950978523735541e-06,
+      "loss": 0.8716,
+      "step": 10676
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.950299553993223e-06,
+      "loss": 0.9136,
+      "step": 10677
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.949620566070177e-06,
+      "loss": 0.8975,
+      "step": 10678
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.948941559979391e-06,
+      "loss": 0.7964,
+      "step": 10679
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9482625357338605e-06,
+      "loss": 0.9619,
+      "step": 10680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.947583493346572e-06,
+      "loss": 0.8857,
+      "step": 10681
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.946904432830518e-06,
+      "loss": 0.8857,
+      "step": 10682
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.94622535419869e-06,
+      "loss": 0.8379,
+      "step": 10683
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.945546257464081e-06,
+      "loss": 0.8267,
+      "step": 10684
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944867142639681e-06,
+      "loss": 0.9258,
+      "step": 10685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944188009738483e-06,
+      "loss": 0.8794,
+      "step": 10686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.943508858773482e-06,
+      "loss": 0.8804,
+      "step": 10687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.942829689757669e-06,
+      "loss": 0.8223,
+      "step": 10688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.942150502704038e-06,
+      "loss": 0.302,
+      "step": 10689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.941471297625585e-06,
+      "loss": 0.9458,
+      "step": 10690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9407920745353e-06,
+      "loss": 0.8286,
+      "step": 10691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.940112833446181e-06,
+      "loss": 0.8911,
+      "step": 10692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.939433574371224e-06,
+      "loss": 0.3452,
+      "step": 10693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93875429732342e-06,
+      "loss": 0.8145,
+      "step": 10694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.938075002315768e-06,
+      "loss": 0.853,
+      "step": 10695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.937395689361263e-06,
+      "loss": 0.8774,
+      "step": 10696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9367163584729016e-06,
+      "loss": 0.8462,
+      "step": 10697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.936037009663679e-06,
+      "loss": 0.8936,
+      "step": 10698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.935357642946594e-06,
+      "loss": 0.8882,
+      "step": 10699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.934678258334643e-06,
+      "loss": 0.9185,
+      "step": 10700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.933998855840824e-06,
+      "loss": 0.877,
+      "step": 10701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9333194354781364e-06,
+      "loss": 0.8447,
+      "step": 10702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.932639997259577e-06,
+      "loss": 0.9517,
+      "step": 10703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.931960541198143e-06,
+      "loss": 0.9136,
+      "step": 10704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.931281067306836e-06,
+      "loss": 0.854,
+      "step": 10705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9306015755986565e-06,
+      "loss": 0.9155,
+      "step": 10706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.929922066086601e-06,
+      "loss": 0.876,
+      "step": 10707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9292425387836705e-06,
+      "loss": 0.8125,
+      "step": 10708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9285629937028664e-06,
+      "loss": 0.8442,
+      "step": 10709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9278834308571896e-06,
+      "loss": 0.8364,
+      "step": 10710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.927203850259639e-06,
+      "loss": 0.8457,
+      "step": 10711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92652425192322e-06,
+      "loss": 0.8218,
+      "step": 10712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92584463586093e-06,
+      "loss": 0.8018,
+      "step": 10713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.925165002085774e-06,
+      "loss": 0.8613,
+      "step": 10714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.924485350610753e-06,
+      "loss": 0.894,
+      "step": 10715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.923805681448872e-06,
+      "loss": 0.895,
+      "step": 10716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.923125994613132e-06,
+      "loss": 0.8672,
+      "step": 10717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922446290116536e-06,
+      "loss": 0.9233,
+      "step": 10718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.921766567972091e-06,
+      "loss": 0.8608,
+      "step": 10719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.921086828192796e-06,
+      "loss": 0.8779,
+      "step": 10720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9204070707916596e-06,
+      "loss": 0.9438,
+      "step": 10721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.919727295781685e-06,
+      "loss": 0.8594,
+      "step": 10722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.919047503175881e-06,
+      "loss": 0.8325,
+      "step": 10723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.918367692987247e-06,
+      "loss": 0.8955,
+      "step": 10724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9176878652287925e-06,
+      "loss": 0.8467,
+      "step": 10725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.917008019913524e-06,
+      "loss": 0.8369,
+      "step": 10726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.916328157054447e-06,
+      "loss": 0.9072,
+      "step": 10727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.915648276664566e-06,
+      "loss": 0.8164,
+      "step": 10728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.914968378756894e-06,
+      "loss": 0.9282,
+      "step": 10729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.914288463344433e-06,
+      "loss": 0.8799,
+      "step": 10730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9136085304401935e-06,
+      "loss": 0.8936,
+      "step": 10731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.912928580057184e-06,
+      "loss": 0.437,
+      "step": 10732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9122486122084096e-06,
+      "loss": 0.8452,
+      "step": 10733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.911568626906884e-06,
+      "loss": 0.8584,
+      "step": 10734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.910888624165613e-06,
+      "loss": 0.8608,
+      "step": 10735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.910208603997609e-06,
+      "loss": 0.8706,
+      "step": 10736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.909528566415878e-06,
+      "loss": 0.8599,
+      "step": 10737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.908848511433434e-06,
+      "loss": 0.9028,
+      "step": 10738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.908168439063285e-06,
+      "loss": 0.8726,
+      "step": 10739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.907488349318442e-06,
+      "loss": 0.8975,
+      "step": 10740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.906808242211917e-06,
+      "loss": 0.8057,
+      "step": 10741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9061281177567235e-06,
+      "loss": 0.8472,
+      "step": 10742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.905447975965869e-06,
+      "loss": 0.8779,
+      "step": 10743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.904767816852368e-06,
+      "loss": 0.9233,
+      "step": 10744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.904087640429234e-06,
+      "loss": 0.9146,
+      "step": 10745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.903407446709477e-06,
+      "loss": 0.3625,
+      "step": 10746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.902727235706114e-06,
+      "loss": 0.856,
+      "step": 10747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.902047007432156e-06,
+      "loss": 0.8413,
+      "step": 10748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.901366761900615e-06,
+      "loss": 0.8926,
+      "step": 10749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.900686499124511e-06,
+      "loss": 0.876,
+      "step": 10750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9000062191168525e-06,
+      "loss": 0.8911,
+      "step": 10751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8993259218906584e-06,
+      "loss": 0.8354,
+      "step": 10752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.898645607458941e-06,
+      "loss": 0.8901,
+      "step": 10753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8979652758347175e-06,
+      "loss": 0.8545,
+      "step": 10754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.897284927031004e-06,
+      "loss": 0.8647,
+      "step": 10755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.896604561060815e-06,
+      "loss": 0.8525,
+      "step": 10756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.895924177937167e-06,
+      "loss": 0.8877,
+      "step": 10757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.89524377767308e-06,
+      "loss": 0.853,
+      "step": 10758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.894563360281567e-06,
+      "loss": 0.8696,
+      "step": 10759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.893882925775648e-06,
+      "loss": 0.8774,
+      "step": 10760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.893202474168339e-06,
+      "loss": 0.8608,
+      "step": 10761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8925220054726605e-06,
+      "loss": 0.8501,
+      "step": 10762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.891841519701629e-06,
+      "loss": 0.8574,
+      "step": 10763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8911610168682635e-06,
+      "loss": 0.8667,
+      "step": 10764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.890480496985585e-06,
+      "loss": 0.9131,
+      "step": 10765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.889799960066609e-06,
+      "loss": 0.8657,
+      "step": 10766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.889119406124359e-06,
+      "loss": 0.895,
+      "step": 10767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.888438835171854e-06,
+      "loss": 0.8105,
+      "step": 10768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.887758247222114e-06,
+      "loss": 0.8848,
+      "step": 10769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.887077642288158e-06,
+      "loss": 0.8169,
+      "step": 10770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.886397020383012e-06,
+      "loss": 0.9277,
+      "step": 10771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.885716381519693e-06,
+      "loss": 0.8477,
+      "step": 10772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8850357257112225e-06,
+      "loss": 0.8438,
+      "step": 10773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.884355052970624e-06,
+      "loss": 0.9038,
+      "step": 10774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.883674363310922e-06,
+      "loss": 0.8809,
+      "step": 10775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.882993656745135e-06,
+      "loss": 0.8926,
+      "step": 10776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8823129332862885e-06,
+      "loss": 0.8604,
+      "step": 10777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.881632192947404e-06,
+      "loss": 0.8608,
+      "step": 10778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.880951435741507e-06,
+      "loss": 0.9497,
+      "step": 10779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.880270661681622e-06,
+      "loss": 0.8394,
+      "step": 10780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8795898707807716e-06,
+      "loss": 0.8647,
+      "step": 10781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.878909063051981e-06,
+      "loss": 0.8696,
+      "step": 10782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.878228238508275e-06,
+      "loss": 0.8687,
+      "step": 10783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8775473971626785e-06,
+      "loss": 0.9004,
+      "step": 10784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.876866539028218e-06,
+      "loss": 0.8477,
+      "step": 10785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.876185664117918e-06,
+      "loss": 0.9033,
+      "step": 10786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.875504772444806e-06,
+      "loss": 0.8247,
+      "step": 10787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.874823864021909e-06,
+      "loss": 0.8838,
+      "step": 10788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8741429388622515e-06,
+      "loss": 0.8779,
+      "step": 10789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.873461996978862e-06,
+      "loss": 0.8394,
+      "step": 10790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.87278103838477e-06,
+      "loss": 0.8618,
+      "step": 10791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.872100063093e-06,
+      "loss": 0.9048,
+      "step": 10792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.871419071116581e-06,
+      "loss": 0.8364,
+      "step": 10793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.870738062468543e-06,
+      "loss": 0.7175,
+      "step": 10794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8700570371619136e-06,
+      "loss": 0.8354,
+      "step": 10795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8693759952097215e-06,
+      "loss": 0.8623,
+      "step": 10796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8686949366249965e-06,
+      "loss": 0.8623,
+      "step": 10797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.868013861420769e-06,
+      "loss": 0.8218,
+      "step": 10798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.867332769610069e-06,
+      "loss": 0.8662,
+      "step": 10799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8666516612059255e-06,
+      "loss": 0.875,
+      "step": 10800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.865970536221371e-06,
+      "loss": 0.9414,
+      "step": 10801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.865289394669435e-06,
+      "loss": 0.8906,
+      "step": 10802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8646082365631485e-06,
+      "loss": 0.9185,
+      "step": 10803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.863927061915545e-06,
+      "loss": 0.8784,
+      "step": 10804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.863245870739655e-06,
+      "loss": 0.8813,
+      "step": 10805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.862564663048512e-06,
+      "loss": 0.894,
+      "step": 10806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8618834388551475e-06,
+      "loss": 0.8857,
+      "step": 10807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8612021981725944e-06,
+      "loss": 0.7725,
+      "step": 10808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860520941013885e-06,
+      "loss": 0.835,
+      "step": 10809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.859839667392055e-06,
+      "loss": 0.917,
+      "step": 10810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.859158377320138e-06,
+      "loss": 0.9097,
+      "step": 10811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.858477070811166e-06,
+      "loss": 0.8735,
+      "step": 10812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857795747878176e-06,
+      "loss": 0.8442,
+      "step": 10813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857114408534202e-06,
+      "loss": 0.8872,
+      "step": 10814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.856433052792276e-06,
+      "loss": 0.9014,
+      "step": 10815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.855751680665438e-06,
+      "loss": 0.876,
+      "step": 10816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8550702921667235e-06,
+      "loss": 0.8774,
+      "step": 10817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.854388887309165e-06,
+      "loss": 0.394,
+      "step": 10818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8537074661058005e-06,
+      "loss": 0.9165,
+      "step": 10819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8530260285696674e-06,
+      "loss": 0.856,
+      "step": 10820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.852344574713802e-06,
+      "loss": 0.8823,
+      "step": 10821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851663104551242e-06,
+      "loss": 0.8979,
+      "step": 10822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.850981618095024e-06,
+      "loss": 0.8853,
+      "step": 10823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.850300115358188e-06,
+      "loss": 0.8325,
+      "step": 10824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.84961859635377e-06,
+      "loss": 0.8398,
+      "step": 10825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.848937061094809e-06,
+      "loss": 0.8623,
+      "step": 10826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.848255509594345e-06,
+      "loss": 0.8887,
+      "step": 10827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.847573941865417e-06,
+      "loss": 0.8594,
+      "step": 10828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.846892357921063e-06,
+      "loss": 0.8652,
+      "step": 10829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8462107577743245e-06,
+      "loss": 0.958,
+      "step": 10830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.845529141438239e-06,
+      "loss": 0.8125,
+      "step": 10831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.84484750892585e-06,
+      "loss": 0.8765,
+      "step": 10832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.844165860250196e-06,
+      "loss": 0.9546,
+      "step": 10833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.843484195424321e-06,
+      "loss": 0.8677,
+      "step": 10834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.842802514461262e-06,
+      "loss": 0.9033,
+      "step": 10835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8421208173740616e-06,
+      "loss": 0.8916,
+      "step": 10836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.841439104175767e-06,
+      "loss": 0.8818,
+      "step": 10837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840757374879413e-06,
+      "loss": 0.9312,
+      "step": 10838
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840075629498047e-06,
+      "loss": 0.9185,
+      "step": 10839
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.83939386804471e-06,
+      "loss": 0.8408,
+      "step": 10840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838712090532446e-06,
+      "loss": 0.873,
+      "step": 10841
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838030296974298e-06,
+      "loss": 0.8901,
+      "step": 10842
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.837348487383309e-06,
+      "loss": 0.8765,
+      "step": 10843
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.836666661772524e-06,
+      "loss": 0.9331,
+      "step": 10844
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.835984820154988e-06,
+      "loss": 0.9199,
+      "step": 10845
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.835302962543745e-06,
+      "loss": 0.8394,
+      "step": 10846
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.834621088951841e-06,
+      "loss": 0.9224,
+      "step": 10847
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8339391993923175e-06,
+      "loss": 0.8423,
+      "step": 10848
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.833257293878225e-06,
+      "loss": 0.8418,
+      "step": 10849
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.832575372422608e-06,
+      "loss": 0.874,
+      "step": 10850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831893435038511e-06,
+      "loss": 0.8638,
+      "step": 10851
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831211481738982e-06,
+      "loss": 0.8442,
+      "step": 10852
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.830529512537069e-06,
+      "loss": 0.876,
+      "step": 10853
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.829847527445815e-06,
+      "loss": 0.8535,
+      "step": 10854
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.829165526478272e-06,
+      "loss": 0.9346,
+      "step": 10855
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.828483509647484e-06,
+      "loss": 0.8535,
+      "step": 10856
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.827801476966503e-06,
+      "loss": 0.8857,
+      "step": 10857
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.827119428448375e-06,
+      "loss": 0.9146,
+      "step": 10858
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.826437364106148e-06,
+      "loss": 0.8594,
+      "step": 10859
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825755283952872e-06,
+      "loss": 0.8838,
+      "step": 10860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825073188001597e-06,
+      "loss": 0.8447,
+      "step": 10861
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.824391076265371e-06,
+      "loss": 0.8052,
+      "step": 10862
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823708948757246e-06,
+      "loss": 0.8105,
+      "step": 10863
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823026805490268e-06,
+      "loss": 0.8794,
+      "step": 10864
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8223446464774925e-06,
+      "loss": 0.8394,
+      "step": 10865
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.821662471731968e-06,
+      "loss": 0.8501,
+      "step": 10866
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.820980281266746e-06,
+      "loss": 0.8291,
+      "step": 10867
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.820298075094874e-06,
+      "loss": 0.9204,
+      "step": 10868
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.819615853229412e-06,
+      "loss": 0.9082,
+      "step": 10869
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.818933615683406e-06,
+      "loss": 0.8774,
+      "step": 10870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.818251362469908e-06,
+      "loss": 0.8516,
+      "step": 10871
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.817569093601972e-06,
+      "loss": 0.3943,
+      "step": 10872
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.816886809092651e-06,
+      "loss": 0.8926,
+      "step": 10873
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.816204508954999e-06,
+      "loss": 0.8838,
+      "step": 10874
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.815522193202068e-06,
+      "loss": 0.8599,
+      "step": 10875
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.814839861846914e-06,
+      "loss": 0.9014,
+      "step": 10876
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8141575149025864e-06,
+      "loss": 0.896,
+      "step": 10877
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.813475152382144e-06,
+      "loss": 0.8574,
+      "step": 10878
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.812792774298641e-06,
+      "loss": 0.856,
+      "step": 10879
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81211038066513e-06,
+      "loss": 0.8643,
+      "step": 10880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.811427971494667e-06,
+      "loss": 0.8267,
+      "step": 10881
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.810745546800311e-06,
+      "loss": 0.8486,
+      "step": 10882
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.810063106595114e-06,
+      "loss": 0.8818,
+      "step": 10883
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.809380650892134e-06,
+      "loss": 0.8735,
+      "step": 10884
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.808698179704424e-06,
+      "loss": 0.8892,
+      "step": 10885
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.808015693045046e-06,
+      "loss": 0.873,
+      "step": 10886
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.807333190927054e-06,
+      "loss": 0.8667,
+      "step": 10887
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.806650673363504e-06,
+      "loss": 0.9521,
+      "step": 10888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.805968140367457e-06,
+      "loss": 0.9199,
+      "step": 10889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.80528559195197e-06,
+      "loss": 0.8955,
+      "step": 10890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.8046030281301e-06,
+      "loss": 0.8853,
+      "step": 10891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803920448914906e-06,
+      "loss": 0.7759,
+      "step": 10892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803237854319447e-06,
+      "loss": 0.9019,
+      "step": 10893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.802555244356782e-06,
+      "loss": 0.9312,
+      "step": 10894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.801872619039971e-06,
+      "loss": 0.7886,
+      "step": 10895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.801189978382075e-06,
+      "loss": 0.9038,
+      "step": 10896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.800507322396149e-06,
+      "loss": 0.8413,
+      "step": 10897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799824651095257e-06,
+      "loss": 0.8115,
+      "step": 10898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799141964492461e-06,
+      "loss": 0.8579,
+      "step": 10899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7984592626008185e-06,
+      "loss": 0.9165,
+      "step": 10900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.797776545433391e-06,
+      "loss": 0.8035,
+      "step": 10901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.797093813003243e-06,
+      "loss": 0.9019,
+      "step": 10902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.796411065323434e-06,
+      "loss": 0.8159,
+      "step": 10903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795728302407025e-06,
+      "loss": 0.9199,
+      "step": 10904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795045524267081e-06,
+      "loss": 0.8794,
+      "step": 10905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7943627309166625e-06,
+      "loss": 0.8687,
+      "step": 10906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7936799223688345e-06,
+      "loss": 0.8755,
+      "step": 10907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.792997098636658e-06,
+      "loss": 0.8921,
+      "step": 10908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.792314259733198e-06,
+      "loss": 0.8267,
+      "step": 10909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.791631405671516e-06,
+      "loss": 0.895,
+      "step": 10910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.790948536464679e-06,
+      "loss": 0.8843,
+      "step": 10911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.79026565212575e-06,
+      "loss": 0.8438,
+      "step": 10912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.789582752667795e-06,
+      "loss": 0.9224,
+      "step": 10913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.788899838103874e-06,
+      "loss": 0.8921,
+      "step": 10914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.78821690844706e-06,
+      "loss": 0.8896,
+      "step": 10915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.787533963710413e-06,
+      "loss": 0.8999,
+      "step": 10916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.786851003907e-06,
+      "loss": 0.8911,
+      "step": 10917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.786168029049887e-06,
+      "loss": 0.8687,
+      "step": 10918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.785485039152141e-06,
+      "loss": 0.8896,
+      "step": 10919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.78480203422683e-06,
+      "loss": 0.8535,
+      "step": 10920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.784119014287017e-06,
+      "loss": 0.856,
+      "step": 10921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.783435979345773e-06,
+      "loss": 0.9165,
+      "step": 10922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782752929416163e-06,
+      "loss": 0.8745,
+      "step": 10923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782069864511257e-06,
+      "loss": 0.812,
+      "step": 10924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7813867846441226e-06,
+      "loss": 0.8574,
+      "step": 10925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.780703689827827e-06,
+      "loss": 0.9248,
+      "step": 10926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7800205800754396e-06,
+      "loss": 0.8882,
+      "step": 10927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.779337455400031e-06,
+      "loss": 0.897,
+      "step": 10928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.778654315814667e-06,
+      "loss": 0.8901,
+      "step": 10929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7779711613324175e-06,
+      "loss": 0.8774,
+      "step": 10930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.777287991966356e-06,
+      "loss": 0.8843,
+      "step": 10931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.77660480772955e-06,
+      "loss": 0.8857,
+      "step": 10932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.77592160863507e-06,
+      "loss": 0.9253,
+      "step": 10933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.775238394695987e-06,
+      "loss": 0.9053,
+      "step": 10934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.774555165925372e-06,
+      "loss": 0.9082,
+      "step": 10935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.773871922336295e-06,
+      "loss": 0.8193,
+      "step": 10936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.77318866394183e-06,
+      "loss": 0.7949,
+      "step": 10937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.772505390755047e-06,
+      "loss": 0.8862,
+      "step": 10938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.771822102789018e-06,
+      "loss": 0.8984,
+      "step": 10939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.771138800056817e-06,
+      "loss": 0.8926,
+      "step": 10940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.770455482571516e-06,
+      "loss": 0.8721,
+      "step": 10941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.769772150346187e-06,
+      "loss": 0.8359,
+      "step": 10942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.769088803393902e-06,
+      "loss": 0.8564,
+      "step": 10943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.768405441727738e-06,
+      "loss": 0.9199,
+      "step": 10944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.767722065360767e-06,
+      "loss": 0.8298,
+      "step": 10945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.767038674306062e-06,
+      "loss": 0.855,
+      "step": 10946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7663552685766985e-06,
+      "loss": 0.8301,
+      "step": 10947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7656718481857535e-06,
+      "loss": 0.9004,
+      "step": 10948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.764988413146297e-06,
+      "loss": 0.8721,
+      "step": 10949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.764304963471409e-06,
+      "loss": 0.8628,
+      "step": 10950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7636214991741605e-06,
+      "loss": 0.8735,
+      "step": 10951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.76293802026763e-06,
+      "loss": 0.8984,
+      "step": 10952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.762254526764893e-06,
+      "loss": 0.8916,
+      "step": 10953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.761571018679025e-06,
+      "loss": 0.8809,
+      "step": 10954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.760887496023104e-06,
+      "loss": 0.8706,
+      "step": 10955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.760203958810206e-06,
+      "loss": 0.873,
+      "step": 10956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.759520407053409e-06,
+      "loss": 0.8604,
+      "step": 10957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.758836840765791e-06,
+      "loss": 0.8911,
+      "step": 10958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.758153259960426e-06,
+      "loss": 0.855,
+      "step": 10959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.757469664650395e-06,
+      "loss": 0.8774,
+      "step": 10960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.756786054848777e-06,
+      "loss": 0.8799,
+      "step": 10961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.756102430568648e-06,
+      "loss": 0.8599,
+      "step": 10962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.755418791823087e-06,
+      "loss": 0.8267,
+      "step": 10963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.754735138625176e-06,
+      "loss": 0.8423,
+      "step": 10964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.754051470987993e-06,
+      "loss": 0.9414,
+      "step": 10965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7533677889246155e-06,
+      "loss": 0.8872,
+      "step": 10966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752684092448126e-06,
+      "loss": 0.8896,
+      "step": 10967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752000381571602e-06,
+      "loss": 0.832,
+      "step": 10968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.751316656308127e-06,
+      "loss": 0.8701,
+      "step": 10969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.750632916670781e-06,
+      "loss": 0.8804,
+      "step": 10970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.749949162672644e-06,
+      "loss": 0.9102,
+      "step": 10971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.749265394326797e-06,
+      "loss": 0.957,
+      "step": 10972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.748581611646322e-06,
+      "loss": 0.9473,
+      "step": 10973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.747897814644303e-06,
+      "loss": 0.8071,
+      "step": 10974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.74721400333382e-06,
+      "loss": 0.8955,
+      "step": 10975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.746530177727953e-06,
+      "loss": 0.8364,
+      "step": 10976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.745846337839791e-06,
+      "loss": 0.8384,
+      "step": 10977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.745162483682411e-06,
+      "loss": 0.8672,
+      "step": 10978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.744478615268899e-06,
+      "loss": 0.8457,
+      "step": 10979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743794732612338e-06,
+      "loss": 0.9268,
+      "step": 10980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743110835725811e-06,
+      "loss": 0.8682,
+      "step": 10981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.742426924622404e-06,
+      "loss": 0.897,
+      "step": 10982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.741742999315198e-06,
+      "loss": 0.8882,
+      "step": 10983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.741059059817282e-06,
+      "loss": 0.8843,
+      "step": 10984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.740375106141737e-06,
+      "loss": 0.8345,
+      "step": 10985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.73969113830165e-06,
+      "loss": 0.9434,
+      "step": 10986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739007156310107e-06,
+      "loss": 0.8191,
+      "step": 10987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.73832316018019e-06,
+      "loss": 0.8335,
+      "step": 10988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.737639149924988e-06,
+      "loss": 0.8818,
+      "step": 10989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.736955125557589e-06,
+      "loss": 0.8643,
+      "step": 10990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.736271087091077e-06,
+      "loss": 0.897,
+      "step": 10991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.735587034538536e-06,
+      "loss": 0.8462,
+      "step": 10992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.734902967913058e-06,
+      "loss": 0.9272,
+      "step": 10993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.734218887227727e-06,
+      "loss": 0.8882,
+      "step": 10994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.733534792495633e-06,
+      "loss": 0.8887,
+      "step": 10995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.73285068372986e-06,
+      "loss": 1.0005,
+      "step": 10996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732166560943502e-06,
+      "loss": 0.8857,
+      "step": 10997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.731482424149642e-06,
+      "loss": 0.9229,
+      "step": 10998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730798273361371e-06,
+      "loss": 0.8286,
+      "step": 10999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7301141085917765e-06,
+      "loss": 0.9282,
+      "step": 11000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.72942992985395e-06,
+      "loss": 0.8955,
+      "step": 11001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.728745737160979e-06,
+      "loss": 0.8691,
+      "step": 11002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7280615305259535e-06,
+      "loss": 0.8687,
+      "step": 11003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7273773099619645e-06,
+      "loss": 0.8232,
+      "step": 11004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726693075482099e-06,
+      "loss": 0.8428,
+      "step": 11005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726008827099452e-06,
+      "loss": 0.8618,
+      "step": 11006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.725324564827113e-06,
+      "loss": 0.9155,
+      "step": 11007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724640288678171e-06,
+      "loss": 0.8989,
+      "step": 11008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.723955998665716e-06,
+      "loss": 0.8706,
+      "step": 11009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.723271694802846e-06,
+      "loss": 0.9141,
+      "step": 11010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7225873771026455e-06,
+      "loss": 0.9058,
+      "step": 11011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721903045578211e-06,
+      "loss": 0.8979,
+      "step": 11012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7212187002426335e-06,
+      "loss": 0.8491,
+      "step": 11013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.720534341109006e-06,
+      "loss": 0.8796,
+      "step": 11014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7198499681904205e-06,
+      "loss": 0.8564,
+      "step": 11015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719165581499973e-06,
+      "loss": 0.8657,
+      "step": 11016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.718481181050752e-06,
+      "loss": 0.8779,
+      "step": 11017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.717796766855854e-06,
+      "loss": 0.7627,
+      "step": 11018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.717112338928375e-06,
+      "loss": 0.9072,
+      "step": 11019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.716427897281405e-06,
+      "loss": 0.8357,
+      "step": 11020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.715743441928041e-06,
+      "loss": 0.9214,
+      "step": 11021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.715058972881376e-06,
+      "loss": 0.8584,
+      "step": 11022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.714374490154508e-06,
+      "loss": 0.9238,
+      "step": 11023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713689993760529e-06,
+      "loss": 0.9146,
+      "step": 11024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713005483712536e-06,
+      "loss": 0.793,
+      "step": 11025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.712320960023625e-06,
+      "loss": 0.8594,
+      "step": 11026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.711636422706891e-06,
+      "loss": 0.874,
+      "step": 11027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710951871775432e-06,
+      "loss": 0.8472,
+      "step": 11028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710267307242342e-06,
+      "loss": 0.896,
+      "step": 11029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70958272912072e-06,
+      "loss": 0.7739,
+      "step": 11030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7088981374236605e-06,
+      "loss": 0.8374,
+      "step": 11031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708213532164264e-06,
+      "loss": 0.9048,
+      "step": 11032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.707528913355627e-06,
+      "loss": 0.8545,
+      "step": 11033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7068442810108456e-06,
+      "loss": 0.8716,
+      "step": 11034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7061596351430206e-06,
+      "loss": 0.8511,
+      "step": 11035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70547497576525e-06,
+      "loss": 0.8716,
+      "step": 11036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7047903028906295e-06,
+      "loss": 0.8369,
+      "step": 11037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.704105616532259e-06,
+      "loss": 0.8013,
+      "step": 11038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.703420916703241e-06,
+      "loss": 0.9062,
+      "step": 11039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.702736203416671e-06,
+      "loss": 0.9336,
+      "step": 11040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.702051476685651e-06,
+      "loss": 0.8892,
+      "step": 11041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.701366736523278e-06,
+      "loss": 0.8657,
+      "step": 11042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.700681982942655e-06,
+      "loss": 0.9028,
+      "step": 11043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.699997215956883e-06,
+      "loss": 0.9004,
+      "step": 11044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.69931243557906e-06,
+      "loss": 0.9609,
+      "step": 11045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6986276418222875e-06,
+      "loss": 0.3153,
+      "step": 11046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697942834699668e-06,
+      "loss": 0.8594,
+      "step": 11047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697258014224303e-06,
+      "loss": 0.8701,
+      "step": 11048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.696573180409293e-06,
+      "loss": 0.8667,
+      "step": 11049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695888333267739e-06,
+      "loss": 0.8745,
+      "step": 11050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695203472812746e-06,
+      "loss": 0.8521,
+      "step": 11051
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.694518599057416e-06,
+      "loss": 0.3779,
+      "step": 11052
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.693833712014849e-06,
+      "loss": 0.9575,
+      "step": 11053
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.693148811698149e-06,
+      "loss": 0.9004,
+      "step": 11054
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.692463898120422e-06,
+      "loss": 0.873,
+      "step": 11055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6917789712947695e-06,
+      "loss": 0.8013,
+      "step": 11056
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.691094031234294e-06,
+      "loss": 0.8809,
+      "step": 11057
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6904090779521e-06,
+      "loss": 0.874,
+      "step": 11058
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.689724111461295e-06,
+      "loss": 0.8638,
+      "step": 11059
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.689039131774978e-06,
+      "loss": 0.8354,
+      "step": 11060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.688354138906258e-06,
+      "loss": 0.8198,
+      "step": 11061
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6876691328682384e-06,
+      "loss": 0.8599,
+      "step": 11062
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.686984113674026e-06,
+      "loss": 0.8545,
+      "step": 11063
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.686299081336724e-06,
+      "loss": 0.9043,
+      "step": 11064
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.685614035869437e-06,
+      "loss": 0.8848,
+      "step": 11065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6849289772852756e-06,
+      "loss": 0.8975,
+      "step": 11066
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.684243905597343e-06,
+      "loss": 0.8633,
+      "step": 11067
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6835588208187465e-06,
+      "loss": 0.9038,
+      "step": 11068
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.682873722962593e-06,
+      "loss": 0.8433,
+      "step": 11069
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.682188612041988e-06,
+      "loss": 0.7903,
+      "step": 11070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.681503488070039e-06,
+      "loss": 0.8901,
+      "step": 11071
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6808183510598565e-06,
+      "loss": 0.8296,
+      "step": 11072
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.680133201024546e-06,
+      "loss": 0.9106,
+      "step": 11073
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.679448037977214e-06,
+      "loss": 0.874,
+      "step": 11074
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.678762861930971e-06,
+      "loss": 0.8657,
+      "step": 11075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.678077672898925e-06,
+      "loss": 0.8408,
+      "step": 11076
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.677392470894186e-06,
+      "loss": 0.9175,
+      "step": 11077
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.676707255929858e-06,
+      "loss": 0.8555,
+      "step": 11078
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.676022028019057e-06,
+      "loss": 0.8994,
+      "step": 11079
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675336787174888e-06,
+      "loss": 0.8652,
+      "step": 11080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.674651533410462e-06,
+      "loss": 0.8555,
+      "step": 11081
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.673966266738891e-06,
+      "loss": 0.7676,
+      "step": 11082
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.673280987173281e-06,
+      "loss": 0.8403,
+      "step": 11083
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6725956947267445e-06,
+      "loss": 0.8862,
+      "step": 11084
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.671910389412395e-06,
+      "loss": 0.8496,
+      "step": 11085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.671225071243339e-06,
+      "loss": 0.8872,
+      "step": 11086
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.670539740232689e-06,
+      "loss": 0.8896,
+      "step": 11087
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669854396393559e-06,
+      "loss": 0.9419,
+      "step": 11088
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669169039739059e-06,
+      "loss": 0.8652,
+      "step": 11089
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.668483670282299e-06,
+      "loss": 0.8271,
+      "step": 11090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.667798288036394e-06,
+      "loss": 0.8262,
+      "step": 11091
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.667112893014455e-06,
+      "loss": 0.8901,
+      "step": 11092
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.666427485229595e-06,
+      "loss": 0.9102,
+      "step": 11093
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665742064694928e-06,
+      "loss": 0.8438,
+      "step": 11094
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665056631423566e-06,
+      "loss": 0.8657,
+      "step": 11095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.664371185428622e-06,
+      "loss": 0.8931,
+      "step": 11096
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66368572672321e-06,
+      "loss": 0.8325,
+      "step": 11097
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6630002553204465e-06,
+      "loss": 0.8545,
+      "step": 11098
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66231477123344e-06,
+      "loss": 0.8325,
+      "step": 11099
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66162927447531e-06,
+      "loss": 0.8574,
+      "step": 11100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66094376505917e-06,
+      "loss": 0.894,
+      "step": 11101
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.660258242998133e-06,
+      "loss": 0.8447,
+      "step": 11102
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.659572708305315e-06,
+      "loss": 0.9233,
+      "step": 11103
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.658887160993835e-06,
+      "loss": 0.9019,
+      "step": 11104
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.658201601076803e-06,
+      "loss": 0.8613,
+      "step": 11105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.657516028567337e-06,
+      "loss": 0.8535,
+      "step": 11106
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.656830443478553e-06,
+      "loss": 0.9102,
+      "step": 11107
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.656144845823568e-06,
+      "loss": 0.8921,
+      "step": 11108
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.655459235615498e-06,
+      "loss": 0.8818,
+      "step": 11109
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65477361286746e-06,
+      "loss": 0.8965,
+      "step": 11110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.654087977592573e-06,
+      "loss": 0.9043,
+      "step": 11111
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.653402329803949e-06,
+      "loss": 0.9219,
+      "step": 11112
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.652716669514709e-06,
+      "loss": 0.8633,
+      "step": 11113
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.652030996737972e-06,
+      "loss": 0.8579,
+      "step": 11114
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.651345311486855e-06,
+      "loss": 0.325,
+      "step": 11115
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.650659613774474e-06,
+      "loss": 0.8594,
+      "step": 11116
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6499739036139515e-06,
+      "loss": 0.8921,
+      "step": 11117
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.649288181018403e-06,
+      "loss": 0.8438,
+      "step": 11118
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.648602446000948e-06,
+      "loss": 0.9082,
+      "step": 11119
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.647916698574706e-06,
+      "loss": 0.8271,
+      "step": 11120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.647230938752798e-06,
+      "loss": 0.894,
+      "step": 11121
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.646545166548341e-06,
+      "loss": 0.897,
+      "step": 11122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.645859381974457e-06,
+      "loss": 0.9141,
+      "step": 11123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.645173585044263e-06,
+      "loss": 0.894,
+      "step": 11124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.644487775770884e-06,
+      "loss": 0.8926,
+      "step": 11125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.643801954167437e-06,
+      "loss": 0.8989,
+      "step": 11126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.643116120247045e-06,
+      "loss": 0.8774,
+      "step": 11127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.642430274022827e-06,
+      "loss": 0.8735,
+      "step": 11128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641744415507907e-06,
+      "loss": 0.9248,
+      "step": 11129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6410585447154045e-06,
+      "loss": 0.8462,
+      "step": 11130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.640372661658443e-06,
+      "loss": 0.8403,
+      "step": 11131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.639686766350141e-06,
+      "loss": 0.8672,
+      "step": 11132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6390008588036234e-06,
+      "loss": 0.8071,
+      "step": 11133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.638314939032014e-06,
+      "loss": 0.9229,
+      "step": 11134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.637629007048434e-06,
+      "loss": 0.918,
+      "step": 11135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.636943062866005e-06,
+      "loss": 0.8506,
+      "step": 11136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6362571064978535e-06,
+      "loss": 0.8911,
+      "step": 11137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635571137957099e-06,
+      "loss": 0.8911,
+      "step": 11138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.634885157256867e-06,
+      "loss": 0.8809,
+      "step": 11139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.634199164410283e-06,
+      "loss": 0.8345,
+      "step": 11140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.633513159430468e-06,
+      "loss": 0.8594,
+      "step": 11141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.632827142330549e-06,
+      "loss": 0.9053,
+      "step": 11142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.632141113123649e-06,
+      "loss": 0.875,
+      "step": 11143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6314550718228954e-06,
+      "loss": 0.8726,
+      "step": 11144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630769018441408e-06,
+      "loss": 0.8872,
+      "step": 11145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630082952992317e-06,
+      "loss": 0.9233,
+      "step": 11146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.629396875488746e-06,
+      "loss": 0.9141,
+      "step": 11147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628710785943821e-06,
+      "loss": 0.8696,
+      "step": 11148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628024684370668e-06,
+      "loss": 0.8696,
+      "step": 11149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.627338570782414e-06,
+      "loss": 0.9346,
+      "step": 11150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.626652445192183e-06,
+      "loss": 0.8447,
+      "step": 11151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.625966307613104e-06,
+      "loss": 0.8105,
+      "step": 11152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6252801580583025e-06,
+      "loss": 0.8711,
+      "step": 11153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624593996540907e-06,
+      "loss": 0.9199,
+      "step": 11154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.623907823074044e-06,
+      "loss": 0.854,
+      "step": 11155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.623221637670842e-06,
+      "loss": 0.8535,
+      "step": 11156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.622535440344426e-06,
+      "loss": 0.8394,
+      "step": 11157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.621849231107928e-06,
+      "loss": 0.8555,
+      "step": 11158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.621163009974473e-06,
+      "loss": 0.8281,
+      "step": 11159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.620476776957192e-06,
+      "loss": 0.8398,
+      "step": 11160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.61979053206921e-06,
+      "loss": 0.9072,
+      "step": 11161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.61910427532366e-06,
+      "loss": 0.8726,
+      "step": 11162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.618418006733672e-06,
+      "loss": 0.8228,
+      "step": 11163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.61773172631237e-06,
+      "loss": 0.8618,
+      "step": 11164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.617045434072886e-06,
+      "loss": 0.8726,
+      "step": 11165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.616359130028353e-06,
+      "loss": 0.8628,
+      "step": 11166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615672814191898e-06,
+      "loss": 1.002,
+      "step": 11167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6149864865766515e-06,
+      "loss": 0.9604,
+      "step": 11168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.614300147195743e-06,
+      "loss": 0.8584,
+      "step": 11169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.613613796062307e-06,
+      "loss": 0.9062,
+      "step": 11170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.612927433189471e-06,
+      "loss": 0.8276,
+      "step": 11171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6122410585903675e-06,
+      "loss": 0.8652,
+      "step": 11172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.611554672278128e-06,
+      "loss": 0.8232,
+      "step": 11173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6108682742658835e-06,
+      "loss": 0.8779,
+      "step": 11174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.610181864566766e-06,
+      "loss": 0.8335,
+      "step": 11175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.609495443193909e-06,
+      "loss": 0.876,
+      "step": 11176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.608809010160444e-06,
+      "loss": 0.894,
+      "step": 11177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.608122565479501e-06,
+      "loss": 0.9053,
+      "step": 11178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.607436109164216e-06,
+      "loss": 0.8779,
+      "step": 11179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.606749641227723e-06,
+      "loss": 0.3578,
+      "step": 11180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.606063161683152e-06,
+      "loss": 0.8525,
+      "step": 11181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.605376670543636e-06,
+      "loss": 0.8262,
+      "step": 11182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6046901678223135e-06,
+      "loss": 0.9077,
+      "step": 11183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.604003653532312e-06,
+      "loss": 0.8574,
+      "step": 11184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.60331712768677e-06,
+      "loss": 0.9155,
+      "step": 11185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.602630590298822e-06,
+      "loss": 0.7397,
+      "step": 11186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6019440413816e-06,
+      "loss": 0.8755,
+      "step": 11187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.60125748094824e-06,
+      "loss": 0.9189,
+      "step": 11188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.600570909011877e-06,
+      "loss": 0.8975,
+      "step": 11189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599884325585646e-06,
+      "loss": 0.4115,
+      "step": 11190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599197730682682e-06,
+      "loss": 0.8267,
+      "step": 11191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.598511124316122e-06,
+      "loss": 0.8135,
+      "step": 11192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597824506499102e-06,
+      "loss": 0.8423,
+      "step": 11193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597137877244755e-06,
+      "loss": 0.9492,
+      "step": 11194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.596451236566221e-06,
+      "loss": 0.9155,
+      "step": 11195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.595764584476635e-06,
+      "loss": 0.8467,
+      "step": 11196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5950779209891324e-06,
+      "loss": 0.8325,
+      "step": 11197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5943912461168516e-06,
+      "loss": 0.8213,
+      "step": 11198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5937045598729314e-06,
+      "loss": 0.8931,
+      "step": 11199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.593017862270505e-06,
+      "loss": 0.8862,
+      "step": 11200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.592331153322714e-06,
+      "loss": 0.9463,
+      "step": 11201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.591644433042694e-06,
+      "loss": 0.8633,
+      "step": 11202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5909577014435845e-06,
+      "loss": 0.9238,
+      "step": 11203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.590270958538522e-06,
+      "loss": 0.8721,
+      "step": 11204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.589584204340647e-06,
+      "loss": 0.8721,
+      "step": 11205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.588897438863097e-06,
+      "loss": 0.9097,
+      "step": 11206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.588210662119009e-06,
+      "loss": 0.8794,
+      "step": 11207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.587523874121526e-06,
+      "loss": 0.894,
+      "step": 11208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.586837074883785e-06,
+      "loss": 0.8799,
+      "step": 11209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5861502644189255e-06,
+      "loss": 0.9082,
+      "step": 11210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.585463442740088e-06,
+      "loss": 0.896,
+      "step": 11211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.584776609860414e-06,
+      "loss": 0.8765,
+      "step": 11212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.58408976579304e-06,
+      "loss": 0.9141,
+      "step": 11213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.583402910551109e-06,
+      "loss": 0.9087,
+      "step": 11214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.58271604414776e-06,
+      "loss": 0.8687,
+      "step": 11215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.582029166596136e-06,
+      "loss": 0.8374,
+      "step": 11216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5813422779093765e-06,
+      "loss": 0.8906,
+      "step": 11217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.580655378100624e-06,
+      "loss": 0.8735,
+      "step": 11218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5799684671830176e-06,
+      "loss": 0.9434,
+      "step": 11219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.579281545169702e-06,
+      "loss": 0.9482,
+      "step": 11220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.578594612073815e-06,
+      "loss": 0.875,
+      "step": 11221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.577907667908505e-06,
+      "loss": 0.8999,
+      "step": 11222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5772207126869085e-06,
+      "loss": 0.9058,
+      "step": 11223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.576533746422171e-06,
+      "loss": 0.8618,
+      "step": 11224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.575846769127432e-06,
+      "loss": 0.9189,
+      "step": 11225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.575159780815841e-06,
+      "loss": 0.8547,
+      "step": 11226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574472781500533e-06,
+      "loss": 0.8516,
+      "step": 11227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5737857711946575e-06,
+      "loss": 0.9111,
+      "step": 11228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.573098749911356e-06,
+      "loss": 0.7729,
+      "step": 11229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5724117176637725e-06,
+      "loss": 0.8521,
+      "step": 11230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.57172467446505e-06,
+      "loss": 0.8447,
+      "step": 11231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.571037620328335e-06,
+      "loss": 0.8921,
+      "step": 11232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.570350555266769e-06,
+      "loss": 0.8428,
+      "step": 11233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.569663479293499e-06,
+      "loss": 0.8555,
+      "step": 11234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.568976392421668e-06,
+      "loss": 0.9019,
+      "step": 11235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.568289294664423e-06,
+      "loss": 0.843,
+      "step": 11236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.567602186034907e-06,
+      "loss": 0.9019,
+      "step": 11237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.566915066546268e-06,
+      "loss": 0.8862,
+      "step": 11238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.566227936211651e-06,
+      "loss": 0.3544,
+      "step": 11239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5655407950442e-06,
+      "loss": 0.8442,
+      "step": 11240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5648536430570624e-06,
+      "loss": 0.8276,
+      "step": 11241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.564166480263385e-06,
+      "loss": 0.8813,
+      "step": 11242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.563479306676314e-06,
+      "loss": 0.873,
+      "step": 11243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.562792122308995e-06,
+      "loss": 0.8511,
+      "step": 11244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.562104927174577e-06,
+      "loss": 0.8374,
+      "step": 11245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561417721286206e-06,
+      "loss": 0.8882,
+      "step": 11246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.560730504657028e-06,
+      "loss": 0.9434,
+      "step": 11247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.560043277300193e-06,
+      "loss": 0.781,
+      "step": 11248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559356039228847e-06,
+      "loss": 0.8657,
+      "step": 11249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5586687904561386e-06,
+      "loss": 0.8711,
+      "step": 11250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.557981530995216e-06,
+      "loss": 0.8823,
+      "step": 11251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.557294260859227e-06,
+      "loss": 0.8872,
+      "step": 11252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.556606980061321e-06,
+      "loss": 0.8892,
+      "step": 11253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.555919688614647e-06,
+      "loss": 0.8989,
+      "step": 11254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.555232386532354e-06,
+      "loss": 0.8306,
+      "step": 11255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.554545073827587e-06,
+      "loss": 0.8516,
+      "step": 11256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.553857750513502e-06,
+      "loss": 0.9233,
+      "step": 11257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.553170416603245e-06,
+      "loss": 0.8569,
+      "step": 11258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552483072109966e-06,
+      "loss": 0.9258,
+      "step": 11259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.551795717046814e-06,
+      "loss": 0.8184,
+      "step": 11260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.551108351426942e-06,
+      "loss": 0.8799,
+      "step": 11261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550420975263498e-06,
+      "loss": 0.8662,
+      "step": 11262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.549733588569635e-06,
+      "loss": 0.8857,
+      "step": 11263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5490461913585e-06,
+      "loss": 0.8799,
+      "step": 11264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.548358783643247e-06,
+      "loss": 0.894,
+      "step": 11265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.547671365437026e-06,
+      "loss": 0.8765,
+      "step": 11266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.54698393675299e-06,
+      "loss": 0.8452,
+      "step": 11267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.546296497604289e-06,
+      "loss": 0.873,
+      "step": 11268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.545609048004074e-06,
+      "loss": 0.9131,
+      "step": 11269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.544921587965499e-06,
+      "loss": 0.8008,
+      "step": 11270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.544234117501716e-06,
+      "loss": 0.8491,
+      "step": 11271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.543546636625875e-06,
+      "loss": 0.8652,
+      "step": 11272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.54285914535113e-06,
+      "loss": 0.8716,
+      "step": 11273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.542171643690637e-06,
+      "loss": 0.8633,
+      "step": 11274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.541484131657544e-06,
+      "loss": 0.8633,
+      "step": 11275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.540796609265006e-06,
+      "loss": 0.8477,
+      "step": 11276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.540109076526177e-06,
+      "loss": 0.8794,
+      "step": 11277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.539421533454211e-06,
+      "loss": 0.8677,
+      "step": 11278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.538733980062261e-06,
+      "loss": 0.7979,
+      "step": 11279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5380464163634805e-06,
+      "loss": 0.8145,
+      "step": 11280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.537358842371024e-06,
+      "loss": 0.8872,
+      "step": 11281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.536671258098046e-06,
+      "loss": 0.8799,
+      "step": 11282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.535983663557702e-06,
+      "loss": 0.8872,
+      "step": 11283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.535296058763145e-06,
+      "loss": 0.3823,
+      "step": 11284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5346084437275316e-06,
+      "loss": 0.8579,
+      "step": 11285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.533920818464017e-06,
+      "loss": 0.7861,
+      "step": 11286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.533233182985754e-06,
+      "loss": 0.79,
+      "step": 11287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.532545537305902e-06,
+      "loss": 0.9458,
+      "step": 11288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.531857881437612e-06,
+      "loss": 0.8374,
+      "step": 11289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.531170215394044e-06,
+      "loss": 0.8896,
+      "step": 11290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.530482539188354e-06,
+      "loss": 0.8677,
+      "step": 11291
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.529794852833695e-06,
+      "loss": 0.9238,
+      "step": 11292
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5291071563432265e-06,
+      "loss": 0.9341,
+      "step": 11293
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.528419449730104e-06,
+      "loss": 0.8208,
+      "step": 11294
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527731733007486e-06,
+      "loss": 0.8281,
+      "step": 11295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527044006188528e-06,
+      "loss": 0.8247,
+      "step": 11296
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.526356269286386e-06,
+      "loss": 0.8115,
+      "step": 11297
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.525668522314221e-06,
+      "loss": 0.8887,
+      "step": 11298
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5249807652851876e-06,
+      "loss": 0.9321,
+      "step": 11299
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.524292998212445e-06,
+      "loss": 0.8633,
+      "step": 11300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.523605221109151e-06,
+      "loss": 0.8594,
+      "step": 11301
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.522917433988465e-06,
+      "loss": 0.8667,
+      "step": 11302
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.522229636863544e-06,
+      "loss": 0.9048,
+      "step": 11303
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.521541829747548e-06,
+      "loss": 0.8418,
+      "step": 11304
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.520854012653634e-06,
+      "loss": 0.834,
+      "step": 11305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.520166185594961e-06,
+      "loss": 0.8818,
+      "step": 11306
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51947834858469e-06,
+      "loss": 0.9194,
+      "step": 11307
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.518790501635981e-06,
+      "loss": 0.8608,
+      "step": 11308
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5181026447619904e-06,
+      "loss": 0.8892,
+      "step": 11309
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51741477797588e-06,
+      "loss": 0.916,
+      "step": 11310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.516726901290809e-06,
+      "loss": 0.3423,
+      "step": 11311
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51603901471994e-06,
+      "loss": 0.856,
+      "step": 11312
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5153511182764306e-06,
+      "loss": 0.8525,
+      "step": 11313
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.514663211973442e-06,
+      "loss": 0.8643,
+      "step": 11314
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.513975295824136e-06,
+      "loss": 0.8516,
+      "step": 11315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.513287369841671e-06,
+      "loss": 0.8359,
+      "step": 11316
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.512599434039212e-06,
+      "loss": 0.8516,
+      "step": 11317
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.511911488429915e-06,
+      "loss": 0.8774,
+      "step": 11318
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.511223533026947e-06,
+      "loss": 0.9148,
+      "step": 11319
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.510535567843467e-06,
+      "loss": 0.8145,
+      "step": 11320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509847592892636e-06,
+      "loss": 0.9429,
+      "step": 11321
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5091596081876175e-06,
+      "loss": 0.8462,
+      "step": 11322
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.508471613741575e-06,
+      "loss": 0.8691,
+      "step": 11323
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.507783609567667e-06,
+      "loss": 0.8652,
+      "step": 11324
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.507095595679059e-06,
+      "loss": 0.8228,
+      "step": 11325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.506407572088914e-06,
+      "loss": 0.8804,
+      "step": 11326
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.505719538810394e-06,
+      "loss": 0.854,
+      "step": 11327
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5050314958566606e-06,
+      "loss": 0.916,
+      "step": 11328
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50434344324088e-06,
+      "loss": 0.8135,
+      "step": 11329
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.503655380976216e-06,
+      "loss": 0.917,
+      "step": 11330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.502967309075829e-06,
+      "loss": 0.8828,
+      "step": 11331
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5022792275528855e-06,
+      "loss": 0.9487,
+      "step": 11332
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50159113642055e-06,
+      "loss": 0.9219,
+      "step": 11333
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.500903035691984e-06,
+      "loss": 0.8994,
+      "step": 11334
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.500214925380353e-06,
+      "loss": 0.8564,
+      "step": 11335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.499526805498824e-06,
+      "loss": 0.8398,
+      "step": 11336
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.498838676060559e-06,
+      "loss": 0.8433,
+      "step": 11337
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.498150537078724e-06,
+      "loss": 0.8901,
+      "step": 11338
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4974623885664844e-06,
+      "loss": 0.8281,
+      "step": 11339
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496774230537007e-06,
+      "loss": 0.8481,
+      "step": 11340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496086063003454e-06,
+      "loss": 0.9399,
+      "step": 11341
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.495397885978994e-06,
+      "loss": 0.8662,
+      "step": 11342
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49470969947679e-06,
+      "loss": 0.8364,
+      "step": 11343
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.494021503510012e-06,
+      "loss": 0.8774,
+      "step": 11344
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.493333298091823e-06,
+      "loss": 0.874,
+      "step": 11345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49264508323539e-06,
+      "loss": 0.8584,
+      "step": 11346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491956858953881e-06,
+      "loss": 0.9126,
+      "step": 11347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491268625260462e-06,
+      "loss": 0.8672,
+      "step": 11348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4905803821683e-06,
+      "loss": 0.8457,
+      "step": 11349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.489892129690564e-06,
+      "loss": 0.8237,
+      "step": 11350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4892038678404156e-06,
+      "loss": 0.9058,
+      "step": 11351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.488515596631029e-06,
+      "loss": 0.8989,
+      "step": 11352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.48782731607557e-06,
+      "loss": 0.8911,
+      "step": 11353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.487139026187205e-06,
+      "loss": 0.8989,
+      "step": 11354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4864507269791016e-06,
+      "loss": 0.8452,
+      "step": 11355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.48576241846443e-06,
+      "loss": 0.8696,
+      "step": 11356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.485074100656358e-06,
+      "loss": 0.9106,
+      "step": 11357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.484385773568054e-06,
+      "loss": 0.9033,
+      "step": 11358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.483697437212687e-06,
+      "loss": 0.2751,
+      "step": 11359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.483009091603425e-06,
+      "loss": 0.8579,
+      "step": 11360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.482320736753439e-06,
+      "loss": 0.8491,
+      "step": 11361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.481632372675897e-06,
+      "loss": 0.877,
+      "step": 11362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480943999383969e-06,
+      "loss": 0.8789,
+      "step": 11363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480255616890823e-06,
+      "loss": 0.9014,
+      "step": 11364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.479567225209631e-06,
+      "loss": 0.8496,
+      "step": 11365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478878824353564e-06,
+      "loss": 0.9048,
+      "step": 11366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478190414335788e-06,
+      "loss": 0.842,
+      "step": 11367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.477501995169476e-06,
+      "loss": 0.8604,
+      "step": 11368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4768135668678015e-06,
+      "loss": 0.7993,
+      "step": 11369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476125129443929e-06,
+      "loss": 0.8789,
+      "step": 11370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.475436682911033e-06,
+      "loss": 0.8911,
+      "step": 11371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.474748227282285e-06,
+      "loss": 0.9092,
+      "step": 11372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.474059762570855e-06,
+      "loss": 0.8286,
+      "step": 11373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.473371288789914e-06,
+      "loss": 0.8623,
+      "step": 11374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.472682805952634e-06,
+      "loss": 0.8926,
+      "step": 11375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.471994314072187e-06,
+      "loss": 0.9165,
+      "step": 11376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.471305813161747e-06,
+      "loss": 0.9194,
+      "step": 11377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.470617303234482e-06,
+      "loss": 0.8638,
+      "step": 11378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469928784303568e-06,
+      "loss": 0.9136,
+      "step": 11379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469240256382174e-06,
+      "loss": 0.8994,
+      "step": 11380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.468551719483477e-06,
+      "loss": 0.897,
+      "step": 11381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.467863173620646e-06,
+      "loss": 0.9077,
+      "step": 11382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.467174618806856e-06,
+      "loss": 0.8433,
+      "step": 11383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4664860550552765e-06,
+      "loss": 0.9233,
+      "step": 11384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.465797482379087e-06,
+      "loss": 0.8765,
+      "step": 11385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.465108900791456e-06,
+      "loss": 0.8818,
+      "step": 11386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4644203103055594e-06,
+      "loss": 0.8223,
+      "step": 11387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.463731710934569e-06,
+      "loss": 0.9111,
+      "step": 11388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.463043102691662e-06,
+      "loss": 0.9097,
+      "step": 11389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4623544855900105e-06,
+      "loss": 0.8242,
+      "step": 11390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.461665859642788e-06,
+      "loss": 0.9233,
+      "step": 11391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.460977224863171e-06,
+      "loss": 0.8667,
+      "step": 11392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.460288581264333e-06,
+      "loss": 0.3156,
+      "step": 11393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4595999288594496e-06,
+      "loss": 0.8589,
+      "step": 11394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.458911267661694e-06,
+      "loss": 0.791,
+      "step": 11395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.458222597684244e-06,
+      "loss": 0.855,
+      "step": 11396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.457533918940273e-06,
+      "loss": 0.8208,
+      "step": 11397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.456845231442958e-06,
+      "loss": 0.8452,
+      "step": 11398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.456156535205473e-06,
+      "loss": 0.8921,
+      "step": 11399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4554678302409945e-06,
+      "loss": 0.8335,
+      "step": 11400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454779116562699e-06,
+      "loss": 0.8765,
+      "step": 11401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454090394183764e-06,
+      "loss": 0.8755,
+      "step": 11402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.453401663117362e-06,
+      "loss": 0.8711,
+      "step": 11403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.452712923376672e-06,
+      "loss": 0.9497,
+      "step": 11404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.452024174974872e-06,
+      "loss": 0.8433,
+      "step": 11405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.451335417925135e-06,
+      "loss": 0.8857,
+      "step": 11406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4506466522406406e-06,
+      "loss": 0.8887,
+      "step": 11407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.449957877934568e-06,
+      "loss": 0.8423,
+      "step": 11408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.449269095020089e-06,
+      "loss": 0.3645,
+      "step": 11409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.448580303510386e-06,
+      "loss": 0.9087,
+      "step": 11410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.447891503418634e-06,
+      "loss": 0.8618,
+      "step": 11411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.447202694758014e-06,
+      "loss": 0.856,
+      "step": 11412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.446513877541699e-06,
+      "loss": 0.8486,
+      "step": 11413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44582505178287e-06,
+      "loss": 0.8521,
+      "step": 11414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.445136217494707e-06,
+      "loss": 0.7944,
+      "step": 11415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.444447374690387e-06,
+      "loss": 0.8569,
+      "step": 11416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443758523383086e-06,
+      "loss": 0.8765,
+      "step": 11417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443069663585987e-06,
+      "loss": 0.8652,
+      "step": 11418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.442380795312267e-06,
+      "loss": 0.8408,
+      "step": 11419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.441691918575105e-06,
+      "loss": 0.856,
+      "step": 11420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44100303338768e-06,
+      "loss": 0.918,
+      "step": 11421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.440314139763172e-06,
+      "loss": 0.8901,
+      "step": 11422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4396252377147615e-06,
+      "loss": 0.8398,
+      "step": 11423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.438936327255628e-06,
+      "loss": 0.8916,
+      "step": 11424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.438247408398949e-06,
+      "loss": 0.8379,
+      "step": 11425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.437558481157907e-06,
+      "loss": 0.8799,
+      "step": 11426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.436869545545682e-06,
+      "loss": 0.8994,
+      "step": 11427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4361806015754555e-06,
+      "loss": 0.9121,
+      "step": 11428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.435491649260405e-06,
+      "loss": 0.9038,
+      "step": 11429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.434802688613713e-06,
+      "loss": 0.8491,
+      "step": 11430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.434113719648563e-06,
+      "loss": 0.8838,
+      "step": 11431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4334247423781304e-06,
+      "loss": 0.8354,
+      "step": 11432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.432735756815601e-06,
+      "loss": 0.8828,
+      "step": 11433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.432046762974155e-06,
+      "loss": 0.8521,
+      "step": 11434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.431357760866972e-06,
+      "loss": 0.3478,
+      "step": 11435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430668750507236e-06,
+      "loss": 0.8569,
+      "step": 11436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4299797319081285e-06,
+      "loss": 0.9136,
+      "step": 11437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.42929070508283e-06,
+      "loss": 0.9033,
+      "step": 11438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4286016700445255e-06,
+      "loss": 0.8569,
+      "step": 11439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.427912626806394e-06,
+      "loss": 0.8198,
+      "step": 11440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.427223575381623e-06,
+      "loss": 0.875,
+      "step": 11441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426534515783387e-06,
+      "loss": 0.8809,
+      "step": 11442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.425845448024877e-06,
+      "loss": 0.9204,
+      "step": 11443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.425156372119272e-06,
+      "loss": 0.791,
+      "step": 11444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.424467288079755e-06,
+      "loss": 0.8521,
+      "step": 11445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.423778195919509e-06,
+      "loss": 0.9351,
+      "step": 11446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.423089095651719e-06,
+      "loss": 0.8833,
+      "step": 11447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.422399987289569e-06,
+      "loss": 0.8828,
+      "step": 11448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.421710870846242e-06,
+      "loss": 0.8198,
+      "step": 11449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.421021746334918e-06,
+      "loss": 0.8867,
+      "step": 11450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.420332613768788e-06,
+      "loss": 0.8789,
+      "step": 11451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.419643473161031e-06,
+      "loss": 0.8525,
+      "step": 11452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.418954324524834e-06,
+      "loss": 0.8848,
+      "step": 11453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.418265167873381e-06,
+      "loss": 0.8516,
+      "step": 11454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.417576003219856e-06,
+      "loss": 0.8926,
+      "step": 11455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.416886830577442e-06,
+      "loss": 0.9121,
+      "step": 11456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4161976499593285e-06,
+      "loss": 0.8237,
+      "step": 11457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.415508461378697e-06,
+      "loss": 0.8633,
+      "step": 11458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4148192648487344e-06,
+      "loss": 0.835,
+      "step": 11459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.414130060382625e-06,
+      "loss": 0.916,
+      "step": 11460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.413440847993556e-06,
+      "loss": 0.9551,
+      "step": 11461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.41275162769471e-06,
+      "loss": 0.8071,
+      "step": 11462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.412062399499276e-06,
+      "loss": 0.8555,
+      "step": 11463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.411373163420441e-06,
+      "loss": 0.8916,
+      "step": 11464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.410683919471387e-06,
+      "loss": 0.8584,
+      "step": 11465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.409994667665301e-06,
+      "loss": 0.8647,
+      "step": 11466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.409305408015375e-06,
+      "loss": 0.9155,
+      "step": 11467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.408616140534789e-06,
+      "loss": 0.3521,
+      "step": 11468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.407926865236732e-06,
+      "loss": 0.8716,
+      "step": 11469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.407237582134393e-06,
+      "loss": 0.9292,
+      "step": 11470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.406548291240955e-06,
+      "loss": 0.8501,
+      "step": 11471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.405858992569609e-06,
+      "loss": 0.8325,
+      "step": 11472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.40516968613354e-06,
+      "loss": 0.8965,
+      "step": 11473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.404480371945939e-06,
+      "loss": 0.8364,
+      "step": 11474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403791050019988e-06,
+      "loss": 0.8799,
+      "step": 11475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.40310172036888e-06,
+      "loss": 0.8477,
+      "step": 11476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4024123830057995e-06,
+      "loss": 0.8511,
+      "step": 11477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.401723037943937e-06,
+      "loss": 0.8652,
+      "step": 11478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.401033685196478e-06,
+      "loss": 0.9214,
+      "step": 11479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.400344324776615e-06,
+      "loss": 0.8535,
+      "step": 11480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.399654956697533e-06,
+      "loss": 0.897,
+      "step": 11481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.398965580972421e-06,
+      "loss": 0.8853,
+      "step": 11482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39827619761447e-06,
+      "loss": 0.8896,
+      "step": 11483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.397586806636868e-06,
+      "loss": 0.8438,
+      "step": 11484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.396897408052804e-06,
+      "loss": 0.8701,
+      "step": 11485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.396208001875466e-06,
+      "loss": 0.8628,
+      "step": 11486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3955185881180446e-06,
+      "loss": 0.9097,
+      "step": 11487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3948291667937306e-06,
+      "loss": 0.8281,
+      "step": 11488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3941397379157114e-06,
+      "loss": 0.8696,
+      "step": 11489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39345030149718e-06,
+      "loss": 0.9067,
+      "step": 11490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39276085755132e-06,
+      "loss": 0.813,
+      "step": 11491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.392071406091329e-06,
+      "loss": 0.8413,
+      "step": 11492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.391381947130394e-06,
+      "loss": 0.9102,
+      "step": 11493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390692480681706e-06,
+      "loss": 0.8472,
+      "step": 11494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390003006758453e-06,
+      "loss": 0.9355,
+      "step": 11495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.38931352537383e-06,
+      "loss": 0.9072,
+      "step": 11496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.388624036541026e-06,
+      "loss": 0.8379,
+      "step": 11497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.387934540273231e-06,
+      "loss": 0.8555,
+      "step": 11498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.387245036583635e-06,
+      "loss": 0.9014,
+      "step": 11499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.386555525485434e-06,
+      "loss": 0.8569,
+      "step": 11500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3858660069918156e-06,
+      "loss": 0.9199,
+      "step": 11501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.385176481115973e-06,
+      "loss": 0.8628,
+      "step": 11502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.384486947871097e-06,
+      "loss": 0.856,
+      "step": 11503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.383797407270379e-06,
+      "loss": 0.855,
+      "step": 11504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.383107859327012e-06,
+      "loss": 0.739,
+      "step": 11505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.382418304054189e-06,
+      "loss": 0.8618,
+      "step": 11506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.381728741465102e-06,
+      "loss": 0.834,
+      "step": 11507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.38103917157294e-06,
+      "loss": 0.8286,
+      "step": 11508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3803495943909e-06,
+      "loss": 0.8657,
+      "step": 11509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.379660009932173e-06,
+      "loss": 0.8179,
+      "step": 11510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.378970418209951e-06,
+      "loss": 0.8403,
+      "step": 11511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.378280819237427e-06,
+      "loss": 0.8359,
+      "step": 11512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.377591213027797e-06,
+      "loss": 0.8623,
+      "step": 11513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376901599594251e-06,
+      "loss": 0.9048,
+      "step": 11514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376211978949983e-06,
+      "loss": 0.8052,
+      "step": 11515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.375522351108186e-06,
+      "loss": 0.832,
+      "step": 11516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.374832716082057e-06,
+      "loss": 0.8516,
+      "step": 11517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.374143073884787e-06,
+      "loss": 0.8647,
+      "step": 11518
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.373453424529569e-06,
+      "loss": 0.8604,
+      "step": 11519
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3727637680296e-06,
+      "loss": 0.8916,
+      "step": 11520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3720741043980715e-06,
+      "loss": 0.8916,
+      "step": 11521
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37138443364818e-06,
+      "loss": 0.8198,
+      "step": 11522
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37069475579312e-06,
+      "loss": 0.8799,
+      "step": 11523
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.370005070846083e-06,
+      "loss": 0.8906,
+      "step": 11524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.369315378820266e-06,
+      "loss": 0.8286,
+      "step": 11525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.368625679728866e-06,
+      "loss": 0.9146,
+      "step": 11526
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367935973585074e-06,
+      "loss": 0.8525,
+      "step": 11527
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367246260402087e-06,
+      "loss": 0.9263,
+      "step": 11528
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.366556540193102e-06,
+      "loss": 0.8926,
+      "step": 11529
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.365866812971311e-06,
+      "loss": 0.8999,
+      "step": 11530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.365177078749912e-06,
+      "loss": 0.8877,
+      "step": 11531
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3644873375421e-06,
+      "loss": 0.9292,
+      "step": 11532
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.363797589361071e-06,
+      "loss": 0.854,
+      "step": 11533
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.363107834220021e-06,
+      "loss": 0.8652,
+      "step": 11534
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.362418072132145e-06,
+      "loss": 0.8516,
+      "step": 11535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.361728303110642e-06,
+      "loss": 0.7959,
+      "step": 11536
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.361038527168705e-06,
+      "loss": 0.8643,
+      "step": 11537
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.360348744319532e-06,
+      "loss": 0.8394,
+      "step": 11538
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.359658954576321e-06,
+      "loss": 0.7837,
+      "step": 11539
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.358969157952267e-06,
+      "loss": 0.8208,
+      "step": 11540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.358279354460566e-06,
+      "loss": 0.9058,
+      "step": 11541
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3575895441144176e-06,
+      "loss": 0.8584,
+      "step": 11542
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.356899726927017e-06,
+      "loss": 0.8857,
+      "step": 11543
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3562099029115635e-06,
+      "loss": 0.855,
+      "step": 11544
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.355520072081253e-06,
+      "loss": 0.856,
+      "step": 11545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.354830234449282e-06,
+      "loss": 0.9097,
+      "step": 11546
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35414039002885e-06,
+      "loss": 0.8071,
+      "step": 11547
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.353450538833155e-06,
+      "loss": 0.8198,
+      "step": 11548
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.352760680875393e-06,
+      "loss": 0.8662,
+      "step": 11549
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.352070816168764e-06,
+      "loss": 0.7939,
+      "step": 11550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.351380944726465e-06,
+      "loss": 0.9072,
+      "step": 11551
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.350691066561696e-06,
+      "loss": 0.8174,
+      "step": 11552
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.350001181687653e-06,
+      "loss": 0.725,
+      "step": 11553
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.349311290117537e-06,
+      "loss": 0.8926,
+      "step": 11554
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3486213918645455e-06,
+      "loss": 0.9585,
+      "step": 11555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347931486941877e-06,
+      "loss": 0.8887,
+      "step": 11556
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347241575362729e-06,
+      "loss": 0.854,
+      "step": 11557
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.346551657140304e-06,
+      "loss": 0.9258,
+      "step": 11558
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.345861732287801e-06,
+      "loss": 0.9038,
+      "step": 11559
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.345171800818419e-06,
+      "loss": 0.8853,
+      "step": 11560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3444818627453525e-06,
+      "loss": 0.8257,
+      "step": 11561
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343791918081807e-06,
+      "loss": 0.2739,
+      "step": 11562
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343101966840981e-06,
+      "loss": 0.8662,
+      "step": 11563
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.342412009036073e-06,
+      "loss": 0.8784,
+      "step": 11564
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341722044680284e-06,
+      "loss": 0.9106,
+      "step": 11565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341032073786814e-06,
+      "loss": 0.8584,
+      "step": 11566
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.340342096368862e-06,
+      "loss": 0.8901,
+      "step": 11567
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.33965211243963e-06,
+      "loss": 0.8613,
+      "step": 11568
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.338962122012319e-06,
+      "loss": 0.9023,
+      "step": 11569
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3382721251001275e-06,
+      "loss": 0.8105,
+      "step": 11570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.337582121716257e-06,
+      "loss": 0.9136,
+      "step": 11571
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3368921118739105e-06,
+      "loss": 0.8652,
+      "step": 11572
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3362020955862835e-06,
+      "loss": 0.8521,
+      "step": 11573
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.335512072866582e-06,
+      "loss": 0.8896,
+      "step": 11574
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334822043728008e-06,
+      "loss": 0.9326,
+      "step": 11575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334132008183759e-06,
+      "loss": 0.7876,
+      "step": 11576
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.333441966247037e-06,
+      "loss": 0.896,
+      "step": 11577
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332751917931046e-06,
+      "loss": 0.8794,
+      "step": 11578
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332061863248987e-06,
+      "loss": 0.8799,
+      "step": 11579
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.331371802214061e-06,
+      "loss": 0.9199,
+      "step": 11580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3306817348394694e-06,
+      "loss": 0.8164,
+      "step": 11581
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.329991661138415e-06,
+      "loss": 0.9038,
+      "step": 11582
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3293015811241e-06,
+      "loss": 0.8564,
+      "step": 11583
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328611494809728e-06,
+      "loss": 0.9019,
+      "step": 11584
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3279214022085e-06,
+      "loss": 0.8926,
+      "step": 11585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.327231303333618e-06,
+      "loss": 0.9009,
+      "step": 11586
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3265411981982865e-06,
+      "loss": 0.9282,
+      "step": 11587
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.325851086815708e-06,
+      "loss": 0.8247,
+      "step": 11588
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.325160969199084e-06,
+      "loss": 0.8433,
+      "step": 11589
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.324470845361616e-06,
+      "loss": 0.8936,
+      "step": 11590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.323780715316512e-06,
+      "loss": 0.9048,
+      "step": 11591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.323090579076973e-06,
+      "loss": 0.8438,
+      "step": 11592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.322400436656201e-06,
+      "loss": 0.8481,
+      "step": 11593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3217102880673995e-06,
+      "loss": 0.8652,
+      "step": 11594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.321020133323775e-06,
+      "loss": 0.9189,
+      "step": 11595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.32032997243853e-06,
+      "loss": 0.8247,
+      "step": 11596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.319639805424865e-06,
+      "loss": 0.8999,
+      "step": 11597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.31894963229599e-06,
+      "loss": 0.8501,
+      "step": 11598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.318259453065104e-06,
+      "loss": 0.8423,
+      "step": 11599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.317569267745414e-06,
+      "loss": 0.8975,
+      "step": 11600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.316879076350124e-06,
+      "loss": 0.856,
+      "step": 11601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.316188878892436e-06,
+      "loss": 0.3876,
+      "step": 11602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.315498675385557e-06,
+      "loss": 0.9443,
+      "step": 11603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314808465842694e-06,
+      "loss": 0.8071,
+      "step": 11604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314118250277046e-06,
+      "loss": 0.8247,
+      "step": 11605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3134280287018204e-06,
+      "loss": 0.9126,
+      "step": 11606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312737801130226e-06,
+      "loss": 0.8999,
+      "step": 11607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312047567575463e-06,
+      "loss": 0.897,
+      "step": 11608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.311357328050737e-06,
+      "loss": 0.9072,
+      "step": 11609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.310667082569255e-06,
+      "loss": 0.8379,
+      "step": 11610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.309976831144224e-06,
+      "loss": 0.877,
+      "step": 11611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.309286573788846e-06,
+      "loss": 0.8687,
+      "step": 11612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.308596310516328e-06,
+      "loss": 0.8906,
+      "step": 11613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3079060413398786e-06,
+      "loss": 0.8506,
+      "step": 11614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3072157662727e-06,
+      "loss": 0.4285,
+      "step": 11615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.306525485327999e-06,
+      "loss": 0.9346,
+      "step": 11616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.305835198518984e-06,
+      "loss": 0.9253,
+      "step": 11617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.305144905858859e-06,
+      "loss": 0.8774,
+      "step": 11618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.304454607360831e-06,
+      "loss": 0.8833,
+      "step": 11619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303764303038106e-06,
+      "loss": 0.8779,
+      "step": 11620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303073992903893e-06,
+      "loss": 0.9072,
+      "step": 11621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.302383676971395e-06,
+      "loss": 0.9121,
+      "step": 11622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.301693355253821e-06,
+      "loss": 0.8633,
+      "step": 11623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.30100302776438e-06,
+      "loss": 0.894,
+      "step": 11624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.300312694516274e-06,
+      "loss": 0.7186,
+      "step": 11625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.299622355522714e-06,
+      "loss": 0.402,
+      "step": 11626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2989320107969065e-06,
+      "loss": 0.8569,
+      "step": 11627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2982416603520585e-06,
+      "loss": 0.9023,
+      "step": 11628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.297551304201377e-06,
+      "loss": 0.9082,
+      "step": 11629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2968609423580705e-06,
+      "loss": 0.8789,
+      "step": 11630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.296170574835347e-06,
+      "loss": 0.8384,
+      "step": 11631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.295480201646412e-06,
+      "loss": 0.8398,
+      "step": 11632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294789822804476e-06,
+      "loss": 0.8618,
+      "step": 11633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294099438322747e-06,
+      "loss": 0.8604,
+      "step": 11634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.29340904821443e-06,
+      "loss": 0.8281,
+      "step": 11635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.292718652492737e-06,
+      "loss": 0.9106,
+      "step": 11636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2920282511708775e-06,
+      "loss": 0.8774,
+      "step": 11637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.291337844262054e-06,
+      "loss": 0.8184,
+      "step": 11638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.290647431779479e-06,
+      "loss": 0.856,
+      "step": 11639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.289957013736362e-06,
+      "loss": 0.7983,
+      "step": 11640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2892665901459095e-06,
+      "loss": 0.8618,
+      "step": 11641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2885761610213316e-06,
+      "loss": 0.2843,
+      "step": 11642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.287885726375837e-06,
+      "loss": 0.876,
+      "step": 11643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.287195286222634e-06,
+      "loss": 0.8608,
+      "step": 11644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.286504840574932e-06,
+      "loss": 0.8643,
+      "step": 11645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.285814389445942e-06,
+      "loss": 0.8589,
+      "step": 11646
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.285123932848873e-06,
+      "loss": 0.8613,
+      "step": 11647
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.284433470796933e-06,
+      "loss": 0.8672,
+      "step": 11648
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.283743003303332e-06,
+      "loss": 0.8076,
+      "step": 11649
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.283052530381283e-06,
+      "loss": 0.7842,
+      "step": 11650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.28236205204399e-06,
+      "loss": 0.8848,
+      "step": 11651
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.281671568304666e-06,
+      "loss": 0.8818,
+      "step": 11652
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.280981079176524e-06,
+      "loss": 0.8501,
+      "step": 11653
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2802905846727684e-06,
+      "loss": 0.8984,
+      "step": 11654
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2796000848066135e-06,
+      "loss": 0.8799,
+      "step": 11655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.278909579591267e-06,
+      "loss": 0.8193,
+      "step": 11656
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.278219069039941e-06,
+      "loss": 0.8389,
+      "step": 11657
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.277528553165847e-06,
+      "loss": 0.9043,
+      "step": 11658
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2768380319821944e-06,
+      "loss": 0.856,
+      "step": 11659
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.276147505502192e-06,
+      "loss": 0.9102,
+      "step": 11660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.275456973739055e-06,
+      "loss": 0.8638,
+      "step": 11661
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2747664367059915e-06,
+      "loss": 0.8843,
+      "step": 11662
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.274075894416214e-06,
+      "loss": 0.8965,
+      "step": 11663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2733853468829295e-06,
+      "loss": 0.8228,
+      "step": 11664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.272694794119355e-06,
+      "loss": 0.7876,
+      "step": 11665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2720042361387e-06,
+      "loss": 0.8633,
+      "step": 11666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.271313672954174e-06,
+      "loss": 0.8813,
+      "step": 11667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.270623104578989e-06,
+      "loss": 0.8794,
+      "step": 11668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2699325310263585e-06,
+      "loss": 0.9141,
+      "step": 11669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.269241952309495e-06,
+      "loss": 0.8379,
+      "step": 11670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.268551368441607e-06,
+      "loss": 0.8755,
+      "step": 11671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.267860779435907e-06,
+      "loss": 0.894,
+      "step": 11672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.267170185305611e-06,
+      "loss": 0.9194,
+      "step": 11673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.266479586063926e-06,
+      "loss": 0.8613,
+      "step": 11674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2657889817240685e-06,
+      "loss": 0.8721,
+      "step": 11675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.265098372299248e-06,
+      "loss": 0.8901,
+      "step": 11676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.264407757802677e-06,
+      "loss": 0.8862,
+      "step": 11677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.263717138247569e-06,
+      "loss": 0.8657,
+      "step": 11678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2630265136471384e-06,
+      "loss": 0.8521,
+      "step": 11679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.262335884014596e-06,
+      "loss": 0.8735,
+      "step": 11680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.261645249363152e-06,
+      "loss": 0.8413,
+      "step": 11681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.260954609706024e-06,
+      "loss": 0.896,
+      "step": 11682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.260263965056425e-06,
+      "loss": 0.9639,
+      "step": 11683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.259573315427563e-06,
+      "loss": 0.8564,
+      "step": 11684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2588826608326545e-06,
+      "loss": 0.8364,
+      "step": 11685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258192001284916e-06,
+      "loss": 0.8252,
+      "step": 11686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.257501336797555e-06,
+      "loss": 0.8584,
+      "step": 11687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256810667383788e-06,
+      "loss": 0.8618,
+      "step": 11688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256119993056829e-06,
+      "loss": 0.8408,
+      "step": 11689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2554293138298895e-06,
+      "loss": 0.8809,
+      "step": 11690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.254738629716186e-06,
+      "loss": 0.3638,
+      "step": 11691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.25404794072893e-06,
+      "loss": 0.8555,
+      "step": 11692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2533572468813375e-06,
+      "loss": 0.9175,
+      "step": 11693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.252666548186622e-06,
+      "loss": 0.8682,
+      "step": 11694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251975844657996e-06,
+      "loss": 0.9136,
+      "step": 11695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251285136308677e-06,
+      "loss": 0.876,
+      "step": 11696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.250594423151875e-06,
+      "loss": 0.8862,
+      "step": 11697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2499037052008075e-06,
+      "loss": 0.8799,
+      "step": 11698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.24921298246869e-06,
+      "loss": 0.9087,
+      "step": 11699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.248522254968734e-06,
+      "loss": 0.8574,
+      "step": 11700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.247831522714154e-06,
+      "loss": 0.9155,
+      "step": 11701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.24714078571817e-06,
+      "loss": 0.9292,
+      "step": 11702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.246450043993991e-06,
+      "loss": 0.8652,
+      "step": 11703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.245759297554835e-06,
+      "loss": 0.873,
+      "step": 11704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.245068546413916e-06,
+      "loss": 0.8789,
+      "step": 11705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.24437779058445e-06,
+      "loss": 0.9102,
+      "step": 11706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.243687030079652e-06,
+      "loss": 0.8804,
+      "step": 11707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.242996264912736e-06,
+      "loss": 0.7715,
+      "step": 11708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.24230549509692e-06,
+      "loss": 0.8589,
+      "step": 11709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.241614720645416e-06,
+      "loss": 0.8428,
+      "step": 11710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240923941571443e-06,
+      "loss": 0.8267,
+      "step": 11711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240233157888216e-06,
+      "loss": 0.8853,
+      "step": 11712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.23954236960895e-06,
+      "loss": 0.8066,
+      "step": 11713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.238851576746858e-06,
+      "loss": 0.9058,
+      "step": 11714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.238160779315163e-06,
+      "loss": 0.8262,
+      "step": 11715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.237469977327075e-06,
+      "loss": 0.8564,
+      "step": 11716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236779170795813e-06,
+      "loss": 0.9155,
+      "step": 11717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236088359734591e-06,
+      "loss": 0.8711,
+      "step": 11718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.235397544156628e-06,
+      "loss": 0.8789,
+      "step": 11719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234706724075138e-06,
+      "loss": 0.9102,
+      "step": 11720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234015899503339e-06,
+      "loss": 0.8687,
+      "step": 11721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.233325070454447e-06,
+      "loss": 0.9038,
+      "step": 11722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2326342369416785e-06,
+      "loss": 0.8623,
+      "step": 11723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.231943398978251e-06,
+      "loss": 0.8799,
+      "step": 11724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2312525565773805e-06,
+      "loss": 0.897,
+      "step": 11725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.230561709752284e-06,
+      "loss": 0.8745,
+      "step": 11726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.229870858516178e-06,
+      "loss": 0.8682,
+      "step": 11727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.229180002882281e-06,
+      "loss": 0.8706,
+      "step": 11728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.228489142863812e-06,
+      "loss": 0.8857,
+      "step": 11729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.227798278473981e-06,
+      "loss": 0.936,
+      "step": 11730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.227107409726012e-06,
+      "loss": 0.8643,
+      "step": 11731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.226416536633122e-06,
+      "loss": 0.8374,
+      "step": 11732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.225725659208525e-06,
+      "loss": 0.8521,
+      "step": 11733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.22503477746544e-06,
+      "loss": 0.894,
+      "step": 11734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.224343891417086e-06,
+      "loss": 0.8345,
+      "step": 11735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.223653001076681e-06,
+      "loss": 0.8242,
+      "step": 11736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2229621064574395e-06,
+      "loss": 0.8599,
+      "step": 11737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.222271207572583e-06,
+      "loss": 0.8325,
+      "step": 11738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.221580304435327e-06,
+      "loss": 0.8164,
+      "step": 11739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2208893970588915e-06,
+      "loss": 0.8633,
+      "step": 11740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.220198485456494e-06,
+      "loss": 0.8291,
+      "step": 11741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.219507569641353e-06,
+      "loss": 0.4058,
+      "step": 11742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218816649626684e-06,
+      "loss": 0.8618,
+      "step": 11743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218125725425708e-06,
+      "loss": 0.876,
+      "step": 11744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.217434797051645e-06,
+      "loss": 0.8433,
+      "step": 11745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.21674386451771e-06,
+      "loss": 0.855,
+      "step": 11746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.216052927837123e-06,
+      "loss": 0.8354,
+      "step": 11747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.215361987023106e-06,
+      "loss": 0.8711,
+      "step": 11748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.214671042088872e-06,
+      "loss": 0.8818,
+      "step": 11749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.213980093047644e-06,
+      "loss": 0.8496,
+      "step": 11750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.213289139912638e-06,
+      "loss": 0.8027,
+      "step": 11751
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.212598182697076e-06,
+      "loss": 0.8809,
+      "step": 11752
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.211907221414176e-06,
+      "loss": 0.8662,
+      "step": 11753
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.211216256077156e-06,
+      "loss": 0.8779,
+      "step": 11754
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2105252866992375e-06,
+      "loss": 0.8662,
+      "step": 11755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.209834313293637e-06,
+      "loss": 0.8267,
+      "step": 11756
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.209143335873576e-06,
+      "loss": 0.8853,
+      "step": 11757
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.208452354452275e-06,
+      "loss": 0.9292,
+      "step": 11758
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.20776136904295e-06,
+      "loss": 0.8882,
+      "step": 11759
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.207070379658822e-06,
+      "loss": 0.873,
+      "step": 11760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.206379386313114e-06,
+      "loss": 0.8662,
+      "step": 11761
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.205688389019041e-06,
+      "loss": 0.9199,
+      "step": 11762
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2049973877898244e-06,
+      "loss": 0.8545,
+      "step": 11763
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.204306382638687e-06,
+      "loss": 0.8726,
+      "step": 11764
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.203615373578846e-06,
+      "loss": 0.9385,
+      "step": 11765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.202924360623521e-06,
+      "loss": 0.8794,
+      "step": 11766
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.202233343785934e-06,
+      "loss": 0.9048,
+      "step": 11767
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.201542323079304e-06,
+      "loss": 0.9009,
+      "step": 11768
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.200851298516852e-06,
+      "loss": 0.8213,
+      "step": 11769
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.200160270111798e-06,
+      "loss": 0.8711,
+      "step": 11770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.199469237877363e-06,
+      "loss": 0.8979,
+      "step": 11771
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.198778201826766e-06,
+      "loss": 0.8486,
+      "step": 11772
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19808716197323e-06,
+      "loss": 0.8765,
+      "step": 11773
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1973961183299745e-06,
+      "loss": 0.8105,
+      "step": 11774
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1967050709102205e-06,
+      "loss": 0.9082,
+      "step": 11775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.196014019727187e-06,
+      "loss": 0.8682,
+      "step": 11776
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.195322964794098e-06,
+      "loss": 0.8623,
+      "step": 11777
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194631906124172e-06,
+      "loss": 0.8945,
+      "step": 11778
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.193940843730631e-06,
+      "loss": 0.8555,
+      "step": 11779
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.193249777626694e-06,
+      "loss": 0.8262,
+      "step": 11780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192558707825588e-06,
+      "loss": 0.9126,
+      "step": 11781
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19186763434053e-06,
+      "loss": 0.8506,
+      "step": 11782
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1911765571847395e-06,
+      "loss": 0.8892,
+      "step": 11783
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190485476371442e-06,
+      "loss": 0.8413,
+      "step": 11784
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189794391913856e-06,
+      "loss": 0.8975,
+      "step": 11785
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189103303825205e-06,
+      "loss": 0.9038,
+      "step": 11786
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.188412212118708e-06,
+      "loss": 0.8604,
+      "step": 11787
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.187721116807591e-06,
+      "loss": 0.8579,
+      "step": 11788
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.187030017905073e-06,
+      "loss": 0.8677,
+      "step": 11789
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.186338915424375e-06,
+      "loss": 0.8823,
+      "step": 11790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185647809378721e-06,
+      "loss": 0.9233,
+      "step": 11791
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1849566997813295e-06,
+      "loss": 0.8838,
+      "step": 11792
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.184265586645426e-06,
+      "loss": 0.8804,
+      "step": 11793
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.183574469984234e-06,
+      "loss": 0.812,
+      "step": 11794
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.18288334981097e-06,
+      "loss": 0.7876,
+      "step": 11795
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182192226138859e-06,
+      "loss": 0.7891,
+      "step": 11796
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1815010989811265e-06,
+      "loss": 0.8911,
+      "step": 11797
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.18080996835099e-06,
+      "loss": 0.8315,
+      "step": 11798
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.180118834261675e-06,
+      "loss": 0.9482,
+      "step": 11799
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1794276967264015e-06,
+      "loss": 0.8926,
+      "step": 11800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.178736555758395e-06,
+      "loss": 0.875,
+      "step": 11801
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.178045411370874e-06,
+      "loss": 0.853,
+      "step": 11802
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.177354263577065e-06,
+      "loss": 0.8599,
+      "step": 11803
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176663112390191e-06,
+      "loss": 0.7905,
+      "step": 11804
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175971957823471e-06,
+      "loss": 0.8496,
+      "step": 11805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175280799890132e-06,
+      "loss": 0.8604,
+      "step": 11806
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.174589638603395e-06,
+      "loss": 0.8569,
+      "step": 11807
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.173898473976483e-06,
+      "loss": 0.8564,
+      "step": 11808
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.173207306022617e-06,
+      "loss": 0.8291,
+      "step": 11809
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172516134755024e-06,
+      "loss": 0.8208,
+      "step": 11810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.171824960186927e-06,
+      "loss": 0.3737,
+      "step": 11811
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.171133782331546e-06,
+      "loss": 0.8628,
+      "step": 11812
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.170442601202106e-06,
+      "loss": 0.9312,
+      "step": 11813
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.16975141681183e-06,
+      "loss": 0.8999,
+      "step": 11814
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169060229173942e-06,
+      "loss": 0.8716,
+      "step": 11815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.168369038301666e-06,
+      "loss": 0.8701,
+      "step": 11816
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1676778442082255e-06,
+      "loss": 0.8354,
+      "step": 11817
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.166986646906843e-06,
+      "loss": 0.8271,
+      "step": 11818
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.166295446410743e-06,
+      "loss": 0.8921,
+      "step": 11819
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1656042427331494e-06,
+      "loss": 0.8647,
+      "step": 11820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164913035887284e-06,
+      "loss": 0.8613,
+      "step": 11821
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164221825886373e-06,
+      "loss": 0.8838,
+      "step": 11822
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.163530612743641e-06,
+      "loss": 0.8486,
+      "step": 11823
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.16283939647231e-06,
+      "loss": 0.8975,
+      "step": 11824
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.162148177085604e-06,
+      "loss": 0.895,
+      "step": 11825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.16145695459675e-06,
+      "loss": 0.8247,
+      "step": 11826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.160765729018968e-06,
+      "loss": 0.9048,
+      "step": 11827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.160074500365485e-06,
+      "loss": 0.877,
+      "step": 11828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1593832686495235e-06,
+      "loss": 0.8521,
+      "step": 11829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1586920338843105e-06,
+      "loss": 0.9082,
+      "step": 11830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.158000796083068e-06,
+      "loss": 0.9521,
+      "step": 11831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.157309555259022e-06,
+      "loss": 0.915,
+      "step": 11832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.156618311425395e-06,
+      "loss": 0.9302,
+      "step": 11833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.155927064595413e-06,
+      "loss": 0.8257,
+      "step": 11834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.155235814782301e-06,
+      "loss": 0.9141,
+      "step": 11835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.154544561999285e-06,
+      "loss": 0.9229,
+      "step": 11836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153853306259586e-06,
+      "loss": 0.8154,
+      "step": 11837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.15316204757643e-06,
+      "loss": 0.8545,
+      "step": 11838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.152470785963044e-06,
+      "loss": 0.9111,
+      "step": 11839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.151779521432653e-06,
+      "loss": 0.8677,
+      "step": 11840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1510882539984775e-06,
+      "loss": 0.8726,
+      "step": 11841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.150396983673747e-06,
+      "loss": 0.8945,
+      "step": 11842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.149705710471686e-06,
+      "loss": 0.8516,
+      "step": 11843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.149014434405519e-06,
+      "loss": 0.8882,
+      "step": 11844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.148323155488471e-06,
+      "loss": 0.832,
+      "step": 11845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.147631873733766e-06,
+      "loss": 0.8735,
+      "step": 11846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1469405891546305e-06,
+      "loss": 0.8306,
+      "step": 11847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.146249301764291e-06,
+      "loss": 0.8774,
+      "step": 11848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.145558011575972e-06,
+      "loss": 0.9116,
+      "step": 11849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144866718602898e-06,
+      "loss": 0.8555,
+      "step": 11850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144175422858297e-06,
+      "loss": 0.915,
+      "step": 11851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.14348412435539e-06,
+      "loss": 0.9194,
+      "step": 11852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.142792823107409e-06,
+      "loss": 0.8486,
+      "step": 11853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.142101519127574e-06,
+      "loss": 0.8486,
+      "step": 11854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1414102124291135e-06,
+      "loss": 0.8413,
+      "step": 11855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.140718903025254e-06,
+      "loss": 0.8613,
+      "step": 11856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.140027590929218e-06,
+      "loss": 0.9072,
+      "step": 11857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1393362761542345e-06,
+      "loss": 0.8403,
+      "step": 11858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1386449587135304e-06,
+      "loss": 0.8921,
+      "step": 11859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1379536386203275e-06,
+      "loss": 0.8608,
+      "step": 11860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1372623158878546e-06,
+      "loss": 0.8765,
+      "step": 11861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.136570990529337e-06,
+      "loss": 0.8359,
+      "step": 11862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.135879662558001e-06,
+      "loss": 0.9204,
+      "step": 11863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1351883319870734e-06,
+      "loss": 0.8477,
+      "step": 11864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.13449699882978e-06,
+      "loss": 0.8784,
+      "step": 11865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.133805663099347e-06,
+      "loss": 0.8657,
+      "step": 11866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.133114324809001e-06,
+      "loss": 0.875,
+      "step": 11867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.132422983971967e-06,
+      "loss": 0.8467,
+      "step": 11868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131731640601476e-06,
+      "loss": 0.916,
+      "step": 11869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131040294710747e-06,
+      "loss": 0.8198,
+      "step": 11870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.130348946313013e-06,
+      "loss": 0.8911,
+      "step": 11871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.129657595421499e-06,
+      "loss": 0.8472,
+      "step": 11872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.128966242049429e-06,
+      "loss": 0.8901,
+      "step": 11873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.128274886210031e-06,
+      "loss": 0.8315,
+      "step": 11874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.127583527916535e-06,
+      "loss": 0.7896,
+      "step": 11875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.126892167182164e-06,
+      "loss": 0.8369,
+      "step": 11876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.126200804020145e-06,
+      "loss": 0.8477,
+      "step": 11877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1255094384437055e-06,
+      "loss": 0.896,
+      "step": 11878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.124818070466073e-06,
+      "loss": 0.8789,
+      "step": 11879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1241267001004755e-06,
+      "loss": 0.8496,
+      "step": 11880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.123435327360138e-06,
+      "loss": 0.8306,
+      "step": 11881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122743952258288e-06,
+      "loss": 0.8608,
+      "step": 11882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122052574808151e-06,
+      "loss": 0.8101,
+      "step": 11883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.121361195022958e-06,
+      "loss": 0.8945,
+      "step": 11884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.120669812915935e-06,
+      "loss": 0.8696,
+      "step": 11885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.119978428500306e-06,
+      "loss": 0.8652,
+      "step": 11886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1192870417893e-06,
+      "loss": 0.9419,
+      "step": 11887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118595652796148e-06,
+      "loss": 0.8374,
+      "step": 11888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.117904261534073e-06,
+      "loss": 0.8677,
+      "step": 11889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.117212868016303e-06,
+      "loss": 0.895,
+      "step": 11890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.116521472256065e-06,
+      "loss": 0.8833,
+      "step": 11891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.115830074266592e-06,
+      "loss": 0.8345,
+      "step": 11892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.115138674061103e-06,
+      "loss": 0.9019,
+      "step": 11893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.114447271652832e-06,
+      "loss": 0.8789,
+      "step": 11894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.113755867055002e-06,
+      "loss": 0.8633,
+      "step": 11895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.113064460280844e-06,
+      "loss": 0.8481,
+      "step": 11896
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1123730513435855e-06,
+      "loss": 0.9082,
+      "step": 11897
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.111681640256453e-06,
+      "loss": 0.8428,
+      "step": 11898
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.110990227032675e-06,
+      "loss": 0.8516,
+      "step": 11899
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.110298811685479e-06,
+      "loss": 0.9321,
+      "step": 11900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.109607394228092e-06,
+      "loss": 0.8794,
+      "step": 11901
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.108915974673745e-06,
+      "loss": 0.9004,
+      "step": 11902
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.108224553035661e-06,
+      "loss": 0.3683,
+      "step": 11903
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1075331293270725e-06,
+      "loss": 0.8413,
+      "step": 11904
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.106841703561206e-06,
+      "loss": 0.8037,
+      "step": 11905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.106150275751289e-06,
+      "loss": 0.8984,
+      "step": 11906
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.105458845910548e-06,
+      "loss": 0.8247,
+      "step": 11907
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.104767414052217e-06,
+      "loss": 0.8813,
+      "step": 11908
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1040759801895165e-06,
+      "loss": 0.8394,
+      "step": 11909
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.103384544335681e-06,
+      "loss": 0.8999,
+      "step": 11910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102693106503935e-06,
+      "loss": 0.8423,
+      "step": 11911
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1020016667075075e-06,
+      "loss": 0.8979,
+      "step": 11912
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1013102249596295e-06,
+      "loss": 0.8569,
+      "step": 11913
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.100618781273525e-06,
+      "loss": 0.8271,
+      "step": 11914
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.099927335662427e-06,
+      "loss": 0.8726,
+      "step": 11915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.099235888139559e-06,
+      "loss": 0.8345,
+      "step": 11916
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0985444387181535e-06,
+      "loss": 0.8657,
+      "step": 11917
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.097852987411438e-06,
+      "loss": 0.7876,
+      "step": 11918
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.097161534232641e-06,
+      "loss": 0.8713,
+      "step": 11919
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0964700791949885e-06,
+      "loss": 0.8135,
+      "step": 11920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.095778622311713e-06,
+      "loss": 0.8916,
+      "step": 11921
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.095087163596042e-06,
+      "loss": 0.318,
+      "step": 11922
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0943957030612024e-06,
+      "loss": 0.8154,
+      "step": 11923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093704240720425e-06,
+      "loss": 0.8296,
+      "step": 11924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093012776586938e-06,
+      "loss": 0.8813,
+      "step": 11925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.092321310673969e-06,
+      "loss": 0.8457,
+      "step": 11926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091629842994749e-06,
+      "loss": 0.873,
+      "step": 11927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.090938373562505e-06,
+      "loss": 0.7959,
+      "step": 11928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.090246902390466e-06,
+      "loss": 0.8491,
+      "step": 11929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089555429491862e-06,
+      "loss": 0.8257,
+      "step": 11930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.088863954879923e-06,
+      "loss": 0.9072,
+      "step": 11931
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.088172478567874e-06,
+      "loss": 0.896,
+      "step": 11932
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.087481000568949e-06,
+      "loss": 0.8535,
+      "step": 11933
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0867895208963746e-06,
+      "loss": 0.8311,
+      "step": 11934
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.086098039563378e-06,
+      "loss": 0.8955,
+      "step": 11935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08540655658319e-06,
+      "loss": 0.8628,
+      "step": 11936
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084715071969042e-06,
+      "loss": 0.8442,
+      "step": 11937
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08402358573416e-06,
+      "loss": 0.8896,
+      "step": 11938
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.083332097891776e-06,
+      "loss": 0.8955,
+      "step": 11939
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.082640608455115e-06,
+      "loss": 0.8447,
+      "step": 11940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08194911743741e-06,
+      "loss": 0.8745,
+      "step": 11941
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.081257624851889e-06,
+      "loss": 0.8291,
+      "step": 11942
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080566130711783e-06,
+      "loss": 0.2806,
+      "step": 11943
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.079874635030319e-06,
+      "loss": 0.8618,
+      "step": 11944
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.079183137820727e-06,
+      "loss": 0.8408,
+      "step": 11945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.078491639096237e-06,
+      "loss": 0.856,
+      "step": 11946
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077800138870079e-06,
+      "loss": 0.8555,
+      "step": 11947
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077108637155482e-06,
+      "loss": 0.9092,
+      "step": 11948
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.076417133965673e-06,
+      "loss": 0.9087,
+      "step": 11949
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.075725629313886e-06,
+      "loss": 0.8784,
+      "step": 11950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.075034123213348e-06,
+      "loss": 0.874,
+      "step": 11951
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.074342615677289e-06,
+      "loss": 0.8379,
+      "step": 11952
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.073651106718937e-06,
+      "loss": 0.8501,
+      "step": 11953
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072959596351525e-06,
+      "loss": 0.8398,
+      "step": 11954
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072268084588281e-06,
+      "loss": 0.897,
+      "step": 11955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.071576571442434e-06,
+      "loss": 0.8262,
+      "step": 11956
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070885056927214e-06,
+      "loss": 0.8804,
+      "step": 11957
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070193541055852e-06,
+      "loss": 0.8911,
+      "step": 11958
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.069502023841576e-06,
+      "loss": 0.8501,
+      "step": 11959
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068810505297618e-06,
+      "loss": 0.856,
+      "step": 11960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068118985437206e-06,
+      "loss": 0.8223,
+      "step": 11961
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06742746427357e-06,
+      "loss": 0.8369,
+      "step": 11962
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0667359418199406e-06,
+      "loss": 0.897,
+      "step": 11963
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.066044418089549e-06,
+      "loss": 0.8623,
+      "step": 11964
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.065352893095621e-06,
+      "loss": 0.8608,
+      "step": 11965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06466136685139e-06,
+      "loss": 0.8794,
+      "step": 11966
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063969839370086e-06,
+      "loss": 0.876,
+      "step": 11967
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063278310664938e-06,
+      "loss": 0.8472,
+      "step": 11968
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.062586780749176e-06,
+      "loss": 0.8657,
+      "step": 11969
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.061895249636031e-06,
+      "loss": 0.8745,
+      "step": 11970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.061203717338732e-06,
+      "loss": 0.8857,
+      "step": 11971
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.060512183870508e-06,
+      "loss": 0.8208,
+      "step": 11972
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059820649244592e-06,
+      "loss": 0.8525,
+      "step": 11973
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059129113474212e-06,
+      "loss": 0.8804,
+      "step": 11974
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.058437576572599e-06,
+      "loss": 0.874,
+      "step": 11975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057746038552983e-06,
+      "loss": 0.8345,
+      "step": 11976
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057054499428597e-06,
+      "loss": 0.8594,
+      "step": 11977
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.056362959212664e-06,
+      "loss": 0.8306,
+      "step": 11978
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.055671417918421e-06,
+      "loss": 0.8843,
+      "step": 11979
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.054979875559096e-06,
+      "loss": 0.8706,
+      "step": 11980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0542883321479195e-06,
+      "loss": 0.853,
+      "step": 11981
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.053596787698121e-06,
+      "loss": 0.8521,
+      "step": 11982
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052905242222933e-06,
+      "loss": 0.897,
+      "step": 11983
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052213695735583e-06,
+      "loss": 0.9067,
+      "step": 11984
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.051522148249302e-06,
+      "loss": 0.8809,
+      "step": 11985
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050830599777322e-06,
+      "loss": 0.8477,
+      "step": 11986
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050139050332873e-06,
+      "loss": 0.8652,
+      "step": 11987
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.049447499929184e-06,
+      "loss": 0.856,
+      "step": 11988
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048755948579488e-06,
+      "loss": 0.9033,
+      "step": 11989
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048064396297012e-06,
+      "loss": 0.9053,
+      "step": 11990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04737284309499e-06,
+      "loss": 0.8306,
+      "step": 11991
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.046681288986649e-06,
+      "loss": 0.8784,
+      "step": 11992
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.045989733985223e-06,
+      "loss": 0.8843,
+      "step": 11993
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04529817810394e-06,
+      "loss": 0.8491,
+      "step": 11994
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.044606621356032e-06,
+      "loss": 0.9062,
+      "step": 11995
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0439150637547294e-06,
+      "loss": 0.3068,
+      "step": 11996
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.043223505313263e-06,
+      "loss": 0.8511,
+      "step": 11997
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04253194604486e-06,
+      "loss": 0.8765,
+      "step": 11998
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.041840385962756e-06,
+      "loss": 0.8755,
+      "step": 11999
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04114882508018e-06,
+      "loss": 0.8779,
+      "step": 12000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.040457263410361e-06,
+      "loss": 0.9102,
+      "step": 12001
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.039765700966529e-06,
+      "loss": 0.8281,
+      "step": 12002
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.03907413776192e-06,
+      "loss": 0.3817,
+      "step": 12003
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.038382573809759e-06,
+      "loss": 0.79,
+      "step": 12004
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.03769100912328e-06,
+      "loss": 0.8833,
+      "step": 12005
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.036999443715712e-06,
+      "loss": 0.8491,
+      "step": 12006
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.036307877600286e-06,
+      "loss": 0.793,
+      "step": 12007
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035616310790234e-06,
+      "loss": 0.8496,
+      "step": 12008
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.034924743298786e-06,
+      "loss": 0.9663,
+      "step": 12009
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.034233175139172e-06,
+      "loss": 0.833,
+      "step": 12010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0335416063246225e-06,
+      "loss": 0.8647,
+      "step": 12011
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032850036868371e-06,
+      "loss": 0.8452,
+      "step": 12012
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032158466783647e-06,
+      "loss": 0.7715,
+      "step": 12013
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.031466896083679e-06,
+      "loss": 0.875,
+      "step": 12014
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030775324781698e-06,
+      "loss": 0.8691,
+      "step": 12015
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0300837528909404e-06,
+      "loss": 0.8501,
+      "step": 12016
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.029392180424632e-06,
+      "loss": 0.856,
+      "step": 12017
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028700607396004e-06,
+      "loss": 0.7831,
+      "step": 12018
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028009033818289e-06,
+      "loss": 0.8848,
+      "step": 12019
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.027317459704716e-06,
+      "loss": 0.8535,
+      "step": 12020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026625885068518e-06,
+      "loss": 0.854,
+      "step": 12021
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.025934309922924e-06,
+      "loss": 0.8828,
+      "step": 12022
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.025242734281166e-06,
+      "loss": 0.7949,
+      "step": 12023
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.024551158156475e-06,
+      "loss": 0.8511,
+      "step": 12024
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.023859581562082e-06,
+      "loss": 0.7817,
+      "step": 12025
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0231680045112174e-06,
+      "loss": 0.8989,
+      "step": 12026
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.022476427017111e-06,
+      "loss": 0.8604,
+      "step": 12027
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.021784849092997e-06,
+      "loss": 0.8848,
+      "step": 12028
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.021093270752104e-06,
+      "loss": 0.8311,
+      "step": 12029
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.020401692007664e-06,
+      "loss": 0.8911,
+      "step": 12030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.019710112872905e-06,
+      "loss": 0.8477,
+      "step": 12031
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.019018533361064e-06,
+      "loss": 0.8823,
+      "step": 12032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0183269534853665e-06,
+      "loss": 0.8813,
+      "step": 12033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.017635373259046e-06,
+      "loss": 0.8457,
+      "step": 12034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0169437926953325e-06,
+      "loss": 0.8247,
+      "step": 12035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.016252211807457e-06,
+      "loss": 0.3811,
+      "step": 12036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.015560630608652e-06,
+      "loss": 0.9336,
+      "step": 12037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.014869049112149e-06,
+      "loss": 0.873,
+      "step": 12038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0141774673311745e-06,
+      "loss": 0.8589,
+      "step": 12039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.013485885278965e-06,
+      "loss": 0.9062,
+      "step": 12040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012794302968749e-06,
+      "loss": 0.8765,
+      "step": 12041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012102720413758e-06,
+      "loss": 0.916,
+      "step": 12042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.011411137627222e-06,
+      "loss": 0.8857,
+      "step": 12043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.010719554622373e-06,
+      "loss": 0.8643,
+      "step": 12044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0100279714124435e-06,
+      "loss": 0.825,
+      "step": 12045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0093363880106615e-06,
+      "loss": 0.7739,
+      "step": 12046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008644804430261e-06,
+      "loss": 0.8604,
+      "step": 12047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.007953220684471e-06,
+      "loss": 0.812,
+      "step": 12048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.007261636786522e-06,
+      "loss": 0.8682,
+      "step": 12049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.006570052749648e-06,
+      "loss": 0.8179,
+      "step": 12050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0058784685870786e-06,
+      "loss": 0.8521,
+      "step": 12051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.005186884312045e-06,
+      "loss": 0.8364,
+      "step": 12052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.004495299937776e-06,
+      "loss": 0.8726,
+      "step": 12053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.003803715477507e-06,
+      "loss": 0.8525,
+      "step": 12054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.003112130944467e-06,
+      "loss": 0.8198,
+      "step": 12055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.002420546351886e-06,
+      "loss": 0.8408,
+      "step": 12056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.001728961712996e-06,
+      "loss": 0.9111,
+      "step": 12057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.001037377041031e-06,
+      "loss": 0.9312,
+      "step": 12058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.000345792349216e-06,
+      "loss": 0.9097,
+      "step": 12059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.999654207650786e-06,
+      "loss": 0.8589,
+      "step": 12060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.998962622958973e-06,
+      "loss": 0.8999,
+      "step": 12061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9982710382870055e-06,
+      "loss": 0.8154,
+      "step": 12062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.997579453648115e-06,
+      "loss": 0.8701,
+      "step": 12063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.996887869055535e-06,
+      "loss": 0.9248,
+      "step": 12064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.996196284522494e-06,
+      "loss": 0.8525,
+      "step": 12065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9955047000622234e-06,
+      "loss": 0.834,
+      "step": 12066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.994813115687958e-06,
+      "loss": 0.8887,
+      "step": 12067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.994121531412923e-06,
+      "loss": 0.8779,
+      "step": 12068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993429947250355e-06,
+      "loss": 0.9199,
+      "step": 12069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.992738363213479e-06,
+      "loss": 0.8594,
+      "step": 12070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.992046779315531e-06,
+      "loss": 0.8579,
+      "step": 12071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.991355195569742e-06,
+      "loss": 0.7852,
+      "step": 12072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.99066361198934e-06,
+      "loss": 0.8882,
+      "step": 12073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.989972028587558e-06,
+      "loss": 0.7881,
+      "step": 12074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.989280445377628e-06,
+      "loss": 0.8882,
+      "step": 12075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.988588862372779e-06,
+      "loss": 0.833,
+      "step": 12076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.987897279586244e-06,
+      "loss": 0.8999,
+      "step": 12077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.987205697031253e-06,
+      "loss": 0.3726,
+      "step": 12078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.986514114721035e-06,
+      "loss": 0.8633,
+      "step": 12079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.985822532668826e-06,
+      "loss": 0.8804,
+      "step": 12080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.985130950887853e-06,
+      "loss": 0.3041,
+      "step": 12081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98443936939135e-06,
+      "loss": 0.9443,
+      "step": 12082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983747788192544e-06,
+      "loss": 0.9072,
+      "step": 12083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983056207304669e-06,
+      "loss": 0.8315,
+      "step": 12084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.982364626740957e-06,
+      "loss": 0.8716,
+      "step": 12085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.981673046514634e-06,
+      "loss": 0.8799,
+      "step": 12086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.980981466638937e-06,
+      "loss": 0.8926,
+      "step": 12087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.980289887127096e-06,
+      "loss": 0.9136,
+      "step": 12088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.979598307992338e-06,
+      "loss": 0.3555,
+      "step": 12089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978906729247898e-06,
+      "loss": 0.7505,
+      "step": 12090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978215150907005e-06,
+      "loss": 0.856,
+      "step": 12091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.977523572982888e-06,
+      "loss": 0.8628,
+      "step": 12092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.976831995488784e-06,
+      "loss": 0.8179,
+      "step": 12093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97614041843792e-06,
+      "loss": 0.2866,
+      "step": 12094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975448841843526e-06,
+      "loss": 0.8882,
+      "step": 12095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.974757265718835e-06,
+      "loss": 0.8623,
+      "step": 12096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.974065690077078e-06,
+      "loss": 0.856,
+      "step": 12097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.973374114931485e-06,
+      "loss": 0.8188,
+      "step": 12098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9726825402952855e-06,
+      "loss": 0.8989,
+      "step": 12099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971990966181713e-06,
+      "loss": 0.8374,
+      "step": 12100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9712993926039985e-06,
+      "loss": 0.8657,
+      "step": 12101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97060781957537e-06,
+      "loss": 0.3362,
+      "step": 12102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969916247109061e-06,
+      "loss": 0.8882,
+      "step": 12103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969224675218303e-06,
+      "loss": 0.8486,
+      "step": 12104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.968533103916323e-06,
+      "loss": 0.8569,
+      "step": 12105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.967841533216357e-06,
+      "loss": 0.8945,
+      "step": 12106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.967149963131631e-06,
+      "loss": 0.3566,
+      "step": 12107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9664583936753774e-06,
+      "loss": 0.875,
+      "step": 12108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96576682486083e-06,
+      "loss": 0.8418,
+      "step": 12109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965075256701216e-06,
+      "loss": 0.8198,
+      "step": 12110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9643836892097684e-06,
+      "loss": 0.8794,
+      "step": 12111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9636921223997144e-06,
+      "loss": 0.856,
+      "step": 12112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9630005562842885e-06,
+      "loss": 0.8062,
+      "step": 12113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.962308990876723e-06,
+      "loss": 0.8638,
+      "step": 12114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9616174261902414e-06,
+      "loss": 0.8892,
+      "step": 12115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96092586223808e-06,
+      "loss": 0.8755,
+      "step": 12116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.960234299033472e-06,
+      "loss": 0.8198,
+      "step": 12117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9595427365896405e-06,
+      "loss": 0.8887,
+      "step": 12118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958851174919823e-06,
+      "loss": 0.8242,
+      "step": 12119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958159614037245e-06,
+      "loss": 0.8403,
+      "step": 12120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.957468053955141e-06,
+      "loss": 0.8818,
+      "step": 12121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.956776494686741e-06,
+      "loss": 0.8877,
+      "step": 12122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.956084936245271e-06,
+      "loss": 0.8716,
+      "step": 12123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.955393378643968e-06,
+      "loss": 0.9048,
+      "step": 12124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.954701821896062e-06,
+      "loss": 0.9023,
+      "step": 12125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9540102660147774e-06,
+      "loss": 0.9214,
+      "step": 12126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.953318711013352e-06,
+      "loss": 0.7803,
+      "step": 12127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.952627156905012e-06,
+      "loss": 0.9243,
+      "step": 12128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951935603702988e-06,
+      "loss": 0.8247,
+      "step": 12129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951244051420514e-06,
+      "loss": 0.8096,
+      "step": 12130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.950552500070817e-06,
+      "loss": 0.8711,
+      "step": 12131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949860949667128e-06,
+      "loss": 0.8608,
+      "step": 12132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949169400222679e-06,
+      "loss": 0.8901,
+      "step": 12133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9484778517506986e-06,
+      "loss": 0.8369,
+      "step": 12134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94778630426442e-06,
+      "loss": 0.9136,
+      "step": 12135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.947094757777069e-06,
+      "loss": 0.8633,
+      "step": 12136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94640321230188e-06,
+      "loss": 0.8315,
+      "step": 12137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.945711667852083e-06,
+      "loss": 0.8662,
+      "step": 12138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.945020124440905e-06,
+      "loss": 0.8342,
+      "step": 12139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.944328582081579e-06,
+      "loss": 0.8442,
+      "step": 12140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9436370407873375e-06,
+      "loss": 0.8843,
+      "step": 12141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942945500571406e-06,
+      "loss": 0.8521,
+      "step": 12142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942253961447018e-06,
+      "loss": 0.8613,
+      "step": 12143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.941562423427403e-06,
+      "loss": 0.8374,
+      "step": 12144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.940870886525788e-06,
+      "loss": 0.8965,
+      "step": 12145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94017935075541e-06,
+      "loss": 0.917,
+      "step": 12146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.939487816129493e-06,
+      "loss": 0.8975,
+      "step": 12147
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938796282661271e-06,
+      "loss": 0.9463,
+      "step": 12148
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938104750363971e-06,
+      "loss": 0.8413,
+      "step": 12149
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.937413219250825e-06,
+      "loss": 0.8276,
+      "step": 12150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936721689335064e-06,
+      "loss": 0.9087,
+      "step": 12151
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936030160629915e-06,
+      "loss": 0.8477,
+      "step": 12152
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93533863314861e-06,
+      "loss": 0.8828,
+      "step": 12153
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93464710690438e-06,
+      "loss": 0.8477,
+      "step": 12154
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.933955581910453e-06,
+      "loss": 0.8159,
+      "step": 12155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93326405818006e-06,
+      "loss": 0.8481,
+      "step": 12156
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9325725357264316e-06,
+      "loss": 0.8633,
+      "step": 12157
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.931881014562795e-06,
+      "loss": 0.9014,
+      "step": 12158
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.931189494702384e-06,
+      "loss": 0.8921,
+      "step": 12159
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9304979761584256e-06,
+      "loss": 0.8701,
+      "step": 12160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929806458944149e-06,
+      "loss": 0.9521,
+      "step": 12161
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929114943072788e-06,
+      "loss": 0.8037,
+      "step": 12162
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.928423428557568e-06,
+      "loss": 0.7793,
+      "step": 12163
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927731915411722e-06,
+      "loss": 0.3304,
+      "step": 12164
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927040403648476e-06,
+      "loss": 0.8545,
+      "step": 12165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.926348893281064e-06,
+      "loss": 0.8843,
+      "step": 12166
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.925657384322714e-06,
+      "loss": 0.8716,
+      "step": 12167
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.924965876786653e-06,
+      "loss": 0.915,
+      "step": 12168
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.924274370686115e-06,
+      "loss": 0.9067,
+      "step": 12169
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.923582866034329e-06,
+      "loss": 0.8804,
+      "step": 12170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9228913628445195e-06,
+      "loss": 0.8931,
+      "step": 12171
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9221998611299225e-06,
+      "loss": 0.8335,
+      "step": 12172
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.921508360903764e-06,
+      "loss": 0.8833,
+      "step": 12173
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.920816862179273e-06,
+      "loss": 0.7939,
+      "step": 12174
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.920125364969683e-06,
+      "loss": 0.8545,
+      "step": 12175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.919433869288219e-06,
+      "loss": 0.8213,
+      "step": 12176
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.918742375148113e-06,
+      "loss": 0.8691,
+      "step": 12177
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.918050882562591e-06,
+      "loss": 0.8291,
+      "step": 12178
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.917359391544886e-06,
+      "loss": 0.9077,
+      "step": 12179
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.916667902108228e-06,
+      "loss": 0.8481,
+      "step": 12180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9159764142658415e-06,
+      "loss": 0.9229,
+      "step": 12181
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.915284928030959e-06,
+      "loss": 0.8584,
+      "step": 12182
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.914593443416812e-06,
+      "loss": 0.894,
+      "step": 12183
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.913901960436623e-06,
+      "loss": 0.9331,
+      "step": 12184
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.913210479103628e-06,
+      "loss": 0.8691,
+      "step": 12185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.912518999431053e-06,
+      "loss": 0.8105,
+      "step": 12186
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911827521432125e-06,
+      "loss": 0.9463,
+      "step": 12187
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911136045120079e-06,
+      "loss": 0.8306,
+      "step": 12188
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.910444570508139e-06,
+      "loss": 0.8345,
+      "step": 12189
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.909753097609535e-06,
+      "loss": 0.8691,
+      "step": 12190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.909061626437497e-06,
+      "loss": 0.8843,
+      "step": 12191
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908370157005253e-06,
+      "loss": 0.8975,
+      "step": 12192
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.907678689326034e-06,
+      "loss": 0.8506,
+      "step": 12193
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906987223413064e-06,
+      "loss": 0.8848,
+      "step": 12194
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906295759279577e-06,
+      "loss": 0.8535,
+      "step": 12195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9056042969388e-06,
+      "loss": 0.9272,
+      "step": 12196
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.90491283640396e-06,
+      "loss": 0.8706,
+      "step": 12197
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.904221377688288e-06,
+      "loss": 0.9028,
+      "step": 12198
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903529920805013e-06,
+      "loss": 0.8599,
+      "step": 12199
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.902838465767361e-06,
+      "loss": 0.8711,
+      "step": 12200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.902147012588564e-06,
+      "loss": 0.8706,
+      "step": 12201
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.901455561281848e-06,
+      "loss": 0.8521,
+      "step": 12202
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.900764111860441e-06,
+      "loss": 0.8125,
+      "step": 12203
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9000726643375755e-06,
+      "loss": 0.8828,
+      "step": 12204
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.899381218726476e-06,
+      "loss": 0.8154,
+      "step": 12205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.898689775040371e-06,
+      "loss": 0.8794,
+      "step": 12206
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897998333292493e-06,
+      "loss": 0.856,
+      "step": 12207
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8973068934960666e-06,
+      "loss": 0.9258,
+      "step": 12208
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.896615455664322e-06,
+      "loss": 0.8394,
+      "step": 12209
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.895924019810484e-06,
+      "loss": 0.9146,
+      "step": 12210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.895232585947785e-06,
+      "loss": 0.8477,
+      "step": 12211
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.894541154089453e-06,
+      "loss": 0.9087,
+      "step": 12212
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.893849724248713e-06,
+      "loss": 0.8384,
+      "step": 12213
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8931582964387964e-06,
+      "loss": 0.8457,
+      "step": 12214
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892466870672929e-06,
+      "loss": 0.8765,
+      "step": 12215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.891775446964339e-06,
+      "loss": 0.8291,
+      "step": 12216
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.891084025326257e-06,
+      "loss": 0.8057,
+      "step": 12217
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8903926057719096e-06,
+      "loss": 0.8232,
+      "step": 12218
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8897011883145216e-06,
+      "loss": 0.9248,
+      "step": 12219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8890097729673264e-06,
+      "loss": 0.8765,
+      "step": 12220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.888318359743549e-06,
+      "loss": 0.8311,
+      "step": 12221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.887626948656417e-06,
+      "loss": 0.8228,
+      "step": 12222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.886935539719157e-06,
+      "loss": 0.3029,
+      "step": 12223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.886244132944998e-06,
+      "loss": 0.8828,
+      "step": 12224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.885552728347172e-06,
+      "loss": 0.8989,
+      "step": 12225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8848613259388986e-06,
+      "loss": 0.8662,
+      "step": 12226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.884169925733409e-06,
+      "loss": 0.8828,
+      "step": 12227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.883478527743936e-06,
+      "loss": 0.833,
+      "step": 12228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882787131983698e-06,
+      "loss": 0.8794,
+      "step": 12229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88209573846593e-06,
+      "loss": 0.8579,
+      "step": 12230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881404347203853e-06,
+      "loss": 0.8691,
+      "step": 12231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880712958210699e-06,
+      "loss": 0.9448,
+      "step": 12232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880021571499696e-06,
+      "loss": 0.8887,
+      "step": 12233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.879330187084067e-06,
+      "loss": 0.9097,
+      "step": 12234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.878638804977042e-06,
+      "loss": 0.832,
+      "step": 12235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.87794742519185e-06,
+      "loss": 0.7812,
+      "step": 12236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.877256047741713e-06,
+      "loss": 0.7959,
+      "step": 12237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.876564672639864e-06,
+      "loss": 0.8223,
+      "step": 12238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.875873299899526e-06,
+      "loss": 0.8057,
+      "step": 12239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.875181929533927e-06,
+      "loss": 0.8481,
+      "step": 12240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.874490561556295e-06,
+      "loss": 0.8682,
+      "step": 12241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873799195979857e-06,
+      "loss": 0.3472,
+      "step": 12242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873107832817837e-06,
+      "loss": 0.7925,
+      "step": 12243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.872416472083466e-06,
+      "loss": 0.8569,
+      "step": 12244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.871725113789969e-06,
+      "loss": 0.8428,
+      "step": 12245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.871033757950573e-06,
+      "loss": 0.835,
+      "step": 12246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.870342404578502e-06,
+      "loss": 0.8501,
+      "step": 12247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.869651053686987e-06,
+      "loss": 0.8696,
+      "step": 12248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.868959705289254e-06,
+      "loss": 0.8525,
+      "step": 12249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.868268359398526e-06,
+      "loss": 0.8511,
+      "step": 12250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.867577016028034e-06,
+      "loss": 0.8867,
+      "step": 12251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.866885675191001e-06,
+      "loss": 0.9092,
+      "step": 12252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.866194336900653e-06,
+      "loss": 0.8379,
+      "step": 12253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.865503001170221e-06,
+      "loss": 0.9009,
+      "step": 12254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864811668012927e-06,
+      "loss": 0.8296,
+      "step": 12255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864120337441999e-06,
+      "loss": 0.8584,
+      "step": 12256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8634290094706644e-06,
+      "loss": 0.7676,
+      "step": 12257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862737684112147e-06,
+      "loss": 0.8716,
+      "step": 12258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862046361379675e-06,
+      "loss": 0.8687,
+      "step": 12259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.861355041286472e-06,
+      "loss": 0.877,
+      "step": 12260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.860663723845766e-06,
+      "loss": 0.9092,
+      "step": 12261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859972409070784e-06,
+      "loss": 0.833,
+      "step": 12262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859281096974748e-06,
+      "loss": 0.8584,
+      "step": 12263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.858589787570887e-06,
+      "loss": 0.8545,
+      "step": 12264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.857898480872428e-06,
+      "loss": 0.8418,
+      "step": 12265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.857207176892593e-06,
+      "loss": 0.8799,
+      "step": 12266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.856515875644611e-06,
+      "loss": 0.7783,
+      "step": 12267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855824577141706e-06,
+      "loss": 0.8086,
+      "step": 12268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855133281397103e-06,
+      "loss": 0.8594,
+      "step": 12269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8544419884240305e-06,
+      "loss": 0.9131,
+      "step": 12270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.85375069823571e-06,
+      "loss": 0.8457,
+      "step": 12271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8530594108453695e-06,
+      "loss": 0.8794,
+      "step": 12272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.852368126266236e-06,
+      "loss": 0.8706,
+      "step": 12273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851676844511532e-06,
+      "loss": 0.9082,
+      "step": 12274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.850985565594484e-06,
+      "loss": 0.8257,
+      "step": 12275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.850294289528315e-06,
+      "loss": 0.8354,
+      "step": 12276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8496030163262534e-06,
+      "loss": 0.8652,
+      "step": 12277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.848911746001524e-06,
+      "loss": 0.8691,
+      "step": 12278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.848220478567349e-06,
+      "loss": 0.897,
+      "step": 12279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.847529214036957e-06,
+      "loss": 0.8911,
+      "step": 12280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846837952423571e-06,
+      "loss": 0.8804,
+      "step": 12281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846146693740415e-06,
+      "loss": 0.8726,
+      "step": 12282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.845455438000717e-06,
+      "loss": 0.9263,
+      "step": 12283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8447641852177e-06,
+      "loss": 0.8564,
+      "step": 12284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.844072935404587e-06,
+      "loss": 0.9233,
+      "step": 12285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8433816885746065e-06,
+      "loss": 0.3743,
+      "step": 12286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84269044474098e-06,
+      "loss": 0.8506,
+      "step": 12287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8419992039169345e-06,
+      "loss": 0.8877,
+      "step": 12288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.841307966115691e-06,
+      "loss": 0.8491,
+      "step": 12289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.840616731350477e-06,
+      "loss": 0.7803,
+      "step": 12290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.839925499634518e-06,
+      "loss": 0.8682,
+      "step": 12291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.839234270981034e-06,
+      "loss": 0.7944,
+      "step": 12292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.838543045403252e-06,
+      "loss": 0.8662,
+      "step": 12293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.837851822914397e-06,
+      "loss": 0.938,
+      "step": 12294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.837160603527691e-06,
+      "loss": 0.8384,
+      "step": 12295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.836469387256361e-06,
+      "loss": 0.9082,
+      "step": 12296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835778174113628e-06,
+      "loss": 0.8604,
+      "step": 12297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835086964112717e-06,
+      "loss": 0.8359,
+      "step": 12298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.834395757266853e-06,
+      "loss": 0.8496,
+      "step": 12299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.833704553589259e-06,
+      "loss": 0.8545,
+      "step": 12300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.833013353093158e-06,
+      "loss": 0.8237,
+      "step": 12301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.832322155791776e-06,
+      "loss": 0.8384,
+      "step": 12302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.831630961698335e-06,
+      "loss": 0.8066,
+      "step": 12303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83093977082606e-06,
+      "loss": 0.8872,
+      "step": 12304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8302485831881714e-06,
+      "loss": 0.8916,
+      "step": 12305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.829557398797896e-06,
+      "loss": 0.8857,
+      "step": 12306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8288662176684574e-06,
+      "loss": 0.8579,
+      "step": 12307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.828175039813075e-06,
+      "loss": 0.8687,
+      "step": 12308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8274838652449765e-06,
+      "loss": 0.8091,
+      "step": 12309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.826792693977384e-06,
+      "loss": 0.8569,
+      "step": 12310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.826101526023519e-06,
+      "loss": 0.8589,
+      "step": 12311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.825410361396607e-06,
+      "loss": 0.8589,
+      "step": 12312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824719200109869e-06,
+      "loss": 0.8662,
+      "step": 12313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824028042176528e-06,
+      "loss": 0.9155,
+      "step": 12314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.823336887609811e-06,
+      "loss": 0.8354,
+      "step": 12315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8226457364229355e-06,
+      "loss": 0.8794,
+      "step": 12316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.821954588629128e-06,
+      "loss": 0.9194,
+      "step": 12317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8212634442416075e-06,
+      "loss": 0.811,
+      "step": 12318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.820572303273599e-06,
+      "loss": 0.8652,
+      "step": 12319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8198811657383285e-06,
+      "loss": 0.7856,
+      "step": 12320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8191900316490105e-06,
+      "loss": 0.8584,
+      "step": 12321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.818498901018875e-06,
+      "loss": 0.8828,
+      "step": 12322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.817807773861142e-06,
+      "loss": 0.8203,
+      "step": 12323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.817116650189031e-06,
+      "loss": 0.8643,
+      "step": 12324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8164255300157695e-06,
+      "loss": 0.3438,
+      "step": 12325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.815734413354575e-06,
+      "loss": 0.8252,
+      "step": 12326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8150433002186705e-06,
+      "loss": 0.8809,
+      "step": 12327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.814352190621282e-06,
+      "loss": 0.8511,
+      "step": 12328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.813661084575627e-06,
+      "loss": 0.8882,
+      "step": 12329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.812969982094928e-06,
+      "loss": 0.9038,
+      "step": 12330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8122788831924105e-06,
+      "loss": 0.3933,
+      "step": 12331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8115877878812925e-06,
+      "loss": 0.8535,
+      "step": 12332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.810896696174798e-06,
+      "loss": 0.8882,
+      "step": 12333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.810205608086145e-06,
+      "loss": 0.8765,
+      "step": 12334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.809514523628559e-06,
+      "loss": 0.9028,
+      "step": 12335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808823442815263e-06,
+      "loss": 0.8047,
+      "step": 12336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808132365659473e-06,
+      "loss": 0.833,
+      "step": 12337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.807441292174412e-06,
+      "loss": 0.8013,
+      "step": 12338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.806750222373306e-06,
+      "loss": 0.8833,
+      "step": 12339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80605915626937e-06,
+      "loss": 0.874,
+      "step": 12340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80536809387583e-06,
+      "loss": 0.8574,
+      "step": 12341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.804677035205903e-06,
+      "loss": 0.8774,
+      "step": 12342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.803985980272814e-06,
+      "loss": 0.3177,
+      "step": 12343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.803294929089782e-06,
+      "loss": 0.8589,
+      "step": 12344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.802603881670026e-06,
+      "loss": 0.8462,
+      "step": 12345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.801912838026773e-06,
+      "loss": 0.9253,
+      "step": 12346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.801221798173235e-06,
+      "loss": 0.7744,
+      "step": 12347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.800530762122637e-06,
+      "loss": 0.8799,
+      "step": 12348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.799839729888204e-06,
+      "loss": 0.8105,
+      "step": 12349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.799148701483149e-06,
+      "loss": 0.9165,
+      "step": 12350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798457676920696e-06,
+      "loss": 0.8955,
+      "step": 12351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.797766656214068e-06,
+      "loss": 0.8491,
+      "step": 12352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.79707563937648e-06,
+      "loss": 0.8433,
+      "step": 12353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.796384626421157e-06,
+      "loss": 0.854,
+      "step": 12354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.795693617361314e-06,
+      "loss": 0.8838,
+      "step": 12355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7950026122101755e-06,
+      "loss": 0.8657,
+      "step": 12356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.794311610980961e-06,
+      "loss": 0.8794,
+      "step": 12357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.793620613686888e-06,
+      "loss": 0.8818,
+      "step": 12358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.792929620341178e-06,
+      "loss": 0.8657,
+      "step": 12359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7922386309570526e-06,
+      "loss": 0.8818,
+      "step": 12360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.791547645547727e-06,
+      "loss": 0.874,
+      "step": 12361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.790856664126425e-06,
+      "loss": 0.8237,
+      "step": 12362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.790165686706364e-06,
+      "loss": 0.8555,
+      "step": 12363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789474713300763e-06,
+      "loss": 0.7915,
+      "step": 12364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7887837439228456e-06,
+      "loss": 0.8574,
+      "step": 12365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7880927785858256e-06,
+      "loss": 0.8418,
+      "step": 12366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.787401817302924e-06,
+      "loss": 0.8105,
+      "step": 12367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786710860087363e-06,
+      "loss": 0.8848,
+      "step": 12368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786019906952358e-06,
+      "loss": 0.9092,
+      "step": 12369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785328957911131e-06,
+      "loss": 0.8843,
+      "step": 12370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.784638012976896e-06,
+      "loss": 0.856,
+      "step": 12371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.783947072162877e-06,
+      "loss": 0.8516,
+      "step": 12372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.783256135482292e-06,
+      "loss": 0.853,
+      "step": 12373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.782565202948356e-06,
+      "loss": 0.8325,
+      "step": 12374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.781874274574292e-06,
+      "loss": 0.897,
+      "step": 12375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.781183350373319e-06,
+      "loss": 0.8667,
+      "step": 12376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.78049243035865e-06,
+      "loss": 0.8574,
+      "step": 12377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.779801514543509e-06,
+      "loss": 0.8921,
+      "step": 12378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77911060294111e-06,
+      "loss": 0.8584,
+      "step": 12379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7784196955646735e-06,
+      "loss": 0.8945,
+      "step": 12380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.777728792427418e-06,
+      "loss": 0.8379,
+      "step": 12381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.777037893542561e-06,
+      "loss": 0.874,
+      "step": 12382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.776346998923322e-06,
+      "loss": 0.8589,
+      "step": 12383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.775656108582915e-06,
+      "loss": 0.8604,
+      "step": 12384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77496522253456e-06,
+      "loss": 0.8877,
+      "step": 12385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.774274340791477e-06,
+      "loss": 0.8628,
+      "step": 12386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.773583463366879e-06,
+      "loss": 0.8979,
+      "step": 12387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.772892590273988e-06,
+      "loss": 0.8813,
+      "step": 12388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.772201721526021e-06,
+      "loss": 0.9048,
+      "step": 12389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77151085713619e-06,
+      "loss": 0.8979,
+      "step": 12390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7708199971177195e-06,
+      "loss": 0.8813,
+      "step": 12391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.770129141483823e-06,
+      "loss": 0.8623,
+      "step": 12392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.769438290247717e-06,
+      "loss": 0.8447,
+      "step": 12393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.768747443422621e-06,
+      "loss": 0.9165,
+      "step": 12394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.768056601021751e-06,
+      "loss": 0.8438,
+      "step": 12395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7673657630583215e-06,
+      "loss": 0.8784,
+      "step": 12396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.766674929545554e-06,
+      "loss": 0.8081,
+      "step": 12397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.765984100496662e-06,
+      "loss": 0.8711,
+      "step": 12398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7652932759248645e-06,
+      "loss": 0.8101,
+      "step": 12399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.764602455843374e-06,
+      "loss": 0.8335,
+      "step": 12400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.76391164026541e-06,
+      "loss": 0.8794,
+      "step": 12401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.76322082920419e-06,
+      "loss": 0.8857,
+      "step": 12402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.762530022672926e-06,
+      "loss": 0.8657,
+      "step": 12403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761839220684838e-06,
+      "loss": 0.9033,
+      "step": 12404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761148423253143e-06,
+      "loss": 0.853,
+      "step": 12405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7604576303910525e-06,
+      "loss": 0.8633,
+      "step": 12406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.759766842111787e-06,
+      "loss": 0.8936,
+      "step": 12407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.759076058428558e-06,
+      "loss": 0.877,
+      "step": 12408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.758385279354585e-06,
+      "loss": 0.8027,
+      "step": 12409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7576945049030826e-06,
+      "loss": 0.8579,
+      "step": 12410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.757003735087266e-06,
+      "loss": 0.8857,
+      "step": 12411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.756312969920352e-06,
+      "loss": 0.8657,
+      "step": 12412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.755622209415551e-06,
+      "loss": 0.8169,
+      "step": 12413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7549314535860854e-06,
+      "loss": 0.8452,
+      "step": 12414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.754240702445167e-06,
+      "loss": 0.833,
+      "step": 12415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75354995600601e-06,
+      "loss": 0.8945,
+      "step": 12416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752859214281831e-06,
+      "loss": 0.8823,
+      "step": 12417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752168477285846e-06,
+      "loss": 0.8223,
+      "step": 12418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.751477745031268e-06,
+      "loss": 0.8477,
+      "step": 12419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7507870175313126e-06,
+      "loss": 0.8447,
+      "step": 12420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.750096294799193e-06,
+      "loss": 0.9292,
+      "step": 12421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.749405576848125e-06,
+      "loss": 0.7993,
+      "step": 12422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.748714863691326e-06,
+      "loss": 0.897,
+      "step": 12423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.748024155342005e-06,
+      "loss": 0.9102,
+      "step": 12424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.747333451813379e-06,
+      "loss": 0.9189,
+      "step": 12425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746642753118663e-06,
+      "loss": 0.8848,
+      "step": 12426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7459520592710715e-06,
+      "loss": 0.8682,
+      "step": 12427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7452613702838166e-06,
+      "loss": 0.8359,
+      "step": 12428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.744570686170112e-06,
+      "loss": 0.8643,
+      "step": 12429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.743880006943173e-06,
+      "loss": 0.873,
+      "step": 12430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.743189332616214e-06,
+      "loss": 0.895,
+      "step": 12431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.742498663202447e-06,
+      "loss": 0.918,
+      "step": 12432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741807998715086e-06,
+      "loss": 0.8535,
+      "step": 12433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741117339167346e-06,
+      "loss": 0.8257,
+      "step": 12434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.740426684572438e-06,
+      "loss": 0.7568,
+      "step": 12435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.739736034943578e-06,
+      "loss": 0.8813,
+      "step": 12436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.739045390293977e-06,
+      "loss": 0.8691,
+      "step": 12437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7383547506368474e-06,
+      "loss": 0.8032,
+      "step": 12438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.737664115985407e-06,
+      "loss": 0.8682,
+      "step": 12439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.736973486352863e-06,
+      "loss": 0.8096,
+      "step": 12440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.736282861752431e-06,
+      "loss": 0.8921,
+      "step": 12441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.735592242197324e-06,
+      "loss": 0.8599,
+      "step": 12442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.734901627700754e-06,
+      "loss": 0.9048,
+      "step": 12443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.734211018275934e-06,
+      "loss": 0.8945,
+      "step": 12444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.733520413936075e-06,
+      "loss": 0.9106,
+      "step": 12445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73282981469439e-06,
+      "loss": 0.8501,
+      "step": 12446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.732139220564095e-06,
+      "loss": 0.8667,
+      "step": 12447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.731448631558395e-06,
+      "loss": 0.9067,
+      "step": 12448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730758047690508e-06,
+      "loss": 0.8354,
+      "step": 12449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730067468973642e-06,
+      "loss": 0.8892,
+      "step": 12450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7293768954210115e-06,
+      "loss": 0.8472,
+      "step": 12451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.728686327045828e-06,
+      "loss": 0.8481,
+      "step": 12452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7279957638613015e-06,
+      "loss": 0.8213,
+      "step": 12453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7273052058806455e-06,
+      "loss": 0.8564,
+      "step": 12454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.726614653117071e-06,
+      "loss": 0.9106,
+      "step": 12455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725924105583788e-06,
+      "loss": 0.8477,
+      "step": 12456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725233563294011e-06,
+      "loss": 0.8682,
+      "step": 12457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.724543026260946e-06,
+      "loss": 0.8555,
+      "step": 12458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.723852494497807e-06,
+      "loss": 0.8403,
+      "step": 12459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.723161968017807e-06,
+      "loss": 0.8882,
+      "step": 12460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.722471446834154e-06,
+      "loss": 0.854,
+      "step": 12461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.721780930960058e-06,
+      "loss": 0.9219,
+      "step": 12462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.721090420408734e-06,
+      "loss": 0.8535,
+      "step": 12463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.720399915193388e-06,
+      "loss": 0.9116,
+      "step": 12464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.719709415327234e-06,
+      "loss": 0.8892,
+      "step": 12465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.719018920823478e-06,
+      "loss": 0.8555,
+      "step": 12466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.718328431695334e-06,
+      "loss": 0.8721,
+      "step": 12467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.717637947956012e-06,
+      "loss": 0.9106,
+      "step": 12468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.716947469618719e-06,
+      "loss": 0.8276,
+      "step": 12469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.716256996696668e-06,
+      "loss": 0.8745,
+      "step": 12470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.715566529203068e-06,
+      "loss": 0.8896,
+      "step": 12471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.714876067151128e-06,
+      "loss": 0.9336,
+      "step": 12472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.714185610554059e-06,
+      "loss": 0.8413,
+      "step": 12473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.713495159425069e-06,
+      "loss": 0.9204,
+      "step": 12474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712804713777367e-06,
+      "loss": 0.8083,
+      "step": 12475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712114273624166e-06,
+      "loss": 0.873,
+      "step": 12476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.71142383897867e-06,
+      "loss": 0.8765,
+      "step": 12477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7107334098540904e-06,
+      "loss": 0.8828,
+      "step": 12478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7100429862636395e-06,
+      "loss": 0.8218,
+      "step": 12479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.709352568220522e-06,
+      "loss": 0.9019,
+      "step": 12480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.708662155737948e-06,
+      "loss": 0.8564,
+      "step": 12481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.707971748829124e-06,
+      "loss": 0.8833,
+      "step": 12482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.707281347507262e-06,
+      "loss": 0.8931,
+      "step": 12483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.70659095178557e-06,
+      "loss": 0.811,
+      "step": 12484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7059005616772544e-06,
+      "loss": 0.3765,
+      "step": 12485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.705210177195526e-06,
+      "loss": 0.8613,
+      "step": 12486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.704519798353589e-06,
+      "loss": 0.8955,
+      "step": 12487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.703829425164654e-06,
+      "loss": 0.835,
+      "step": 12488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.703139057641931e-06,
+      "loss": 0.8638,
+      "step": 12489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.702448695798625e-06,
+      "loss": 0.8403,
+      "step": 12490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.701758339647942e-06,
+      "loss": 0.8545,
+      "step": 12491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.701067989203095e-06,
+      "loss": 0.335,
+      "step": 12492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700377644477287e-06,
+      "loss": 0.917,
+      "step": 12493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.699687305483728e-06,
+      "loss": 0.8379,
+      "step": 12494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.698996972235622e-06,
+      "loss": 0.8301,
+      "step": 12495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.69830664474618e-06,
+      "loss": 0.981,
+      "step": 12496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6976163230286074e-06,
+      "loss": 0.7793,
+      "step": 12497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6969260070961085e-06,
+      "loss": 0.8564,
+      "step": 12498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.696235696961895e-06,
+      "loss": 0.9272,
+      "step": 12499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.695545392639171e-06,
+      "loss": 0.8457,
+      "step": 12500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.694855094141142e-06,
+      "loss": 0.9082,
+      "step": 12501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6941648014810186e-06,
+      "loss": 0.8203,
+      "step": 12502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.693474514672003e-06,
+      "loss": 0.8354,
+      "step": 12503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.692784233727301e-06,
+      "loss": 0.875,
+      "step": 12504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.692093958660123e-06,
+      "loss": 0.8696,
+      "step": 12505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6914036894836726e-06,
+      "loss": 0.8926,
+      "step": 12506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.690713426211155e-06,
+      "loss": 0.8726,
+      "step": 12507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.690023168855779e-06,
+      "loss": 0.8057,
+      "step": 12508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689332917430746e-06,
+      "loss": 0.9302,
+      "step": 12509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.688642671949265e-06,
+      "loss": 0.8164,
+      "step": 12510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.68795243242454e-06,
+      "loss": 0.8843,
+      "step": 12511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.687262198869776e-06,
+      "loss": 0.7576,
+      "step": 12512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.68657197129818e-06,
+      "loss": 0.9336,
+      "step": 12513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685881749722955e-06,
+      "loss": 0.8604,
+      "step": 12514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685191534157309e-06,
+      "loss": 0.9355,
+      "step": 12515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6845013246144436e-06,
+      "loss": 0.8237,
+      "step": 12516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.683811121107565e-06,
+      "loss": 0.9043,
+      "step": 12517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.683120923649878e-06,
+      "loss": 0.894,
+      "step": 12518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682430732254588e-06,
+      "loss": 0.8799,
+      "step": 12519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6817405469348964e-06,
+      "loss": 0.8979,
+      "step": 12520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.681050367704012e-06,
+      "loss": 0.8276,
+      "step": 12521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.680360194575136e-06,
+      "loss": 0.8804,
+      "step": 12522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.679670027561474e-06,
+      "loss": 0.8296,
+      "step": 12523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.678979866676226e-06,
+      "loss": 0.7827,
+      "step": 12524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.678289711932601e-06,
+      "loss": 0.8608,
+      "step": 12525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.677599563343802e-06,
+      "loss": 0.8628,
+      "step": 12526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6769094209230295e-06,
+      "loss": 0.8623,
+      "step": 12527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.676219284683489e-06,
+      "loss": 0.8286,
+      "step": 12528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.675529154638385e-06,
+      "loss": 0.8716,
+      "step": 12529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.674839030800918e-06,
+      "loss": 0.8667,
+      "step": 12530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.674148913184295e-06,
+      "loss": 0.9194,
+      "step": 12531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.673458801801715e-06,
+      "loss": 0.8887,
+      "step": 12532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.672768696666382e-06,
+      "loss": 0.8564,
+      "step": 12533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.672078597791501e-06,
+      "loss": 0.2953,
+      "step": 12534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.671388505190273e-06,
+      "loss": 0.8066,
+      "step": 12535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6706984188759e-06,
+      "loss": 0.8711,
+      "step": 12536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.670008338861587e-06,
+      "loss": 0.8091,
+      "step": 12537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.669318265160533e-06,
+      "loss": 0.8682,
+      "step": 12538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6686281977859426e-06,
+      "loss": 0.854,
+      "step": 12539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.667938136751015e-06,
+      "loss": 0.8672,
+      "step": 12540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.667248082068955e-06,
+      "loss": 0.8647,
+      "step": 12541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.666558033752966e-06,
+      "loss": 0.8264,
+      "step": 12542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.665867991816243e-06,
+      "loss": 0.8716,
+      "step": 12543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.665177956271994e-06,
+      "loss": 0.8076,
+      "step": 12544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6644879271334195e-06,
+      "loss": 0.8174,
+      "step": 12545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663797904413717e-06,
+      "loss": 0.8325,
+      "step": 12546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663107888126093e-06,
+      "loss": 0.8726,
+      "step": 12547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6624178782837445e-06,
+      "loss": 0.877,
+      "step": 12548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.661727874899873e-06,
+      "loss": 0.895,
+      "step": 12549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.661037877987683e-06,
+      "loss": 0.8428,
+      "step": 12550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.660347887560371e-06,
+      "loss": 0.8721,
+      "step": 12551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.659657903631139e-06,
+      "loss": 0.8701,
+      "step": 12552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.658967926213187e-06,
+      "loss": 0.8525,
+      "step": 12553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.658277955319717e-06,
+      "loss": 0.8555,
+      "step": 12554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6575879909639285e-06,
+      "loss": 0.8916,
+      "step": 12555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6568980331590205e-06,
+      "loss": 0.8438,
+      "step": 12556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.656208081918193e-06,
+      "loss": 0.8242,
+      "step": 12557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.655518137254649e-06,
+      "loss": 0.8696,
+      "step": 12558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.654828199181584e-06,
+      "loss": 0.8735,
+      "step": 12559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.654138267712201e-06,
+      "loss": 0.8955,
+      "step": 12560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6534483428596965e-06,
+      "loss": 0.9321,
+      "step": 12561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652758424637271e-06,
+      "loss": 0.9282,
+      "step": 12562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6520685130581254e-06,
+      "loss": 0.8579,
+      "step": 12563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.651378608135456e-06,
+      "loss": 0.873,
+      "step": 12564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.650688709882464e-06,
+      "loss": 0.9009,
+      "step": 12565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.649998818312349e-06,
+      "loss": 0.876,
+      "step": 12566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.649308933438305e-06,
+      "loss": 0.918,
+      "step": 12567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6486190552735375e-06,
+      "loss": 0.8062,
+      "step": 12568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.647929183831237e-06,
+      "loss": 0.7993,
+      "step": 12569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.647239319124607e-06,
+      "loss": 0.7559,
+      "step": 12570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.646549461166847e-06,
+      "loss": 0.9478,
+      "step": 12571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.645859609971151e-06,
+      "loss": 0.832,
+      "step": 12572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.645169765550718e-06,
+      "loss": 0.8452,
+      "step": 12573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.644479927918749e-06,
+      "loss": 0.8848,
+      "step": 12574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.643790097088438e-06,
+      "loss": 0.8638,
+      "step": 12575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.643100273072985e-06,
+      "loss": 0.834,
+      "step": 12576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.642410455885583e-06,
+      "loss": 0.8535,
+      "step": 12577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641720645539435e-06,
+      "loss": 0.9438,
+      "step": 12578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641030842047735e-06,
+      "loss": 0.8979,
+      "step": 12579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.64034104542368e-06,
+      "loss": 0.875,
+      "step": 12580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6396512556804695e-06,
+      "loss": 0.8809,
+      "step": 12581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.638961472831297e-06,
+      "loss": 0.8657,
+      "step": 12582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.638271696889359e-06,
+      "loss": 0.8823,
+      "step": 12583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637581927867856e-06,
+      "loss": 0.8687,
+      "step": 12584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.63689216577998e-06,
+      "loss": 0.8091,
+      "step": 12585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6362024106389295e-06,
+      "loss": 0.8711,
+      "step": 12586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.635512662457901e-06,
+      "loss": 0.8799,
+      "step": 12587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.634822921250089e-06,
+      "loss": 0.8779,
+      "step": 12588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.634133187028691e-06,
+      "loss": 0.9004,
+      "step": 12589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6334434598068994e-06,
+      "loss": 0.9224,
+      "step": 12590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.632753739597914e-06,
+      "loss": 0.8271,
+      "step": 12591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.632064026414928e-06,
+      "loss": 0.8247,
+      "step": 12592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6313743202711355e-06,
+      "loss": 0.8374,
+      "step": 12593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630684621179734e-06,
+      "loss": 0.9048,
+      "step": 12594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.629994929153919e-06,
+      "loss": 0.8628,
+      "step": 12595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.629305244206882e-06,
+      "loss": 0.8911,
+      "step": 12596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.628615566351822e-06,
+      "loss": 0.9019,
+      "step": 12597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.627925895601929e-06,
+      "loss": 0.8691,
+      "step": 12598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.627236231970401e-06,
+      "loss": 0.9048,
+      "step": 12599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.626546575470432e-06,
+      "loss": 0.9009,
+      "step": 12600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625856926115215e-06,
+      "loss": 0.792,
+      "step": 12601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625167283917943e-06,
+      "loss": 0.8447,
+      "step": 12602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6244776488918145e-06,
+      "loss": 0.8535,
+      "step": 12603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.623788021050018e-06,
+      "loss": 0.8057,
+      "step": 12604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6230984004057525e-06,
+      "loss": 0.8198,
+      "step": 12605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.622408786972204e-06,
+      "loss": 0.9038,
+      "step": 12606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6217191807625734e-06,
+      "loss": 0.7988,
+      "step": 12607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.621029581790051e-06,
+      "loss": 0.7983,
+      "step": 12608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.620339990067828e-06,
+      "loss": 0.8604,
+      "step": 12609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.619650405609101e-06,
+      "loss": 0.8994,
+      "step": 12610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.618960828427061e-06,
+      "loss": 0.8818,
+      "step": 12611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6182712585349e-06,
+      "loss": 0.8862,
+      "step": 12612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.617581695945812e-06,
+      "loss": 0.8154,
+      "step": 12613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6168921406729885e-06,
+      "loss": 0.8921,
+      "step": 12614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6162025927296215e-06,
+      "loss": 0.896,
+      "step": 12615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6155130521289045e-06,
+      "loss": 0.8765,
+      "step": 12616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.614823518884028e-06,
+      "loss": 0.8906,
+      "step": 12617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.614133993008187e-06,
+      "loss": 0.8643,
+      "step": 12618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.613444474514567e-06,
+      "loss": 0.8062,
+      "step": 12619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.612754963416365e-06,
+      "loss": 0.8687,
+      "step": 12620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6120654597267725e-06,
+      "loss": 0.8403,
+      "step": 12621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.611375963458976e-06,
+      "loss": 0.8652,
+      "step": 12622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.610686474626171e-06,
+      "loss": 0.877,
+      "step": 12623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609996993241549e-06,
+      "loss": 0.8638,
+      "step": 12624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609307519318295e-06,
+      "loss": 0.8735,
+      "step": 12625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.608618052869608e-06,
+      "loss": 0.7983,
+      "step": 12626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.607928593908673e-06,
+      "loss": 0.9243,
+      "step": 12627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60723914244868e-06,
+      "loss": 0.8223,
+      "step": 12628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.606549698502824e-06,
+      "loss": 0.8506,
+      "step": 12629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60586026208429e-06,
+      "loss": 0.9004,
+      "step": 12630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60517083320627e-06,
+      "loss": 0.8682,
+      "step": 12631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.604481411881956e-06,
+      "loss": 0.7852,
+      "step": 12632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.603791998124536e-06,
+      "loss": 0.8613,
+      "step": 12633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.603102591947199e-06,
+      "loss": 0.4221,
+      "step": 12634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.602413193363133e-06,
+      "loss": 0.8257,
+      "step": 12635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6017238023855305e-06,
+      "loss": 0.8096,
+      "step": 12636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60103441902758e-06,
+      "loss": 0.8809,
+      "step": 12637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6003450433024685e-06,
+      "loss": 0.8643,
+      "step": 12638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.599655675223386e-06,
+      "loss": 0.8687,
+      "step": 12639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598966314803523e-06,
+      "loss": 0.8149,
+      "step": 12640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598276962056064e-06,
+      "loss": 0.8794,
+      "step": 12641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.597587616994202e-06,
+      "loss": 0.9111,
+      "step": 12642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.596898279631123e-06,
+      "loss": 0.8931,
+      "step": 12643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.596208949980012e-06,
+      "loss": 0.8799,
+      "step": 12644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5955196280540635e-06,
+      "loss": 0.8574,
+      "step": 12645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.594830313866461e-06,
+      "loss": 0.8594,
+      "step": 12646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5941410074303935e-06,
+      "loss": 0.8457,
+      "step": 12647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5934517087590464e-06,
+      "loss": 0.9116,
+      "step": 12648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.592762417865609e-06,
+      "loss": 0.8857,
+      "step": 12649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5920731347632705e-06,
+      "loss": 0.8799,
+      "step": 12650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.591383859465212e-06,
+      "loss": 0.8608,
+      "step": 12651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.590694591984627e-06,
+      "loss": 0.8408,
+      "step": 12652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5900053323346995e-06,
+      "loss": 0.854,
+      "step": 12653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.589316080528614e-06,
+      "loss": 0.8242,
+      "step": 12654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.588626836579562e-06,
+      "loss": 0.8296,
+      "step": 12655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.587937600500725e-06,
+      "loss": 0.8335,
+      "step": 12656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.587248372305291e-06,
+      "loss": 0.8906,
+      "step": 12657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.586559152006446e-06,
+      "loss": 0.9092,
+      "step": 12658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.585869939617377e-06,
+      "loss": 0.8374,
+      "step": 12659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.585180735151266e-06,
+      "loss": 0.8965,
+      "step": 12660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.584491538621304e-06,
+      "loss": 0.8481,
+      "step": 12661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.583802350040673e-06,
+      "loss": 0.8486,
+      "step": 12662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.583113169422559e-06,
+      "loss": 0.9185,
+      "step": 12663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582423996780146e-06,
+      "loss": 0.9302,
+      "step": 12664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58173483212662e-06,
+      "loss": 0.8276,
+      "step": 12665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.581045675475168e-06,
+      "loss": 0.9043,
+      "step": 12666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5803565268389695e-06,
+      "loss": 0.8247,
+      "step": 12667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.579667386231212e-06,
+      "loss": 0.9438,
+      "step": 12668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578978253665083e-06,
+      "loss": 0.3396,
+      "step": 12669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.57828912915376e-06,
+      "loss": 0.8447,
+      "step": 12670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.577600012710433e-06,
+      "loss": 0.8535,
+      "step": 12671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.576910904348282e-06,
+      "loss": 0.877,
+      "step": 12672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.576221804080491e-06,
+      "loss": 0.8213,
+      "step": 12673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.575532711920247e-06,
+      "loss": 0.3303,
+      "step": 12674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.574843627880729e-06,
+      "loss": 0.8438,
+      "step": 12675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.574154551975124e-06,
+      "loss": 0.8162,
+      "step": 12676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5734654842166135e-06,
+      "loss": 0.8667,
+      "step": 12677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5727764246183795e-06,
+      "loss": 0.9028,
+      "step": 12678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.572087373193608e-06,
+      "loss": 0.895,
+      "step": 12679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.571398329955476e-06,
+      "loss": 0.8164,
+      "step": 12680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.57070929491717e-06,
+      "loss": 0.8506,
+      "step": 12681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.570020268091873e-06,
+      "loss": 0.9028,
+      "step": 12682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5693312494927654e-06,
+      "loss": 0.9077,
+      "step": 12683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.568642239133031e-06,
+      "loss": 0.8188,
+      "step": 12684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.567953237025847e-06,
+      "loss": 0.8667,
+      "step": 12685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5672642431844e-06,
+      "loss": 0.8413,
+      "step": 12686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.566575257621872e-06,
+      "loss": 0.8682,
+      "step": 12687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565886280351439e-06,
+      "loss": 0.8462,
+      "step": 12688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565197311386287e-06,
+      "loss": 0.7651,
+      "step": 12689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.564508350739596e-06,
+      "loss": 0.8794,
+      "step": 12690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.563819398424545e-06,
+      "loss": 0.8525,
+      "step": 12691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.563130454454319e-06,
+      "loss": 0.7988,
+      "step": 12692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562441518842094e-06,
+      "loss": 0.8428,
+      "step": 12693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.561752591601051e-06,
+      "loss": 0.8755,
+      "step": 12694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.561063672744375e-06,
+      "loss": 0.8154,
+      "step": 12695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.56037476228524e-06,
+      "loss": 0.8408,
+      "step": 12696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5596858602368276e-06,
+      "loss": 0.8848,
+      "step": 12697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.558996966612321e-06,
+      "loss": 0.8354,
+      "step": 12698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.558308081424897e-06,
+      "loss": 0.8564,
+      "step": 12699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.557619204687736e-06,
+      "loss": 0.79,
+      "step": 12700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556930336414014e-06,
+      "loss": 0.853,
+      "step": 12701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556241476616914e-06,
+      "loss": 0.9204,
+      "step": 12702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.555552625309616e-06,
+      "loss": 0.8955,
+      "step": 12703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.554863782505294e-06,
+      "loss": 0.8301,
+      "step": 12704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.554174948217129e-06,
+      "loss": 0.8472,
+      "step": 12705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.553486122458302e-06,
+      "loss": 0.8442,
+      "step": 12706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.552797305241988e-06,
+      "loss": 0.8335,
+      "step": 12707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5521084965813665e-06,
+      "loss": 0.8975,
+      "step": 12708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551419696489615e-06,
+      "loss": 0.8667,
+      "step": 12709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.550730904979911e-06,
+      "loss": 0.3513,
+      "step": 12710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.550042122065435e-06,
+      "loss": 0.8413,
+      "step": 12711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54935334775936e-06,
+      "loss": 0.8433,
+      "step": 12712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.548664582074865e-06,
+      "loss": 0.9082,
+      "step": 12713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54797582502513e-06,
+      "loss": 0.8232,
+      "step": 12714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.547287076623329e-06,
+      "loss": 0.8486,
+      "step": 12715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54659833688264e-06,
+      "loss": 0.9092,
+      "step": 12716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.545909605816237e-06,
+      "loss": 0.8711,
+      "step": 12717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.545220883437302e-06,
+      "loss": 0.8965,
+      "step": 12718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.544532169759006e-06,
+      "loss": 0.8398,
+      "step": 12719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.543843464794528e-06,
+      "loss": 0.8232,
+      "step": 12720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.543154768557044e-06,
+      "loss": 0.8037,
+      "step": 12721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.542466081059728e-06,
+      "loss": 0.917,
+      "step": 12722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.541777402315756e-06,
+      "loss": 0.8623,
+      "step": 12723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5410887323383065e-06,
+      "loss": 0.8257,
+      "step": 12724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.540400071140552e-06,
+      "loss": 0.9033,
+      "step": 12725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.539711418735667e-06,
+      "loss": 0.8599,
+      "step": 12726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.539022775136831e-06,
+      "loss": 0.8169,
+      "step": 12727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.538334140357213e-06,
+      "loss": 0.8296,
+      "step": 12728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.537645514409992e-06,
+      "loss": 0.8364,
+      "step": 12729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536956897308339e-06,
+      "loss": 0.8955,
+      "step": 12730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536268289065432e-06,
+      "loss": 0.8359,
+      "step": 12731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.535579689694444e-06,
+      "loss": 0.9253,
+      "step": 12732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.534891099208545e-06,
+      "loss": 0.8135,
+      "step": 12733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.534202517620914e-06,
+      "loss": 0.8301,
+      "step": 12734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.533513944944724e-06,
+      "loss": 0.918,
+      "step": 12735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532825381193146e-06,
+      "loss": 0.8579,
+      "step": 12736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532136826379356e-06,
+      "loss": 0.8271,
+      "step": 12737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.531448280516525e-06,
+      "loss": 0.7803,
+      "step": 12738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530759743617825e-06,
+      "loss": 0.9321,
+      "step": 12739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530071215696433e-06,
+      "loss": 0.8481,
+      "step": 12740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.529382696765519e-06,
+      "loss": 0.9053,
+      "step": 12741
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528694186838254e-06,
+      "loss": 0.9268,
+      "step": 12742
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5280056859278135e-06,
+      "loss": 0.9097,
+      "step": 12743
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.527317194047367e-06,
+      "loss": 0.7798,
+      "step": 12744
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.526628711210088e-06,
+      "loss": 0.8662,
+      "step": 12745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525940237429147e-06,
+      "loss": 0.853,
+      "step": 12746
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525251772717717e-06,
+      "loss": 0.855,
+      "step": 12747
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5245633170889695e-06,
+      "loss": 0.8721,
+      "step": 12748
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523874870556073e-06,
+      "loss": 0.8906,
+      "step": 12749
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523186433132202e-06,
+      "loss": 0.8325,
+      "step": 12750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.522498004830525e-06,
+      "loss": 0.8315,
+      "step": 12751
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521809585664213e-06,
+      "loss": 0.8599,
+      "step": 12752
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521121175646439e-06,
+      "loss": 0.8628,
+      "step": 12753
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.52043277479037e-06,
+      "loss": 0.9419,
+      "step": 12754
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519744383109177e-06,
+      "loss": 0.8657,
+      "step": 12755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519056000616034e-06,
+      "loss": 0.8765,
+      "step": 12756
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.518367627324105e-06,
+      "loss": 0.8589,
+      "step": 12757
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5176792632465635e-06,
+      "loss": 0.8535,
+      "step": 12758
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.516990908396576e-06,
+      "loss": 0.8604,
+      "step": 12759
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.516302562787315e-06,
+      "loss": 0.8213,
+      "step": 12760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.515614226431949e-06,
+      "loss": 0.8535,
+      "step": 12761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.514925899343644e-06,
+      "loss": 0.8599,
+      "step": 12762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.514237581535571e-06,
+      "loss": 0.835,
+      "step": 12763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.513549273020901e-06,
+      "loss": 0.8521,
+      "step": 12764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512860973812797e-06,
+      "loss": 0.8384,
+      "step": 12765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512172683924432e-06,
+      "loss": 0.8149,
+      "step": 12766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.511484403368972e-06,
+      "loss": 0.8779,
+      "step": 12767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5107961321595836e-06,
+      "loss": 0.8721,
+      "step": 12768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.510107870309439e-06,
+      "loss": 0.9009,
+      "step": 12769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5094196178317015e-06,
+      "loss": 0.874,
+      "step": 12770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.508731374739539e-06,
+      "loss": 0.8555,
+      "step": 12771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50804314104612e-06,
+      "loss": 0.8828,
+      "step": 12772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.507354916764611e-06,
+      "loss": 0.8696,
+      "step": 12773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50666670190818e-06,
+      "loss": 0.8599,
+      "step": 12774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5059784964899896e-06,
+      "loss": 0.8428,
+      "step": 12775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50529030052321e-06,
+      "loss": 0.9082,
+      "step": 12776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.504602114021009e-06,
+      "loss": 0.8423,
+      "step": 12777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.503913936996548e-06,
+      "loss": 0.8945,
+      "step": 12778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.503225769462993e-06,
+      "loss": 0.8247,
+      "step": 12779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.502537611433516e-06,
+      "loss": 0.8789,
+      "step": 12780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.501849462921276e-06,
+      "loss": 0.8364,
+      "step": 12781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.501161323939442e-06,
+      "loss": 0.8608,
+      "step": 12782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.500473194501178e-06,
+      "loss": 0.7886,
+      "step": 12783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.499785074619647e-06,
+      "loss": 0.8438,
+      "step": 12784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.499096964308019e-06,
+      "loss": 0.8594,
+      "step": 12785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4984088635794515e-06,
+      "loss": 0.8389,
+      "step": 12786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.497720772447117e-06,
+      "loss": 0.8809,
+      "step": 12787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.497032690924172e-06,
+      "loss": 0.9199,
+      "step": 12788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4963446190237845e-06,
+      "loss": 0.7964,
+      "step": 12789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4956565567591215e-06,
+      "loss": 0.8901,
+      "step": 12790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.49496850414334e-06,
+      "loss": 0.9004,
+      "step": 12791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.494280461189607e-06,
+      "loss": 0.8706,
+      "step": 12792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4935924279110875e-06,
+      "loss": 0.8174,
+      "step": 12793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.492904404320942e-06,
+      "loss": 0.8989,
+      "step": 12794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.492216390432335e-06,
+      "loss": 0.9082,
+      "step": 12795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.491528386258427e-06,
+      "loss": 0.9155,
+      "step": 12796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4908403918123825e-06,
+      "loss": 0.877,
+      "step": 12797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.490152407107365e-06,
+      "loss": 0.8823,
+      "step": 12798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.489464432156534e-06,
+      "loss": 0.8525,
+      "step": 12799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.488776466973054e-06,
+      "loss": 0.8589,
+      "step": 12800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.488088511570086e-06,
+      "loss": 0.8755,
+      "step": 12801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.487400565960789e-06,
+      "loss": 0.856,
+      "step": 12802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.48671263015833e-06,
+      "loss": 0.8638,
+      "step": 12803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.486024704175867e-06,
+      "loss": 0.854,
+      "step": 12804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.485336788026558e-06,
+      "loss": 0.8965,
+      "step": 12805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.48464888172357e-06,
+      "loss": 0.8687,
+      "step": 12806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.483960985280061e-06,
+      "loss": 0.8921,
+      "step": 12807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.48327309870919e-06,
+      "loss": 0.875,
+      "step": 12808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.482585222024121e-06,
+      "loss": 0.8931,
+      "step": 12809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.48189735523801e-06,
+      "loss": 0.875,
+      "step": 12810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.481209498364022e-06,
+      "loss": 0.9038,
+      "step": 12811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.480521651415311e-06,
+      "loss": 0.8262,
+      "step": 12812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.479833814405039e-06,
+      "loss": 0.8896,
+      "step": 12813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.479145987346369e-06,
+      "loss": 0.938,
+      "step": 12814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.478458170252453e-06,
+      "loss": 0.8457,
+      "step": 12815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477770363136458e-06,
+      "loss": 0.8608,
+      "step": 12816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477082566011536e-06,
+      "loss": 0.7876,
+      "step": 12817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4763947788908495e-06,
+      "loss": 0.8608,
+      "step": 12818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4757070017875565e-06,
+      "loss": 0.875,
+      "step": 12819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.475019234714814e-06,
+      "loss": 0.8501,
+      "step": 12820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.47433147768578e-06,
+      "loss": 0.8794,
+      "step": 12821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.473643730713615e-06,
+      "loss": 0.8623,
+      "step": 12822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.472955993811474e-06,
+      "loss": 0.9219,
+      "step": 12823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.472268266992517e-06,
+      "loss": 0.8794,
+      "step": 12824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4715805502698965e-06,
+      "loss": 0.8369,
+      "step": 12825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.470892843656774e-06,
+      "loss": 0.8701,
+      "step": 12826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.470205147166307e-06,
+      "loss": 0.9087,
+      "step": 12827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.469517460811648e-06,
+      "loss": 0.8843,
+      "step": 12828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4688297846059554e-06,
+      "loss": 0.8677,
+      "step": 12829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.468142118562389e-06,
+      "loss": 0.8726,
+      "step": 12830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4674544626941e-06,
+      "loss": 0.7905,
+      "step": 12831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4667668170142475e-06,
+      "loss": 0.8423,
+      "step": 12832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.466079181535985e-06,
+      "loss": 0.8501,
+      "step": 12833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.465391556272468e-06,
+      "loss": 0.2924,
+      "step": 12834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.464703941236855e-06,
+      "loss": 0.8604,
+      "step": 12835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.464016336442299e-06,
+      "loss": 0.8979,
+      "step": 12836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.463328741901953e-06,
+      "loss": 0.8647,
+      "step": 12837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4626411576289766e-06,
+      "loss": 0.8696,
+      "step": 12838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46195358363652e-06,
+      "loss": 0.8179,
+      "step": 12839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.461266019937741e-06,
+      "loss": 0.8257,
+      "step": 12840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46057846654579e-06,
+      "loss": 0.8823,
+      "step": 12841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.459890923473824e-06,
+      "loss": 0.853,
+      "step": 12842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.459203390734996e-06,
+      "loss": 0.8789,
+      "step": 12843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.458515868342457e-06,
+      "loss": 0.916,
+      "step": 12844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.457828356309364e-06,
+      "loss": 0.833,
+      "step": 12845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.457140854648871e-06,
+      "loss": 0.8042,
+      "step": 12846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.456453363374126e-06,
+      "loss": 0.9229,
+      "step": 12847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.455765882498287e-06,
+      "loss": 0.8325,
+      "step": 12848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.455078412034503e-06,
+      "loss": 0.8306,
+      "step": 12849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.454390951995927e-06,
+      "loss": 0.876,
+      "step": 12850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.453703502395713e-06,
+      "loss": 0.9004,
+      "step": 12851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.453016063247012e-06,
+      "loss": 0.8296,
+      "step": 12852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.452328634562975e-06,
+      "loss": 0.8394,
+      "step": 12853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451641216356754e-06,
+      "loss": 0.8608,
+      "step": 12854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.450953808641501e-06,
+      "loss": 0.9648,
+      "step": 12855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.450266411430368e-06,
+      "loss": 0.7998,
+      "step": 12856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.449579024736503e-06,
+      "loss": 0.8794,
+      "step": 12857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4488916485730585e-06,
+      "loss": 0.916,
+      "step": 12858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4482042829531875e-06,
+      "loss": 0.8408,
+      "step": 12859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.447516927890035e-06,
+      "loss": 0.9038,
+      "step": 12860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.446829583396758e-06,
+      "loss": 0.9155,
+      "step": 12861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4461422494865e-06,
+      "loss": 0.8066,
+      "step": 12862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.445454926172413e-06,
+      "loss": 0.8232,
+      "step": 12863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.444767613467649e-06,
+      "loss": 0.8965,
+      "step": 12864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.444080311385355e-06,
+      "loss": 0.8784,
+      "step": 12865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.443393019938678e-06,
+      "loss": 0.9106,
+      "step": 12866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.442705739140774e-06,
+      "loss": 0.8735,
+      "step": 12867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.442018469004785e-06,
+      "loss": 0.8677,
+      "step": 12868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.441331209543864e-06,
+      "loss": 0.4098,
+      "step": 12869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4406439607711546e-06,
+      "loss": 0.8149,
+      "step": 12870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.439956722699809e-06,
+      "loss": 0.8418,
+      "step": 12871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.439269495342975e-06,
+      "loss": 0.3332,
+      "step": 12872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.438582278713796e-06,
+      "loss": 0.8457,
+      "step": 12873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4378950728254245e-06,
+      "loss": 0.8345,
+      "step": 12874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.437207877691006e-06,
+      "loss": 0.7983,
+      "step": 12875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.436520693323687e-06,
+      "loss": 0.8608,
+      "step": 12876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.435833519736617e-06,
+      "loss": 0.8721,
+      "step": 12877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.435146356942939e-06,
+      "loss": 0.8955,
+      "step": 12878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434459204955801e-06,
+      "loss": 0.4374,
+      "step": 12879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.433772063788351e-06,
+      "loss": 0.9277,
+      "step": 12880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.433084933453734e-06,
+      "loss": 0.8159,
+      "step": 12881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4323978139650955e-06,
+      "loss": 0.855,
+      "step": 12882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.431710705335579e-06,
+      "loss": 0.8779,
+      "step": 12883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.431023607578334e-06,
+      "loss": 0.855,
+      "step": 12884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.430336520706504e-06,
+      "loss": 0.9214,
+      "step": 12885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4296494447332325e-06,
+      "loss": 0.325,
+      "step": 12886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.428962379671667e-06,
+      "loss": 0.3228,
+      "step": 12887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.428275325534952e-06,
+      "loss": 0.9058,
+      "step": 12888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.427588282336229e-06,
+      "loss": 0.8599,
+      "step": 12889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.426901250088646e-06,
+      "loss": 0.8735,
+      "step": 12890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.426214228805344e-06,
+      "loss": 0.8569,
+      "step": 12891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.425527218499467e-06,
+      "loss": 0.8975,
+      "step": 12892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.424840219184162e-06,
+      "loss": 0.9609,
+      "step": 12893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.424153230872569e-06,
+      "loss": 0.8892,
+      "step": 12894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.42346625357783e-06,
+      "loss": 0.877,
+      "step": 12895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.422779287313094e-06,
+      "loss": 0.8823,
+      "step": 12896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.422092332091497e-06,
+      "loss": 0.8779,
+      "step": 12897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.421405387926186e-06,
+      "loss": 0.8374,
+      "step": 12898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.420718454830301e-06,
+      "loss": 0.8369,
+      "step": 12899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.420031532816983e-06,
+      "loss": 0.8931,
+      "step": 12900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.419344621899379e-06,
+      "loss": 0.8628,
+      "step": 12901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.418657722090624e-06,
+      "loss": 0.8975,
+      "step": 12902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.417970833403864e-06,
+      "loss": 0.8926,
+      "step": 12903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.417283955852241e-06,
+      "loss": 0.9146,
+      "step": 12904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.416597089448893e-06,
+      "loss": 0.843,
+      "step": 12905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.415910234206962e-06,
+      "loss": 0.8794,
+      "step": 12906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.415223390139588e-06,
+      "loss": 0.8579,
+      "step": 12907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.414536557259912e-06,
+      "loss": 0.8477,
+      "step": 12908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.413849735581076e-06,
+      "loss": 0.8408,
+      "step": 12909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.413162925116216e-06,
+      "loss": 0.8232,
+      "step": 12910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.412476125878475e-06,
+      "loss": 0.8042,
+      "step": 12911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.411789337880993e-06,
+      "loss": 0.8193,
+      "step": 12912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4111025611369045e-06,
+      "loss": 0.9033,
+      "step": 12913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.410415795659355e-06,
+      "loss": 0.8433,
+      "step": 12914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.40972904146148e-06,
+      "loss": 0.8584,
+      "step": 12915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.409042298556416e-06,
+      "loss": 0.8032,
+      "step": 12916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.408355566957307e-06,
+      "loss": 0.876,
+      "step": 12917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.407668846677287e-06,
+      "loss": 0.7573,
+      "step": 12918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.406982137729497e-06,
+      "loss": 0.8423,
+      "step": 12919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.406295440127071e-06,
+      "loss": 0.2925,
+      "step": 12920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.405608753883148e-06,
+      "loss": 0.8843,
+      "step": 12921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.40492207901087e-06,
+      "loss": 0.8311,
+      "step": 12922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404235415523367e-06,
+      "loss": 0.7524,
+      "step": 12923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.40354876343378e-06,
+      "loss": 0.8306,
+      "step": 12924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.402862122755246e-06,
+      "loss": 0.9087,
+      "step": 12925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4021754935009e-06,
+      "loss": 0.8115,
+      "step": 12926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.401488875683879e-06,
+      "loss": 0.8474,
+      "step": 12927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.400802269317319e-06,
+      "loss": 0.8979,
+      "step": 12928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4001156744143545e-06,
+      "loss": 0.8809,
+      "step": 12929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.399429090988125e-06,
+      "loss": 0.8159,
+      "step": 12930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398742519051762e-06,
+      "loss": 0.8916,
+      "step": 12931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398055958618401e-06,
+      "loss": 0.9248,
+      "step": 12932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.39736940970118e-06,
+      "loss": 0.8384,
+      "step": 12933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3966828723132306e-06,
+      "loss": 0.8135,
+      "step": 12934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.39599634646769e-06,
+      "loss": 0.9307,
+      "step": 12935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395309832177689e-06,
+      "loss": 0.7886,
+      "step": 12936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.394623329456365e-06,
+      "loss": 0.897,
+      "step": 12937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.393936838316851e-06,
+      "loss": 0.9072,
+      "step": 12938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3932503587722785e-06,
+      "loss": 0.9199,
+      "step": 12939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.392563890835784e-06,
+      "loss": 0.8408,
+      "step": 12940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3918774345205e-06,
+      "loss": 0.8315,
+      "step": 12941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.391190989839558e-06,
+      "loss": 0.8682,
+      "step": 12942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.390504556806093e-06,
+      "loss": 0.9043,
+      "step": 12943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3898181354332355e-06,
+      "loss": 0.8447,
+      "step": 12944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.389131725734117e-06,
+      "loss": 0.7871,
+      "step": 12945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3884453277218745e-06,
+      "loss": 0.8232,
+      "step": 12946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.387758941409634e-06,
+      "loss": 0.8643,
+      "step": 12947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38707256681053e-06,
+      "loss": 0.8755,
+      "step": 12948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.386386203937694e-06,
+      "loss": 0.8589,
+      "step": 12949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.385699852804258e-06,
+      "loss": 0.8608,
+      "step": 12950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.385013513423351e-06,
+      "loss": 0.8784,
+      "step": 12951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.384327185808104e-06,
+      "loss": 0.8521,
+      "step": 12952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.383640869971648e-06,
+      "loss": 0.877,
+      "step": 12953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.382954565927115e-06,
+      "loss": 0.3427,
+      "step": 12954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3822682736876315e-06,
+      "loss": 0.896,
+      "step": 12955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.381581993266331e-06,
+      "loss": 0.8984,
+      "step": 12956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.380895724676341e-06,
+      "loss": 0.8755,
+      "step": 12957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38020946793079e-06,
+      "loss": 0.8857,
+      "step": 12958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.379523223042811e-06,
+      "loss": 0.8633,
+      "step": 12959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.378836990025529e-06,
+      "loss": 0.8872,
+      "step": 12960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.378150768892073e-06,
+      "loss": 0.8779,
+      "step": 12961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3774645596555754e-06,
+      "loss": 0.9043,
+      "step": 12962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.376778362329161e-06,
+      "loss": 0.8823,
+      "step": 12963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3760921769259585e-06,
+      "loss": 0.8521,
+      "step": 12964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.375406003459094e-06,
+      "loss": 0.8425,
+      "step": 12965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.374719841941698e-06,
+      "loss": 0.8384,
+      "step": 12966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.374033692386899e-06,
+      "loss": 0.875,
+      "step": 12967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.373347554807818e-06,
+      "loss": 0.8594,
+      "step": 12968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.372661429217588e-06,
+      "loss": 0.8716,
+      "step": 12969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371975315629334e-06,
+      "loss": 0.8833,
+      "step": 12970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3712892140561795e-06,
+      "loss": 0.8667,
+      "step": 12971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3706031245112556e-06,
+      "loss": 0.9004,
+      "step": 12972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369917047007684e-06,
+      "loss": 0.8345,
+      "step": 12973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369230981558592e-06,
+      "loss": 0.8882,
+      "step": 12974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.368544928177108e-06,
+      "loss": 0.8145,
+      "step": 12975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3678588868763524e-06,
+      "loss": 0.8711,
+      "step": 12976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.367172857669451e-06,
+      "loss": 0.9258,
+      "step": 12977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.366486840569532e-06,
+      "loss": 0.8252,
+      "step": 12978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.365800835589718e-06,
+      "loss": 0.8906,
+      "step": 12979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.365114842743135e-06,
+      "loss": 0.8081,
+      "step": 12980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.364428862042902e-06,
+      "loss": 0.9199,
+      "step": 12981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.363742893502149e-06,
+      "loss": 0.9092,
+      "step": 12982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.363056937133996e-06,
+      "loss": 0.855,
+      "step": 12983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.362370992951567e-06,
+      "loss": 0.8862,
+      "step": 12984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.361685060967987e-06,
+      "loss": 0.4463,
+      "step": 12985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360999141196377e-06,
+      "loss": 0.8525,
+      "step": 12986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360313233649859e-06,
+      "loss": 0.9287,
+      "step": 12987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.35962733834156e-06,
+      "loss": 0.7778,
+      "step": 12988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358941455284596e-06,
+      "loss": 0.8901,
+      "step": 12989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358255584492093e-06,
+      "loss": 0.8721,
+      "step": 12990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.357569725977173e-06,
+      "loss": 0.877,
+      "step": 12991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356883879752957e-06,
+      "loss": 0.8091,
+      "step": 12992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356198045832565e-06,
+      "loss": 0.894,
+      "step": 12993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.355512224229118e-06,
+      "loss": 0.8101,
+      "step": 12994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.354826414955738e-06,
+      "loss": 0.8506,
+      "step": 12995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.354140618025547e-06,
+      "loss": 0.8379,
+      "step": 12996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.353454833451661e-06,
+      "loss": 0.8604,
+      "step": 12997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.352769061247204e-06,
+      "loss": 0.856,
+      "step": 12998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.352083301425296e-06,
+      "loss": 0.8486,
+      "step": 12999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.351397553999054e-06,
+      "loss": 0.8594,
+      "step": 13000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.350711818981599e-06,
+      "loss": 0.8608,
+      "step": 13001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.350026096386051e-06,
+      "loss": 0.8101,
+      "step": 13002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.349340386225526e-06,
+      "loss": 0.8408,
+      "step": 13003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.348654688513147e-06,
+      "loss": 0.877,
+      "step": 13004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347969003262029e-06,
+      "loss": 0.8765,
+      "step": 13005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347283330485291e-06,
+      "loss": 0.7891,
+      "step": 13006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.346597670196054e-06,
+      "loss": 0.8105,
+      "step": 13007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.345912022407429e-06,
+      "loss": 0.9028,
+      "step": 13008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3452263871325425e-06,
+      "loss": 0.8608,
+      "step": 13009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3445407643845035e-06,
+      "loss": 0.8687,
+      "step": 13010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3438551541764325e-06,
+      "loss": 0.8271,
+      "step": 13011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.343169556521449e-06,
+      "loss": 0.8882,
+      "step": 13012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.342483971432665e-06,
+      "loss": 0.7515,
+      "step": 13013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.341798398923197e-06,
+      "loss": 0.8682,
+      "step": 13014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.341112839006167e-06,
+      "loss": 0.8149,
+      "step": 13015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.340427291694685e-06,
+      "loss": 0.8003,
+      "step": 13016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3397417570018684e-06,
+      "loss": 0.834,
+      "step": 13017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.339056234940831e-06,
+      "loss": 0.9424,
+      "step": 13018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.33837072552469e-06,
+      "loss": 0.8599,
+      "step": 13019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.337685228766561e-06,
+      "loss": 0.812,
+      "step": 13020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.336999744679555e-06,
+      "loss": 0.8667,
+      "step": 13021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.336314273276791e-06,
+      "loss": 0.9033,
+      "step": 13022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.33562881457138e-06,
+      "loss": 0.4041,
+      "step": 13023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334943368576435e-06,
+      "loss": 0.8369,
+      "step": 13024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334257935305073e-06,
+      "loss": 0.8354,
+      "step": 13025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.333572514770406e-06,
+      "loss": 0.8936,
+      "step": 13026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.332887106985545e-06,
+      "loss": 0.8242,
+      "step": 13027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.332201711963607e-06,
+      "loss": 0.8447,
+      "step": 13028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.331516329717702e-06,
+      "loss": 0.8701,
+      "step": 13029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3308309602609435e-06,
+      "loss": 0.9067,
+      "step": 13030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3301456036064415e-06,
+      "loss": 0.8774,
+      "step": 13031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.329460259767311e-06,
+      "loss": 0.8896,
+      "step": 13032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3287749287566635e-06,
+      "loss": 0.8906,
+      "step": 13033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3280896105876066e-06,
+      "loss": 0.3654,
+      "step": 13034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.327404305273255e-06,
+      "loss": 0.8916,
+      "step": 13035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3267190128267215e-06,
+      "loss": 0.8804,
+      "step": 13036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.32603373326111e-06,
+      "loss": 0.8877,
+      "step": 13037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.325348466589539e-06,
+      "loss": 0.3644,
+      "step": 13038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.324663212825113e-06,
+      "loss": 0.8369,
+      "step": 13039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.323977971980944e-06,
+      "loss": 0.8618,
+      "step": 13040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3232927440701425e-06,
+      "loss": 0.853,
+      "step": 13041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.322607529105817e-06,
+      "loss": 0.8755,
+      "step": 13042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.321922327101075e-06,
+      "loss": 0.8794,
+      "step": 13043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.32123713806903e-06,
+      "loss": 0.8208,
+      "step": 13044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3205519620227875e-06,
+      "loss": 0.8296,
+      "step": 13045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3198667989754575e-06,
+      "loss": 0.3711,
+      "step": 13046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.319181648940144e-06,
+      "loss": 0.8618,
+      "step": 13047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.318496511929961e-06,
+      "loss": 0.9272,
+      "step": 13048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317811387958014e-06,
+      "loss": 0.8184,
+      "step": 13049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317126277037408e-06,
+      "loss": 0.8735,
+      "step": 13050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.316441179181255e-06,
+      "loss": 0.9165,
+      "step": 13051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315756094402659e-06,
+      "loss": 0.8667,
+      "step": 13052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315071022714724e-06,
+      "loss": 0.8872,
+      "step": 13053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.314385964130563e-06,
+      "loss": 0.7798,
+      "step": 13054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3137009186632785e-06,
+      "loss": 0.8594,
+      "step": 13055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.313015886325975e-06,
+      "loss": 0.8384,
+      "step": 13056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.312330867131762e-06,
+      "loss": 0.9165,
+      "step": 13057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.311645861093743e-06,
+      "loss": 0.8252,
+      "step": 13058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3109608682250234e-06,
+      "loss": 0.8848,
+      "step": 13059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.310275888538707e-06,
+      "loss": 0.8857,
+      "step": 13060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.309590922047901e-06,
+      "loss": 0.8853,
+      "step": 13061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3089059687657085e-06,
+      "loss": 0.8501,
+      "step": 13062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308221028705232e-06,
+      "loss": 0.8486,
+      "step": 13063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.307536101879579e-06,
+      "loss": 0.8428,
+      "step": 13064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.306851188301852e-06,
+      "loss": 0.7881,
+      "step": 13065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.306166287985152e-06,
+      "loss": 0.834,
+      "step": 13066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3054814009425876e-06,
+      "loss": 0.9307,
+      "step": 13067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.304796527187256e-06,
+      "loss": 0.8276,
+      "step": 13068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3041116667322614e-06,
+      "loss": 0.8994,
+      "step": 13069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.303426819590709e-06,
+      "loss": 0.4082,
+      "step": 13070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302741985775699e-06,
+      "loss": 0.8965,
+      "step": 13071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302057165300332e-06,
+      "loss": 0.8354,
+      "step": 13072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.301372358177713e-06,
+      "loss": 0.8413,
+      "step": 13073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300687564420942e-06,
+      "loss": 0.8955,
+      "step": 13074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.30000278404312e-06,
+      "loss": 0.8906,
+      "step": 13075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.299318017057346e-06,
+      "loss": 0.8794,
+      "step": 13076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.298633263476723e-06,
+      "loss": 0.9058,
+      "step": 13077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2979485233143525e-06,
+      "loss": 0.8916,
+      "step": 13078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.297263796583331e-06,
+      "loss": 0.915,
+      "step": 13079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.296579083296761e-06,
+      "loss": 0.7749,
+      "step": 13080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.295894383467742e-06,
+      "loss": 0.8677,
+      "step": 13081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.295209697109372e-06,
+      "loss": 0.8672,
+      "step": 13082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.294525024234753e-06,
+      "loss": 0.8828,
+      "step": 13083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293840364856981e-06,
+      "loss": 0.8442,
+      "step": 13084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293155718989154e-06,
+      "loss": 0.8169,
+      "step": 13085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.292471086644375e-06,
+      "loss": 0.8306,
+      "step": 13086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291786467835737e-06,
+      "loss": 0.8354,
+      "step": 13087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291101862576341e-06,
+      "loss": 0.8242,
+      "step": 13088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.290417270879283e-06,
+      "loss": 0.7725,
+      "step": 13089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.28973269275766e-06,
+      "loss": 0.8511,
+      "step": 13090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.289048128224571e-06,
+      "loss": 0.854,
+      "step": 13091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.28836357729311e-06,
+      "loss": 0.8945,
+      "step": 13092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.287679039976376e-06,
+      "loss": 0.8306,
+      "step": 13093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286994516287466e-06,
+      "loss": 0.875,
+      "step": 13094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2863100062394715e-06,
+      "loss": 0.896,
+      "step": 13095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.285625509845494e-06,
+      "loss": 0.8237,
+      "step": 13096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284941027118625e-06,
+      "loss": 0.8916,
+      "step": 13097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.28425655807196e-06,
+      "loss": 0.8511,
+      "step": 13098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.283572102718597e-06,
+      "loss": 0.2823,
+      "step": 13099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.282887661071628e-06,
+      "loss": 0.7847,
+      "step": 13100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.282203233144145e-06,
+      "loss": 0.8379,
+      "step": 13101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.281518818949249e-06,
+      "loss": 0.8975,
+      "step": 13102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.28083441850003e-06,
+      "loss": 0.8477,
+      "step": 13103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280150031809581e-06,
+      "loss": 0.8745,
+      "step": 13104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.279465658890996e-06,
+      "loss": 0.8643,
+      "step": 13105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.278781299757367e-06,
+      "loss": 0.8999,
+      "step": 13106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.278096954421791e-06,
+      "loss": 0.8867,
+      "step": 13107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.277412622897355e-06,
+      "loss": 0.7734,
+      "step": 13108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.276728305197157e-06,
+      "loss": 0.856,
+      "step": 13109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.276044001334285e-06,
+      "loss": 0.896,
+      "step": 13110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.275359711321832e-06,
+      "loss": 0.8472,
+      "step": 13111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2746754351728894e-06,
+      "loss": 0.8511,
+      "step": 13112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.273991172900549e-06,
+      "loss": 0.8599,
+      "step": 13113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2733069245179e-06,
+      "loss": 0.8574,
+      "step": 13114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.272622690038038e-06,
+      "loss": 0.8379,
+      "step": 13115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.271938469474048e-06,
+      "loss": 0.835,
+      "step": 13116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.271254262839024e-06,
+      "loss": 0.9272,
+      "step": 13117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.270570070146052e-06,
+      "loss": 0.8833,
+      "step": 13118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2698858914082234e-06,
+      "loss": 0.8818,
+      "step": 13119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.269201726638632e-06,
+      "loss": 0.9399,
+      "step": 13120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.268517575850359e-06,
+      "loss": 0.8652,
+      "step": 13121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267833439056499e-06,
+      "loss": 0.8301,
+      "step": 13122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267149316270141e-06,
+      "loss": 0.877,
+      "step": 13123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2664652075043686e-06,
+      "loss": 0.8984,
+      "step": 13124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2657811127722755e-06,
+      "loss": 0.8667,
+      "step": 13125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.265097032086943e-06,
+      "loss": 0.8403,
+      "step": 13126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.264412965461465e-06,
+      "loss": 0.8735,
+      "step": 13127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.263728912908927e-06,
+      "loss": 0.8267,
+      "step": 13128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2630448744424125e-06,
+      "loss": 0.8525,
+      "step": 13129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.262360850075012e-06,
+      "loss": 0.812,
+      "step": 13130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.261676839819812e-06,
+      "loss": 0.8613,
+      "step": 13131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.260992843689895e-06,
+      "loss": 0.8813,
+      "step": 13132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.260308861698351e-06,
+      "loss": 0.8687,
+      "step": 13133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.259624893858264e-06,
+      "loss": 0.812,
+      "step": 13134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258940940182718e-06,
+      "loss": 0.918,
+      "step": 13135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258257000684803e-06,
+      "loss": 0.8618,
+      "step": 13136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2575730753775975e-06,
+      "loss": 0.8477,
+      "step": 13137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256889164274189e-06,
+      "loss": 0.8784,
+      "step": 13138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256205267387663e-06,
+      "loss": 0.897,
+      "step": 13139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.255521384731102e-06,
+      "loss": 0.9399,
+      "step": 13140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.254837516317591e-06,
+      "loss": 0.8955,
+      "step": 13141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.254153662160211e-06,
+      "loss": 0.8926,
+      "step": 13142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.253469822272047e-06,
+      "loss": 0.8237,
+      "step": 13143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252785996666183e-06,
+      "loss": 0.9463,
+      "step": 13144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252102185355698e-06,
+      "loss": 0.8618,
+      "step": 13145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251418388353678e-06,
+      "loss": 0.833,
+      "step": 13146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.250734605673205e-06,
+      "loss": 0.8335,
+      "step": 13147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.250050837327357e-06,
+      "loss": 0.791,
+      "step": 13148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.249367083329221e-06,
+      "loss": 0.8623,
+      "step": 13149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.248683343691874e-06,
+      "loss": 0.895,
+      "step": 13150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.247999618428398e-06,
+      "loss": 0.873,
+      "step": 13151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.247315907551876e-06,
+      "loss": 0.7896,
+      "step": 13152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.246632211075386e-06,
+      "loss": 0.9102,
+      "step": 13153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.24594852901201e-06,
+      "loss": 0.8501,
+      "step": 13154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.245264861374825e-06,
+      "loss": 0.8604,
+      "step": 13155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.244581208176913e-06,
+      "loss": 0.8354,
+      "step": 13156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.243897569431354e-06,
+      "loss": 0.873,
+      "step": 13157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2432139451512246e-06,
+      "loss": 0.9111,
+      "step": 13158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.242530335349605e-06,
+      "loss": 0.8706,
+      "step": 13159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.241846740039575e-06,
+      "loss": 0.752,
+      "step": 13160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2411631592342106e-06,
+      "loss": 0.9053,
+      "step": 13161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.240479592946592e-06,
+      "loss": 0.9146,
+      "step": 13162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239796041189795e-06,
+      "loss": 0.897,
+      "step": 13163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239112503976896e-06,
+      "loss": 0.8975,
+      "step": 13164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2384289813209754e-06,
+      "loss": 0.8691,
+      "step": 13165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.237745473235109e-06,
+      "loss": 0.8721,
+      "step": 13166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.23706197973237e-06,
+      "loss": 0.8765,
+      "step": 13167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.23637850082584e-06,
+      "loss": 0.9009,
+      "step": 13168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.235695036528593e-06,
+      "loss": 0.8687,
+      "step": 13169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.235011586853705e-06,
+      "loss": 0.8135,
+      "step": 13170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.234328151814247e-06,
+      "loss": 0.8271,
+      "step": 13171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.233644731423301e-06,
+      "loss": 0.897,
+      "step": 13172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2329613256939395e-06,
+      "loss": 0.9087,
+      "step": 13173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.232277934639234e-06,
+      "loss": 0.8604,
+      "step": 13174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.231594558272263e-06,
+      "loss": 0.8384,
+      "step": 13175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2309111966060995e-06,
+      "loss": 0.8965,
+      "step": 13176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.230227849653815e-06,
+      "loss": 0.8008,
+      "step": 13177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.229544517428486e-06,
+      "loss": 0.8716,
+      "step": 13178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2288611999431846e-06,
+      "loss": 0.8408,
+      "step": 13179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2281778972109826e-06,
+      "loss": 0.7998,
+      "step": 13180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227494609244955e-06,
+      "loss": 0.8242,
+      "step": 13181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.226811336058172e-06,
+      "loss": 0.8643,
+      "step": 13182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.226128077663705e-06,
+      "loss": 0.812,
+      "step": 13183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.22544483407463e-06,
+      "loss": 0.8008,
+      "step": 13184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.224761605304015e-06,
+      "loss": 0.8643,
+      "step": 13185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.224078391364933e-06,
+      "loss": 0.8945,
+      "step": 13186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.223395192270451e-06,
+      "loss": 0.3434,
+      "step": 13187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.222712008033646e-06,
+      "loss": 0.8003,
+      "step": 13188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.222028838667584e-06,
+      "loss": 0.8623,
+      "step": 13189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.221345684185335e-06,
+      "loss": 0.8008,
+      "step": 13190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.220662544599972e-06,
+      "loss": 0.897,
+      "step": 13191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.219979419924562e-06,
+      "loss": 0.9067,
+      "step": 13192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.219296310172174e-06,
+      "loss": 0.9033,
+      "step": 13193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.218613215355879e-06,
+      "loss": 0.8569,
+      "step": 13194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2179301354887436e-06,
+      "loss": 0.8721,
+      "step": 13195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.217247070583837e-06,
+      "loss": 0.8623,
+      "step": 13196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.216564020654229e-06,
+      "loss": 0.8135,
+      "step": 13197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.215880985712985e-06,
+      "loss": 0.8809,
+      "step": 13198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.215197965773173e-06,
+      "loss": 0.8535,
+      "step": 13199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21451496084786e-06,
+      "loss": 0.8789,
+      "step": 13200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.213831970950114e-06,
+      "loss": 0.8677,
+      "step": 13201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.213148996093003e-06,
+      "loss": 0.8296,
+      "step": 13202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.212466036289589e-06,
+      "loss": 0.8027,
+      "step": 13203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.211783091552941e-06,
+      "loss": 0.8711,
+      "step": 13204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.211100161896126e-06,
+      "loss": 0.834,
+      "step": 13205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.210417247332207e-06,
+      "loss": 0.9185,
+      "step": 13206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.209734347874252e-06,
+      "loss": 0.8848,
+      "step": 13207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.209051463535323e-06,
+      "loss": 0.8628,
+      "step": 13208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.208368594328485e-06,
+      "loss": 0.8916,
+      "step": 13209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207685740266805e-06,
+      "loss": 0.8667,
+      "step": 13210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207002901363344e-06,
+      "loss": 0.855,
+      "step": 13211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.206320077631166e-06,
+      "loss": 0.8857,
+      "step": 13212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.205637269083338e-06,
+      "loss": 0.9043,
+      "step": 13213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.20495447573292e-06,
+      "loss": 0.8569,
+      "step": 13214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204271697592976e-06,
+      "loss": 0.8052,
+      "step": 13215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.203588934676567e-06,
+      "loss": 0.8843,
+      "step": 13216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202906186996758e-06,
+      "loss": 0.8823,
+      "step": 13217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202223454566611e-06,
+      "loss": 0.894,
+      "step": 13218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.201540737399183e-06,
+      "loss": 0.9214,
+      "step": 13219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2008580355075415e-06,
+      "loss": 0.8872,
+      "step": 13220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.200175348904745e-06,
+      "loss": 0.7764,
+      "step": 13221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.199492677603852e-06,
+      "loss": 0.834,
+      "step": 13222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1988100216179285e-06,
+      "loss": 0.8848,
+      "step": 13223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.19812738096003e-06,
+      "loss": 0.8789,
+      "step": 13224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.197444755643218e-06,
+      "loss": 0.812,
+      "step": 13225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196762145680554e-06,
+      "loss": 0.8271,
+      "step": 13226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196079551085096e-06,
+      "loss": 0.8291,
+      "step": 13227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.195396971869903e-06,
+      "loss": 0.8555,
+      "step": 13228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1947144080480314e-06,
+      "loss": 0.8682,
+      "step": 13229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.194031859632543e-06,
+      "loss": 0.4031,
+      "step": 13230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1933493266364976e-06,
+      "loss": 0.8896,
+      "step": 13231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.192666809072948e-06,
+      "loss": 0.8765,
+      "step": 13232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1919843069549546e-06,
+      "loss": 0.8535,
+      "step": 13233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.191301820295578e-06,
+      "loss": 0.8428,
+      "step": 13234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.190619349107869e-06,
+      "loss": 0.8867,
+      "step": 13235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.189936893404889e-06,
+      "loss": 0.853,
+      "step": 13236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.189254453199689e-06,
+      "loss": 0.8628,
+      "step": 13237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.188572028505332e-06,
+      "loss": 0.8062,
+      "step": 13238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.187889619334872e-06,
+      "loss": 0.8281,
+      "step": 13239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18720722570136e-06,
+      "loss": 0.8218,
+      "step": 13240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.186524847617856e-06,
+      "loss": 0.8882,
+      "step": 13241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.185842485097415e-06,
+      "loss": 0.9072,
+      "step": 13242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.185160138153088e-06,
+      "loss": 0.8623,
+      "step": 13243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.184477806797933e-06,
+      "loss": 0.897,
+      "step": 13244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.183795491045002e-06,
+      "loss": 0.833,
+      "step": 13245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.183113190907349e-06,
+      "loss": 0.8936,
+      "step": 13246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.182430906398029e-06,
+      "loss": 0.9023,
+      "step": 13247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181748637530094e-06,
+      "loss": 0.8926,
+      "step": 13248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181066384316595e-06,
+      "loss": 0.9023,
+      "step": 13249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.180384146770589e-06,
+      "loss": 0.876,
+      "step": 13250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.179701924905125e-06,
+      "loss": 0.8794,
+      "step": 13251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.179019718733257e-06,
+      "loss": 0.8545,
+      "step": 13252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.178337528268033e-06,
+      "loss": 0.8726,
+      "step": 13253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1776553535225075e-06,
+      "loss": 0.8848,
+      "step": 13254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176973194509733e-06,
+      "loss": 0.8262,
+      "step": 13255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176291051242755e-06,
+      "loss": 0.8462,
+      "step": 13256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1756089237346306e-06,
+      "loss": 0.7744,
+      "step": 13257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.174926811998405e-06,
+      "loss": 0.8535,
+      "step": 13258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.174244716047129e-06,
+      "loss": 0.8408,
+      "step": 13259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.173562635893854e-06,
+      "loss": 0.854,
+      "step": 13260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.172880571551627e-06,
+      "loss": 0.8457,
+      "step": 13261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.172198523033497e-06,
+      "loss": 0.9053,
+      "step": 13262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.171516490352517e-06,
+      "loss": 0.814,
+      "step": 13263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17083447352173e-06,
+      "loss": 0.7974,
+      "step": 13264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.170152472554187e-06,
+      "loss": 0.8696,
+      "step": 13265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.169470487462933e-06,
+      "loss": 0.8345,
+      "step": 13266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.168788518261019e-06,
+      "loss": 0.4006,
+      "step": 13267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.168106564961491e-06,
+      "loss": 0.8345,
+      "step": 13268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.167424627577393e-06,
+      "loss": 0.8755,
+      "step": 13269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.166742706121775e-06,
+      "loss": 0.8696,
+      "step": 13270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.166060800607683e-06,
+      "loss": 0.8369,
+      "step": 13271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.165378911048161e-06,
+      "loss": 0.8774,
+      "step": 13272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.164697037456257e-06,
+      "loss": 0.7998,
+      "step": 13273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.164015179845013e-06,
+      "loss": 0.8662,
+      "step": 13274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.163333338227476e-06,
+      "loss": 0.8511,
+      "step": 13275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.162651512616693e-06,
+      "loss": 0.9048,
+      "step": 13276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161969703025704e-06,
+      "loss": 0.8198,
+      "step": 13277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161287909467555e-06,
+      "loss": 0.8423,
+      "step": 13278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.160606131955291e-06,
+      "loss": 0.8291,
+      "step": 13279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.159924370501954e-06,
+      "loss": 0.8613,
+      "step": 13280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.159242625120589e-06,
+      "loss": 0.7817,
+      "step": 13281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1585608958242355e-06,
+      "loss": 0.8286,
+      "step": 13282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1578791826259376e-06,
+      "loss": 0.877,
+      "step": 13283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.157197485538741e-06,
+      "loss": 0.7809,
+      "step": 13284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.156515804575681e-06,
+      "loss": 0.8003,
+      "step": 13285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.155834139749805e-06,
+      "loss": 0.8843,
+      "step": 13286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.155152491074152e-06,
+      "loss": 0.8071,
+      "step": 13287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154470858561761e-06,
+      "loss": 0.7422,
+      "step": 13288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.153789242225678e-06,
+      "loss": 0.8374,
+      "step": 13289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.153107642078938e-06,
+      "loss": 0.876,
+      "step": 13290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.152426058134584e-06,
+      "loss": 0.8813,
+      "step": 13291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.151744490405656e-06,
+      "loss": 0.8901,
+      "step": 13292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.151062938905192e-06,
+      "loss": 0.8667,
+      "step": 13293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1503814036462324e-06,
+      "loss": 0.8594,
+      "step": 13294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.149699884641813e-06,
+      "loss": 0.918,
+      "step": 13295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.149018381904977e-06,
+      "loss": 0.8228,
+      "step": 13296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14833689544876e-06,
+      "loss": 0.875,
+      "step": 13297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1476554252861986e-06,
+      "loss": 0.8789,
+      "step": 13298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146973971430333e-06,
+      "loss": 0.915,
+      "step": 13299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146292533894202e-06,
+      "loss": 0.8916,
+      "step": 13300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.145611112690837e-06,
+      "loss": 0.8477,
+      "step": 13301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.144929707833279e-06,
+      "loss": 0.7612,
+      "step": 13302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1442483193345625e-06,
+      "loss": 0.9028,
+      "step": 13303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.143566947207724e-06,
+      "loss": 0.9048,
+      "step": 13304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.142885591465801e-06,
+      "loss": 0.8794,
+      "step": 13305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.142204252121826e-06,
+      "loss": 0.8604,
+      "step": 13306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.141522929188834e-06,
+      "loss": 0.8945,
+      "step": 13307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.140841622679864e-06,
+      "loss": 0.8359,
+      "step": 13308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.140160332607946e-06,
+      "loss": 0.8691,
+      "step": 13309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139479058986117e-06,
+      "loss": 0.8467,
+      "step": 13310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.138797801827408e-06,
+      "loss": 0.8022,
+      "step": 13311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.138116561144854e-06,
+      "loss": 0.8794,
+      "step": 13312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.137435336951491e-06,
+      "loss": 0.8589,
+      "step": 13313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.136754129260346e-06,
+      "loss": 0.8809,
+      "step": 13314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.136072938084456e-06,
+      "loss": 0.8643,
+      "step": 13315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.135391763436853e-06,
+      "loss": 0.8911,
+      "step": 13316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.134710605330566e-06,
+      "loss": 0.8315,
+      "step": 13317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.134029463778631e-06,
+      "loss": 0.8325,
+      "step": 13318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.133348338794076e-06,
+      "loss": 0.3052,
+      "step": 13319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.132667230389932e-06,
+      "loss": 0.8008,
+      "step": 13320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.131986138579232e-06,
+      "loss": 0.8428,
+      "step": 13321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.131305063375004e-06,
+      "loss": 0.8003,
+      "step": 13322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.130624004790281e-06,
+      "loss": 0.895,
+      "step": 13323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.129942962838088e-06,
+      "loss": 0.8726,
+      "step": 13324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.129261937531459e-06,
+      "loss": 0.8936,
+      "step": 13325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1285809288834215e-06,
+      "loss": 0.8579,
+      "step": 13326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.127899936907002e-06,
+      "loss": 0.8711,
+      "step": 13327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.127218961615232e-06,
+      "loss": 0.8677,
+      "step": 13328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1265380030211395e-06,
+      "loss": 0.873,
+      "step": 13329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.125857061137749e-06,
+      "loss": 0.853,
+      "step": 13330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.125176135978094e-06,
+      "loss": 0.8354,
+      "step": 13331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.124495227555196e-06,
+      "loss": 0.7959,
+      "step": 13332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1238143358820826e-06,
+      "loss": 0.9033,
+      "step": 13333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.123133460971784e-06,
+      "loss": 0.9219,
+      "step": 13334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122452602837323e-06,
+      "loss": 0.8672,
+      "step": 13335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.121771761491726e-06,
+      "loss": 0.8579,
+      "step": 13336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.121090936948021e-06,
+      "loss": 0.9053,
+      "step": 13337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120410129219229e-06,
+      "loss": 0.8286,
+      "step": 13338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.11972933831838e-06,
+      "loss": 0.8457,
+      "step": 13339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1190485642584936e-06,
+      "loss": 0.8496,
+      "step": 13340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.118367807052596e-06,
+      "loss": 0.8525,
+      "step": 13341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117687066713715e-06,
+      "loss": 0.8037,
+      "step": 13342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117006343254867e-06,
+      "loss": 0.8809,
+      "step": 13343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.116325636689079e-06,
+      "loss": 0.874,
+      "step": 13344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.115644947029378e-06,
+      "loss": 0.8281,
+      "step": 13345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.114964274288779e-06,
+      "loss": 0.877,
+      "step": 13346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.11428361848031e-06,
+      "loss": 0.8159,
+      "step": 13347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.11360297961699e-06,
+      "loss": 0.8628,
+      "step": 13348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.112922357711842e-06,
+      "loss": 0.832,
+      "step": 13349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.112241752777889e-06,
+      "loss": 0.853,
+      "step": 13350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1115611648281465e-06,
+      "loss": 0.8862,
+      "step": 13351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.110880593875643e-06,
+      "loss": 0.8501,
+      "step": 13352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.110200039933392e-06,
+      "loss": 0.8408,
+      "step": 13353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.109519503014416e-06,
+      "loss": 0.9307,
+      "step": 13354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.108838983131737e-06,
+      "loss": 0.8828,
+      "step": 13355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1081584802983725e-06,
+      "loss": 0.7759,
+      "step": 13356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.10747799452734e-06,
+      "loss": 0.8921,
+      "step": 13357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1067975258316614e-06,
+      "loss": 0.8394,
+      "step": 13358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.106117074224354e-06,
+      "loss": 0.8369,
+      "step": 13359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.105436639718436e-06,
+      "loss": 0.8394,
+      "step": 13360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.104756222326922e-06,
+      "loss": 0.856,
+      "step": 13361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.104075822062834e-06,
+      "loss": 0.8101,
+      "step": 13362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.103395438939188e-06,
+      "loss": 0.8408,
+      "step": 13363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.102715072968997e-06,
+      "loss": 0.8159,
+      "step": 13364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.102034724165283e-06,
+      "loss": 0.9624,
+      "step": 13365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.101354392541061e-06,
+      "loss": 0.8608,
+      "step": 13366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.100674078109343e-06,
+      "loss": 0.915,
+      "step": 13367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.099993780883149e-06,
+      "loss": 0.8442,
+      "step": 13368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.099313500875492e-06,
+      "loss": 0.8428,
+      "step": 13369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098633238099385e-06,
+      "loss": 0.9043,
+      "step": 13370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.097952992567846e-06,
+      "loss": 0.8147,
+      "step": 13371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.097272764293888e-06,
+      "loss": 0.8384,
+      "step": 13372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.096592553290522e-06,
+      "loss": 0.8413,
+      "step": 13373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.095912359570768e-06,
+      "loss": 0.8198,
+      "step": 13374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.095232183147633e-06,
+      "loss": 0.8887,
+      "step": 13375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.094552024034134e-06,
+      "loss": 0.9136,
+      "step": 13376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.093871882243279e-06,
+      "loss": 0.8564,
+      "step": 13377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.093191757788083e-06,
+      "loss": 0.3528,
+      "step": 13378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.09251165068156e-06,
+      "loss": 0.8574,
+      "step": 13379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0918315609367164e-06,
+      "loss": 0.8892,
+      "step": 13380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.091151488566567e-06,
+      "loss": 0.9458,
+      "step": 13381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.090471433584123e-06,
+      "loss": 0.9126,
+      "step": 13382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.089791396002392e-06,
+      "loss": 0.8564,
+      "step": 13383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.089111375834388e-06,
+      "loss": 0.7788,
+      "step": 13384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.088431373093118e-06,
+      "loss": 0.8101,
+      "step": 13385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0877513877915896e-06,
+      "loss": 0.8247,
+      "step": 13386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.087071419942819e-06,
+      "loss": 0.856,
+      "step": 13387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.086391469559808e-06,
+      "loss": 0.791,
+      "step": 13388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.08571153665557e-06,
+      "loss": 0.874,
+      "step": 13389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.085031621243107e-06,
+      "loss": 0.8462,
+      "step": 13390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.084351723335434e-06,
+      "loss": 0.8481,
+      "step": 13391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.083671842945556e-06,
+      "loss": 0.9458,
+      "step": 13392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.082991980086477e-06,
+      "loss": 0.8936,
+      "step": 13393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.082312134771208e-06,
+      "loss": 0.4304,
+      "step": 13394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.081632307012754e-06,
+      "loss": 0.8267,
+      "step": 13395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.080952496824121e-06,
+      "loss": 0.8232,
+      "step": 13396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.080272704218316e-06,
+      "loss": 0.8506,
+      "step": 13397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.079592929208342e-06,
+      "loss": 0.8809,
+      "step": 13398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078913171807204e-06,
+      "loss": 0.8491,
+      "step": 13399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078233432027912e-06,
+      "loss": 0.8657,
+      "step": 13400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.077553709883466e-06,
+      "loss": 0.8262,
+      "step": 13401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.076874005386869e-06,
+      "loss": 0.875,
+      "step": 13402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.07619431855113e-06,
+      "loss": 0.853,
+      "step": 13403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0755146493892475e-06,
+      "loss": 0.8467,
+      "step": 13404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.074834997914228e-06,
+      "loss": 0.8921,
+      "step": 13405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.074155364139071e-06,
+      "loss": 0.8193,
+      "step": 13406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.073475748076782e-06,
+      "loss": 0.8428,
+      "step": 13407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072796149740362e-06,
+      "loss": 0.8804,
+      "step": 13408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072116569142811e-06,
+      "loss": 0.8086,
+      "step": 13409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.071437006297134e-06,
+      "loss": 0.8096,
+      "step": 13410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.070757461216331e-06,
+      "loss": 0.8677,
+      "step": 13411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0700779339134e-06,
+      "loss": 0.8208,
+      "step": 13412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.069398424401346e-06,
+      "loss": 0.8013,
+      "step": 13413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0687189326931645e-06,
+      "loss": 0.3506,
+      "step": 13414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.068039458801857e-06,
+      "loss": 0.3894,
+      "step": 13415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.067360002740425e-06,
+      "loss": 0.8027,
+      "step": 13416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066680564521865e-06,
+      "loss": 0.9102,
+      "step": 13417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066001144159176e-06,
+      "loss": 0.8687,
+      "step": 13418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.065321741665358e-06,
+      "loss": 0.8232,
+      "step": 13419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.064642357053407e-06,
+      "loss": 0.7993,
+      "step": 13420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.063962990336323e-06,
+      "loss": 0.9014,
+      "step": 13421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0632836415271e-06,
+      "loss": 0.8857,
+      "step": 13422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.062604310638738e-06,
+      "loss": 0.8325,
+      "step": 13423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.061924997684234e-06,
+      "loss": 0.8408,
+      "step": 13424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0612457026765804e-06,
+      "loss": 0.8066,
+      "step": 13425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.060566425628778e-06,
+      "loss": 0.3745,
+      "step": 13426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.05988716655382e-06,
+      "loss": 0.9058,
+      "step": 13427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0592079254647e-06,
+      "loss": 0.8789,
+      "step": 13428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.058528702374418e-06,
+      "loss": 0.8198,
+      "step": 13429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.057849497295963e-06,
+      "loss": 0.8525,
+      "step": 13430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.057170310242331e-06,
+      "loss": 0.8447,
+      "step": 13431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0564911412265195e-06,
+      "loss": 0.8833,
+      "step": 13432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055811990261518e-06,
+      "loss": 0.853,
+      "step": 13433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055132857360322e-06,
+      "loss": 0.8096,
+      "step": 13434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.054453742535921e-06,
+      "loss": 0.8472,
+      "step": 13435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053774645801311e-06,
+      "loss": 0.772,
+      "step": 13436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053095567169484e-06,
+      "loss": 0.8467,
+      "step": 13437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.052416506653429e-06,
+      "loss": 0.8154,
+      "step": 13438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.051737464266141e-06,
+      "loss": 0.9424,
+      "step": 13439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.05105844002061e-06,
+      "loss": 0.8921,
+      "step": 13440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.050379433929825e-06,
+      "loss": 0.8677,
+      "step": 13441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.049700446006779e-06,
+      "loss": 0.8867,
+      "step": 13442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.049021476264461e-06,
+      "loss": 0.8853,
+      "step": 13443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.048342524715858e-06,
+      "loss": 0.8062,
+      "step": 13444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.047663591373966e-06,
+      "loss": 0.8687,
+      "step": 13445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.046984676251768e-06,
+      "loss": 0.8745,
+      "step": 13446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.046305779362254e-06,
+      "loss": 0.854,
+      "step": 13447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.045626900718416e-06,
+      "loss": 0.8291,
+      "step": 13448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044948040333238e-06,
+      "loss": 0.7651,
+      "step": 13449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.04426919821971e-06,
+      "loss": 0.8052,
+      "step": 13450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.043590374390816e-06,
+      "loss": 0.9219,
+      "step": 13451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.042911568859546e-06,
+      "loss": 0.8599,
+      "step": 13452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0422327816388876e-06,
+      "loss": 0.8345,
+      "step": 13453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.041554012741823e-06,
+      "loss": 0.8442,
+      "step": 13454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.040875262181342e-06,
+      "loss": 0.8584,
+      "step": 13455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.040196529970428e-06,
+      "loss": 0.8823,
+      "step": 13456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.039517816122066e-06,
+      "loss": 0.8589,
+      "step": 13457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.038839120649244e-06,
+      "loss": 0.8491,
+      "step": 13458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.038160443564942e-06,
+      "loss": 0.814,
+      "step": 13459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0374817848821466e-06,
+      "loss": 0.8555,
+      "step": 13460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0368031446138435e-06,
+      "loss": 0.8169,
+      "step": 13461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.036124522773011e-06,
+      "loss": 0.8345,
+      "step": 13462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.035445919372639e-06,
+      "loss": 0.9336,
+      "step": 13463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.034767334425703e-06,
+      "loss": 0.8862,
+      "step": 13464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.034088767945189e-06,
+      "loss": 0.8315,
+      "step": 13465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.033410219944082e-06,
+      "loss": 0.8828,
+      "step": 13466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.032731690435359e-06,
+      "loss": 0.8228,
+      "step": 13467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.032053179432002e-06,
+      "loss": 0.8936,
+      "step": 13468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.031374686946995e-06,
+      "loss": 0.856,
+      "step": 13469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.030696212993314e-06,
+      "loss": 0.8354,
+      "step": 13470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.030017757583945e-06,
+      "loss": 0.8638,
+      "step": 13471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.029339320731862e-06,
+      "loss": 0.7734,
+      "step": 13472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.028660902450049e-06,
+      "loss": 0.874,
+      "step": 13473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.027982502751484e-06,
+      "loss": 0.8403,
+      "step": 13474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.027304121649144e-06,
+      "loss": 0.3865,
+      "step": 13475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.026625759156009e-06,
+      "loss": 0.8071,
+      "step": 13476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.025947415285058e-06,
+      "loss": 0.8984,
+      "step": 13477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0252690900492665e-06,
+      "loss": 0.9238,
+      "step": 13478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.024590783461615e-06,
+      "loss": 0.8906,
+      "step": 13479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.023912495535078e-06,
+      "loss": 0.8379,
+      "step": 13480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.023234226282632e-06,
+      "loss": 0.8354,
+      "step": 13481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.022555975717257e-06,
+      "loss": 0.8594,
+      "step": 13482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.021877743851924e-06,
+      "loss": 0.8848,
+      "step": 13483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0211995306996114e-06,
+      "loss": 0.8115,
+      "step": 13484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.020521336273296e-06,
+      "loss": 0.7881,
+      "step": 13485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.01984316058595e-06,
+      "loss": 0.8921,
+      "step": 13486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.01916500365055e-06,
+      "loss": 0.9102,
+      "step": 13487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.018486865480066e-06,
+      "loss": 0.8311,
+      "step": 13488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.017808746087477e-06,
+      "loss": 0.8843,
+      "step": 13489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.017130645485755e-06,
+      "loss": 0.834,
+      "step": 13490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.016452563687871e-06,
+      "loss": 0.8784,
+      "step": 13491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015774500706801e-06,
+      "loss": 0.8848,
+      "step": 13492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015096456555514e-06,
+      "loss": 0.8892,
+      "step": 13493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0144184312469835e-06,
+      "loss": 0.8633,
+      "step": 13494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0137404247941835e-06,
+      "loss": 0.8477,
+      "step": 13495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0130624372100826e-06,
+      "loss": 0.8525,
+      "step": 13496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0123844685076505e-06,
+      "loss": 0.8262,
+      "step": 13497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.011706518699863e-06,
+      "loss": 0.8394,
+      "step": 13498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.011028587799685e-06,
+      "loss": 0.8711,
+      "step": 13499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.010350675820091e-06,
+      "loss": 0.8672,
+      "step": 13500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0096727827740455e-06,
+      "loss": 0.8213,
+      "step": 13501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.008994908674521e-06,
+      "loss": 0.8667,
+      "step": 13502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.008317053534486e-06,
+      "loss": 0.9136,
+      "step": 13503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.007639217366907e-06,
+      "loss": 0.7808,
+      "step": 13504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0069614001847545e-06,
+      "loss": 0.8765,
+      "step": 13505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.006283602000997e-06,
+      "loss": 0.7739,
+      "step": 13506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.005605822828597e-06,
+      "loss": 0.8413,
+      "step": 13507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.004928062680527e-06,
+      "loss": 0.8354,
+      "step": 13508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.00425032156975e-06,
+      "loss": 0.8677,
+      "step": 13509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.003572599509233e-06,
+      "loss": 0.8726,
+      "step": 13510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002894896511944e-06,
+      "loss": 0.8862,
+      "step": 13511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002217212590845e-06,
+      "loss": 0.9082,
+      "step": 13512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.001539547758903e-06,
+      "loss": 0.8296,
+      "step": 13513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.000861902029085e-06,
+      "loss": 0.8506,
+      "step": 13514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0001842754143525e-06,
+      "loss": 0.8921,
+      "step": 13515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.999506667927671e-06,
+      "loss": 0.8447,
+      "step": 13516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9988290795820004e-06,
+      "loss": 0.8804,
+      "step": 13517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.998151510390309e-06,
+      "loss": 0.8481,
+      "step": 13518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.99747396036556e-06,
+      "loss": 0.8477,
+      "step": 13519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.99679642952071e-06,
+      "loss": 0.8599,
+      "step": 13520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9961189178687285e-06,
+      "loss": 0.876,
+      "step": 13521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.995441425422573e-06,
+      "loss": 0.7832,
+      "step": 13522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994763952195204e-06,
+      "loss": 0.8037,
+      "step": 13523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994086498199588e-06,
+      "loss": 0.8657,
+      "step": 13524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.99340906344868e-06,
+      "loss": 0.7618,
+      "step": 13525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.992731647955442e-06,
+      "loss": 0.7998,
+      "step": 13526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.992054251732837e-06,
+      "loss": 0.8696,
+      "step": 13527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9913768747938225e-06,
+      "loss": 0.8286,
+      "step": 13528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.990699517151358e-06,
+      "loss": 0.793,
+      "step": 13529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9900221788184e-06,
+      "loss": 0.8745,
+      "step": 13530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.989344859807911e-06,
+      "loss": 0.8574,
+      "step": 13531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9886675601328475e-06,
+      "loss": 0.8428,
+      "step": 13532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.987990279806166e-06,
+      "loss": 0.8901,
+      "step": 13533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.987313018840826e-06,
+      "loss": 0.896,
+      "step": 13534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.986635777249785e-06,
+      "loss": 0.8369,
+      "step": 13535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985958555045996e-06,
+      "loss": 0.856,
+      "step": 13536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985281352242419e-06,
+      "loss": 0.8276,
+      "step": 13537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.984604168852009e-06,
+      "loss": 0.8818,
+      "step": 13538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9839270048877195e-06,
+      "loss": 0.8433,
+      "step": 13539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98324986036251e-06,
+      "loss": 0.8213,
+      "step": 13540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.982572735289332e-06,
+      "loss": 0.8301,
+      "step": 13541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98189562968114e-06,
+      "loss": 0.8413,
+      "step": 13542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98121854355089e-06,
+      "loss": 0.8584,
+      "step": 13543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.980541476911534e-06,
+      "loss": 0.9121,
+      "step": 13544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.979864429776027e-06,
+      "loss": 0.8896,
+      "step": 13545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.979187402157319e-06,
+      "loss": 0.8857,
+      "step": 13546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.978510394068366e-06,
+      "loss": 0.8867,
+      "step": 13547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.97783340552212e-06,
+      "loss": 0.7983,
+      "step": 13548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.977156436531529e-06,
+      "loss": 0.8999,
+      "step": 13549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.976479487109548e-06,
+      "loss": 0.8579,
+      "step": 13550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9758025572691294e-06,
+      "loss": 0.8657,
+      "step": 13551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9751256470232195e-06,
+      "loss": 0.8838,
+      "step": 13552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.974448756384772e-06,
+      "loss": 0.8496,
+      "step": 13553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.973771885366736e-06,
+      "loss": 0.8813,
+      "step": 13554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9730950339820595e-06,
+      "loss": 0.8311,
+      "step": 13555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.972418202243696e-06,
+      "loss": 0.8013,
+      "step": 13556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.971741390164589e-06,
+      "loss": 0.8311,
+      "step": 13557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9710645977576915e-06,
+      "loss": 0.8154,
+      "step": 13558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9703878250359476e-06,
+      "loss": 0.7861,
+      "step": 13559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.969711072012307e-06,
+      "loss": 0.7566,
+      "step": 13560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.969034338699719e-06,
+      "loss": 0.9048,
+      "step": 13561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.968357625111127e-06,
+      "loss": 0.8335,
+      "step": 13562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.96768093125948e-06,
+      "loss": 0.8813,
+      "step": 13563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.967004257157725e-06,
+      "loss": 0.8867,
+      "step": 13564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9663276028188035e-06,
+      "loss": 0.8228,
+      "step": 13565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.965650968255667e-06,
+      "loss": 0.7974,
+      "step": 13566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.964974353481254e-06,
+      "loss": 0.8794,
+      "step": 13567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.964297758508514e-06,
+      "loss": 0.8843,
+      "step": 13568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.963621183350391e-06,
+      "loss": 0.8267,
+      "step": 13569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.962944628019826e-06,
+      "loss": 0.813,
+      "step": 13570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.962268092529766e-06,
+      "loss": 0.8354,
+      "step": 13571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.961591576893153e-06,
+      "loss": 0.7935,
+      "step": 13572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960915081122928e-06,
+      "loss": 0.7988,
+      "step": 13573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960238605232037e-06,
+      "loss": 0.3842,
+      "step": 13574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.959562149233418e-06,
+      "loss": 0.8262,
+      "step": 13575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.958885713140014e-06,
+      "loss": 0.8408,
+      "step": 13576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.958209296964769e-06,
+      "loss": 0.8662,
+      "step": 13577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.957532900720621e-06,
+      "loss": 0.8872,
+      "step": 13578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.95685652442051e-06,
+      "loss": 0.8721,
+      "step": 13579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.956180168077379e-06,
+      "loss": 0.8462,
+      "step": 13580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.955503831704166e-06,
+      "loss": 0.7903,
+      "step": 13581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.954827515313812e-06,
+      "loss": 0.8672,
+      "step": 13582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.954151218919252e-06,
+      "loss": 0.8809,
+      "step": 13583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.953474942533429e-06,
+      "loss": 0.8613,
+      "step": 13584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952798686169279e-06,
+      "loss": 0.9316,
+      "step": 13585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952122449839739e-06,
+      "loss": 0.8789,
+      "step": 13586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.95144623355775e-06,
+      "loss": 0.9194,
+      "step": 13587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.950770037336246e-06,
+      "loss": 0.8364,
+      "step": 13588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9500938611881635e-06,
+      "loss": 0.895,
+      "step": 13589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.949417705126442e-06,
+      "loss": 0.8477,
+      "step": 13590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.948741569164014e-06,
+      "loss": 0.896,
+      "step": 13591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.948065453313815e-06,
+      "loss": 0.8555,
+      "step": 13592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9473893575887835e-06,
+      "loss": 0.8511,
+      "step": 13593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.946713282001852e-06,
+      "loss": 0.8755,
+      "step": 13594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.946037226565956e-06,
+      "loss": 0.8745,
+      "step": 13595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.945361191294026e-06,
+      "loss": 0.8574,
+      "step": 13596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.944685176199e-06,
+      "loss": 0.855,
+      "step": 13597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.944009181293811e-06,
+      "loss": 0.8403,
+      "step": 13598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.943333206591388e-06,
+      "loss": 0.8447,
+      "step": 13599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942657252104666e-06,
+      "loss": 0.9067,
+      "step": 13600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.941981317846579e-06,
+      "loss": 0.8218,
+      "step": 13601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.941305403830054e-06,
+      "loss": 0.8311,
+      "step": 13602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9406295100680275e-06,
+      "loss": 0.9038,
+      "step": 13603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.939953636573426e-06,
+      "loss": 0.8926,
+      "step": 13604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.939277783359182e-06,
+      "loss": 0.4014,
+      "step": 13605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.938601950438226e-06,
+      "loss": 0.8091,
+      "step": 13606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.937926137823488e-06,
+      "loss": 0.8633,
+      "step": 13607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.937250345527894e-06,
+      "loss": 0.8848,
+      "step": 13608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.936574573564379e-06,
+      "loss": 0.8511,
+      "step": 13609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.935898821945866e-06,
+      "loss": 0.8262,
+      "step": 13610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9352230906852875e-06,
+      "loss": 0.8496,
+      "step": 13611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.934547379795567e-06,
+      "loss": 0.8374,
+      "step": 13612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9338716892896334e-06,
+      "loss": 0.8564,
+      "step": 13613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.933196019180417e-06,
+      "loss": 0.8862,
+      "step": 13614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.932520369480839e-06,
+      "loss": 0.9023,
+      "step": 13615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.93184474020383e-06,
+      "loss": 0.8408,
+      "step": 13616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9311691313623155e-06,
+      "loss": 0.7727,
+      "step": 13617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.930493542969217e-06,
+      "loss": 0.793,
+      "step": 13618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.929817975037465e-06,
+      "loss": 0.8599,
+      "step": 13619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.92914242757998e-06,
+      "loss": 0.835,
+      "step": 13620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.928466900609687e-06,
+      "loss": 0.8818,
+      "step": 13621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.927791394139513e-06,
+      "loss": 0.896,
+      "step": 13622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.927115908182379e-06,
+      "loss": 0.8467,
+      "step": 13623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9264404427512074e-06,
+      "loss": 0.8193,
+      "step": 13624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.925764997858922e-06,
+      "loss": 0.7871,
+      "step": 13625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.925089573518444e-06,
+      "loss": 0.895,
+      "step": 13626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.924414169742698e-06,
+      "loss": 0.7959,
+      "step": 13627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.923738786544602e-06,
+      "loss": 0.8569,
+      "step": 13628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.92306342393708e-06,
+      "loss": 0.9067,
+      "step": 13629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.922388081933053e-06,
+      "loss": 0.9072,
+      "step": 13630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.921712760545437e-06,
+      "loss": 0.8457,
+      "step": 13631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.921037459787159e-06,
+      "loss": 0.8882,
+      "step": 13632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.920362179671132e-06,
+      "loss": 0.8784,
+      "step": 13633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9196869202102775e-06,
+      "loss": 0.8374,
+      "step": 13634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.919011681417516e-06,
+      "loss": 0.8784,
+      "step": 13635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918336463305764e-06,
+      "loss": 0.8169,
+      "step": 13636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.917661265887939e-06,
+      "loss": 0.8418,
+      "step": 13637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916986089176961e-06,
+      "loss": 0.8945,
+      "step": 13638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916310933185745e-06,
+      "loss": 0.8701,
+      "step": 13639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91563579792721e-06,
+      "loss": 0.8701,
+      "step": 13640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.914960683414268e-06,
+      "loss": 0.8276,
+      "step": 13641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91428558965984e-06,
+      "loss": 0.855,
+      "step": 13642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.913610516676839e-06,
+      "loss": 0.8452,
+      "step": 13643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91293546447818e-06,
+      "loss": 0.8584,
+      "step": 13644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91226043307678e-06,
+      "loss": 0.8174,
+      "step": 13645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.911585422485552e-06,
+      "loss": 0.9033,
+      "step": 13646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910910432717409e-06,
+      "loss": 0.9116,
+      "step": 13647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910235463785266e-06,
+      "loss": 0.3182,
+      "step": 13648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.909560515702036e-06,
+      "loss": 0.8091,
+      "step": 13649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.90888558848063e-06,
+      "loss": 0.813,
+      "step": 13650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.908210682133965e-06,
+      "loss": 0.9331,
+      "step": 13651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.907535796674948e-06,
+      "loss": 0.8501,
+      "step": 13652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.906860932116492e-06,
+      "loss": 0.8306,
+      "step": 13653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.906186088471512e-06,
+      "loss": 0.8613,
+      "step": 13654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.905511265752914e-06,
+      "loss": 0.8633,
+      "step": 13655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.904836463973611e-06,
+      "loss": 0.894,
+      "step": 13656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.90416168314651e-06,
+      "loss": 0.833,
+      "step": 13657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.903486923284524e-06,
+      "loss": 0.8496,
+      "step": 13658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.902812184400563e-06,
+      "loss": 0.814,
+      "step": 13659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.902137466507532e-06,
+      "loss": 0.8774,
+      "step": 13660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9014627696183414e-06,
+      "loss": 0.8252,
+      "step": 13661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.900788093745899e-06,
+      "loss": 0.8877,
+      "step": 13662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.900113438903112e-06,
+      "loss": 0.918,
+      "step": 13663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8994388051028895e-06,
+      "loss": 0.8726,
+      "step": 13664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.898764192358137e-06,
+      "loss": 0.938,
+      "step": 13665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.898089600681758e-06,
+      "loss": 0.8564,
+      "step": 13666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.897415030086664e-06,
+      "loss": 0.8838,
+      "step": 13667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.896740480585758e-06,
+      "loss": 0.8516,
+      "step": 13668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.896065952191945e-06,
+      "loss": 0.8462,
+      "step": 13669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8953914449181285e-06,
+      "loss": 0.4823,
+      "step": 13670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.894716958777215e-06,
+      "loss": 0.8936,
+      "step": 13671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.894042493782109e-06,
+      "loss": 0.7764,
+      "step": 13672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.893368049945711e-06,
+      "loss": 0.8457,
+      "step": 13673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.892693627280926e-06,
+      "loss": 0.8135,
+      "step": 13674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.892019225800659e-06,
+      "loss": 0.8916,
+      "step": 13675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.891344845517808e-06,
+      "loss": 0.9243,
+      "step": 13676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.890670486445279e-06,
+      "loss": 0.8516,
+      "step": 13677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.889996148595971e-06,
+      "loss": 0.7935,
+      "step": 13678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.889321831982784e-06,
+      "loss": 0.8721,
+      "step": 13679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.888647536618623e-06,
+      "loss": 0.8525,
+      "step": 13680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.887973262516383e-06,
+      "loss": 0.8042,
+      "step": 13681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.88729900968897e-06,
+      "loss": 0.8315,
+      "step": 13682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8866247781492795e-06,
+      "loss": 0.8428,
+      "step": 13683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.88595056791021e-06,
+      "loss": 0.8154,
+      "step": 13684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.885276378984665e-06,
+      "loss": 0.8291,
+      "step": 13685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.884602211385536e-06,
+      "loss": 0.834,
+      "step": 13686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.883928065125724e-06,
+      "loss": 0.9175,
+      "step": 13687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.883253940218129e-06,
+      "loss": 0.8667,
+      "step": 13688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.882579836675646e-06,
+      "loss": 0.936,
+      "step": 13689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.881905754511171e-06,
+      "loss": 0.8774,
+      "step": 13690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8812316937375996e-06,
+      "loss": 0.8916,
+      "step": 13691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8805576543678305e-06,
+      "loss": 0.8789,
+      "step": 13692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8798836364147576e-06,
+      "loss": 0.8735,
+      "step": 13693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8792096398912735e-06,
+      "loss": 0.8677,
+      "step": 13694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.878535664810277e-06,
+      "loss": 0.8647,
+      "step": 13695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.877861711184661e-06,
+      "loss": 0.8643,
+      "step": 13696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.877187779027317e-06,
+      "loss": 0.8813,
+      "step": 13697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.876513868351142e-06,
+      "loss": 0.9058,
+      "step": 13698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.875839979169026e-06,
+      "loss": 0.8579,
+      "step": 13699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.875166111493862e-06,
+      "loss": 0.8535,
+      "step": 13700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.874492265338544e-06,
+      "loss": 0.8052,
+      "step": 13701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873818440715962e-06,
+      "loss": 0.8267,
+      "step": 13702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873144637639007e-06,
+      "loss": 0.877,
+      "step": 13703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8724708561205725e-06,
+      "loss": 0.8242,
+      "step": 13704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8717970961735464e-06,
+      "loss": 0.7725,
+      "step": 13705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.871123357810821e-06,
+      "loss": 0.8018,
+      "step": 13706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.870449641045283e-06,
+      "loss": 0.8638,
+      "step": 13707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8697759458898234e-06,
+      "loss": 0.8833,
+      "step": 13708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.869102272357332e-06,
+      "loss": 0.876,
+      "step": 13709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.868428620460695e-06,
+      "loss": 0.8857,
+      "step": 13710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867754990212802e-06,
+      "loss": 0.896,
+      "step": 13711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867081381626541e-06,
+      "loss": 0.8501,
+      "step": 13712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.866407794714797e-06,
+      "loss": 0.8325,
+      "step": 13713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.86573422949046e-06,
+      "loss": 0.8589,
+      "step": 13714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.865060685966414e-06,
+      "loss": 0.8594,
+      "step": 13715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8643871641555434e-06,
+      "loss": 0.8008,
+      "step": 13716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863713664070739e-06,
+      "loss": 0.8306,
+      "step": 13717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863040185724881e-06,
+      "loss": 0.8555,
+      "step": 13718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.862366729130855e-06,
+      "loss": 0.7998,
+      "step": 13719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861693294301548e-06,
+      "loss": 0.8623,
+      "step": 13720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861019881249842e-06,
+      "loss": 0.8994,
+      "step": 13721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.860346489988621e-06,
+      "loss": 0.8208,
+      "step": 13722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.859673120530766e-06,
+      "loss": 0.8257,
+      "step": 13723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.858999772889161e-06,
+      "loss": 0.7898,
+      "step": 13724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.858326447076689e-06,
+      "loss": 0.3092,
+      "step": 13725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.857653143106231e-06,
+      "loss": 0.8501,
+      "step": 13726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.856979860990669e-06,
+      "loss": 0.8989,
+      "step": 13727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.856306600742883e-06,
+      "loss": 0.8745,
+      "step": 13728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.855633362375753e-06,
+      "loss": 0.8906,
+      "step": 13729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.854960145902162e-06,
+      "loss": 0.8491,
+      "step": 13730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.854286951334987e-06,
+      "loss": 0.748,
+      "step": 13731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.853613778687107e-06,
+      "loss": 0.8516,
+      "step": 13732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.852940627971403e-06,
+      "loss": 0.8267,
+      "step": 13733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.852267499200752e-06,
+      "loss": 0.897,
+      "step": 13734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.851594392388033e-06,
+      "loss": 0.8652,
+      "step": 13735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.850921307546121e-06,
+      "loss": 0.8984,
+      "step": 13736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.850248244687897e-06,
+      "loss": 0.8662,
+      "step": 13737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.849575203826236e-06,
+      "loss": 0.8804,
+      "step": 13738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8489021849740125e-06,
+      "loss": 0.8833,
+      "step": 13739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.848229188144105e-06,
+      "loss": 0.8032,
+      "step": 13740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.847556213349389e-06,
+      "loss": 0.8481,
+      "step": 13741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.846883260602737e-06,
+      "loss": 0.8101,
+      "step": 13742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.846210329917027e-06,
+      "loss": 0.8706,
+      "step": 13743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.845537421305131e-06,
+      "loss": 0.8442,
+      "step": 13744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844864534779921e-06,
+      "loss": 0.8818,
+      "step": 13745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844191670354276e-06,
+      "loss": 0.8594,
+      "step": 13746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.843518828041065e-06,
+      "loss": 0.8579,
+      "step": 13747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842846007853159e-06,
+      "loss": 0.8633,
+      "step": 13748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842173209803434e-06,
+      "loss": 0.8496,
+      "step": 13749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8415004339047605e-06,
+      "loss": 0.8008,
+      "step": 13750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.84082768017001e-06,
+      "loss": 0.8989,
+      "step": 13751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8401549486120505e-06,
+      "loss": 0.8291,
+      "step": 13752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.839482239243755e-06,
+      "loss": 0.8745,
+      "step": 13753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.838809552077994e-06,
+      "loss": 0.8564,
+      "step": 13754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8381368871276345e-06,
+      "loss": 0.8735,
+      "step": 13755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.83746424440555e-06,
+      "loss": 0.9131,
+      "step": 13756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.836791623924604e-06,
+      "loss": 0.8472,
+      "step": 13757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.836119025697667e-06,
+      "loss": 0.8652,
+      "step": 13758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835446449737608e-06,
+      "loss": 0.9253,
+      "step": 13759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.834773896057294e-06,
+      "loss": 0.8389,
+      "step": 13760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8341013646695894e-06,
+      "loss": 0.9062,
+      "step": 13761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8334288555873654e-06,
+      "loss": 0.8633,
+      "step": 13762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.832756368823485e-06,
+      "loss": 0.853,
+      "step": 13763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.832083904390815e-06,
+      "loss": 0.8481,
+      "step": 13764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8314114623022185e-06,
+      "loss": 0.8501,
+      "step": 13765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8307390425705635e-06,
+      "loss": 0.2998,
+      "step": 13766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.830066645208714e-06,
+      "loss": 0.8555,
+      "step": 13767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.829394270229531e-06,
+      "loss": 0.8003,
+      "step": 13768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.828721917645881e-06,
+      "loss": 0.8579,
+      "step": 13769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8280495874706285e-06,
+      "loss": 0.7476,
+      "step": 13770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.827377279716631e-06,
+      "loss": 0.8706,
+      "step": 13771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.826704994396756e-06,
+      "loss": 0.8286,
+      "step": 13772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.826032731523863e-06,
+      "loss": 0.875,
+      "step": 13773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.825360491110813e-06,
+      "loss": 0.8088,
+      "step": 13774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.82468827317047e-06,
+      "loss": 0.8325,
+      "step": 13775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.82401607771569e-06,
+      "loss": 0.8691,
+      "step": 13776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.823343904759335e-06,
+      "loss": 0.8223,
+      "step": 13777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8226717543142675e-06,
+      "loss": 0.8633,
+      "step": 13778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.821999626393344e-06,
+      "loss": 0.8501,
+      "step": 13779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8213275210094244e-06,
+      "loss": 0.8594,
+      "step": 13780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.820655438175365e-06,
+      "loss": 0.8516,
+      "step": 13781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.819983377904026e-06,
+      "loss": 0.7881,
+      "step": 13782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.819311340208266e-06,
+      "loss": 0.8452,
+      "step": 13783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.818639325100939e-06,
+      "loss": 0.8677,
+      "step": 13784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.817967332594903e-06,
+      "loss": 0.8779,
+      "step": 13785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.817295362703016e-06,
+      "loss": 0.7644,
+      "step": 13786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.816623415438131e-06,
+      "loss": 0.9209,
+      "step": 13787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815951490813107e-06,
+      "loss": 0.75,
+      "step": 13788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815279588840795e-06,
+      "loss": 0.8384,
+      "step": 13789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8146077095340507e-06,
+      "loss": 0.8579,
+      "step": 13790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.813935852905731e-06,
+      "loss": 0.8745,
+      "step": 13791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.813264018968685e-06,
+      "loss": 0.833,
+      "step": 13792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8125922077357726e-06,
+      "loss": 0.7891,
+      "step": 13793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8119204192198376e-06,
+      "loss": 0.8735,
+      "step": 13794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8112486534337383e-06,
+      "loss": 0.8315,
+      "step": 13795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8105769103903283e-06,
+      "loss": 0.8569,
+      "step": 13796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.809905190102453e-06,
+      "loss": 0.8765,
+      "step": 13797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8092334925829673e-06,
+      "loss": 0.8589,
+      "step": 13798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8085618178447232e-06,
+      "loss": 0.8809,
+      "step": 13799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8078901659005673e-06,
+      "loss": 0.895,
+      "step": 13800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.807218536763353e-06,
+      "loss": 0.8809,
+      "step": 13801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8065469304459252e-06,
+      "loss": 0.8242,
+      "step": 13802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8058753469611363e-06,
+      "loss": 0.8984,
+      "step": 13803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.805203786321835e-06,
+      "loss": 0.8164,
+      "step": 13804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8045322485408655e-06,
+      "loss": 0.8306,
+      "step": 13805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.803860733631079e-06,
+      "loss": 0.8691,
+      "step": 13806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.803189241605323e-06,
+      "loss": 0.8364,
+      "step": 13807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.80251777247644e-06,
+      "loss": 0.7861,
+      "step": 13808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.801846326257282e-06,
+      "loss": 0.8423,
+      "step": 13809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8011749029606903e-06,
+      "loss": 0.8794,
+      "step": 13810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.800503502599511e-06,
+      "loss": 0.8726,
+      "step": 13811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.799832125186591e-06,
+      "loss": 0.918,
+      "step": 13812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7991607707347737e-06,
+      "loss": 0.8613,
+      "step": 13813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7984894392569018e-06,
+      "loss": 0.792,
+      "step": 13814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7978181307658226e-06,
+      "loss": 0.7871,
+      "step": 13815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.797146845274375e-06,
+      "loss": 0.8433,
+      "step": 13816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.796475582795406e-06,
+      "loss": 0.8447,
+      "step": 13817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7958043433417523e-06,
+      "loss": 0.8325,
+      "step": 13818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7951331269262614e-06,
+      "loss": 0.8184,
+      "step": 13819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7944619335617727e-06,
+      "loss": 0.8662,
+      "step": 13820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7937907632611246e-06,
+      "loss": 0.8721,
+      "step": 13821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7931196160371625e-06,
+      "loss": 0.814,
+      "step": 13822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7924484919027225e-06,
+      "loss": 0.8467,
+      "step": 13823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7917773908706455e-06,
+      "loss": 0.3527,
+      "step": 13824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7911063129537723e-06,
+      "loss": 0.8193,
+      "step": 13825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7904352581649397e-06,
+      "loss": 0.8447,
+      "step": 13826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7897642265169854e-06,
+      "loss": 0.8091,
+      "step": 13827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.78909321802275e-06,
+      "loss": 0.7976,
+      "step": 13828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.788422232695069e-06,
+      "loss": 0.9263,
+      "step": 13829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.787751270546781e-06,
+      "loss": 0.8208,
+      "step": 13830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7870803315907195e-06,
+      "loss": 0.9028,
+      "step": 13831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7864094158397237e-06,
+      "loss": 0.833,
+      "step": 13832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.785738523306629e-06,
+      "loss": 0.8599,
+      "step": 13833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7850676540042675e-06,
+      "loss": 0.8799,
+      "step": 13834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7843968079454773e-06,
+      "loss": 0.3717,
+      "step": 13835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.783725985143093e-06,
+      "loss": 0.856,
+      "step": 13836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7830551856099452e-06,
+      "loss": 0.8755,
+      "step": 13837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7823844093588714e-06,
+      "loss": 0.8394,
+      "step": 13838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.781713656402701e-06,
+      "loss": 0.8911,
+      "step": 13839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.781042926754268e-06,
+      "loss": 0.875,
+      "step": 13840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7803722204264064e-06,
+      "loss": 0.8193,
+      "step": 13841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7797015374319453e-06,
+      "loss": 0.8862,
+      "step": 13842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.779030877783715e-06,
+      "loss": 0.8242,
+      "step": 13843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.77836024149455e-06,
+      "loss": 0.8979,
+      "step": 13844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7776896285772787e-06,
+      "loss": 0.8833,
+      "step": 13845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7770190390447315e-06,
+      "loss": 0.3105,
+      "step": 13846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7763484729097344e-06,
+      "loss": 0.814,
+      "step": 13847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.775677930185121e-06,
+      "loss": 0.897,
+      "step": 13848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.775007410883719e-06,
+      "loss": 0.8169,
+      "step": 13849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.774336915018354e-06,
+      "loss": 0.8687,
+      "step": 13850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7736664426018554e-06,
+      "loss": 0.8467,
+      "step": 13851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7729959936470517e-06,
+      "loss": 0.8721,
+      "step": 13852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.772325568166766e-06,
+      "loss": 0.897,
+      "step": 13853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.771655166173829e-06,
+      "loss": 0.8398,
+      "step": 13854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7709847876810627e-06,
+      "loss": 0.8896,
+      "step": 13855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7703144327012937e-06,
+      "loss": 0.8477,
+      "step": 13856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7696441012473494e-06,
+      "loss": 0.8008,
+      "step": 13857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.768973793332051e-06,
+      "loss": 0.8623,
+      "step": 13858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7683035089682247e-06,
+      "loss": 0.8496,
+      "step": 13859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7676332481686913e-06,
+      "loss": 0.9043,
+      "step": 13860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.766963010946277e-06,
+      "loss": 0.7856,
+      "step": 13861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.766292797313804e-06,
+      "loss": 0.3157,
+      "step": 13862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.765622607284092e-06,
+      "loss": 0.8145,
+      "step": 13863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7649524408699657e-06,
+      "loss": 0.8486,
+      "step": 13864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7642822980842464e-06,
+      "loss": 0.8647,
+      "step": 13865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.763612178939752e-06,
+      "loss": 0.8242,
+      "step": 13866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.762942083449307e-06,
+      "loss": 0.897,
+      "step": 13867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7622720116257276e-06,
+      "loss": 0.873,
+      "step": 13868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7616019634818344e-06,
+      "loss": 0.8872,
+      "step": 13869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7609319390304498e-06,
+      "loss": 0.8574,
+      "step": 13870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.760261938284388e-06,
+      "loss": 0.8418,
+      "step": 13871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.759591961256468e-06,
+      "loss": 0.8462,
+      "step": 13872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7589220079595113e-06,
+      "loss": 0.855,
+      "step": 13873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7582520784063303e-06,
+      "loss": 0.8486,
+      "step": 13874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.757582172609745e-06,
+      "loss": 0.8579,
+      "step": 13875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7569122905825684e-06,
+      "loss": 0.8413,
+      "step": 13876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.75624243233762e-06,
+      "loss": 0.8296,
+      "step": 13877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7555725978877144e-06,
+      "loss": 0.8267,
+      "step": 13878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.754902787245664e-06,
+      "loss": 0.8335,
+      "step": 13879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7542330004242867e-06,
+      "loss": 0.874,
+      "step": 13880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7535632374363955e-06,
+      "loss": 0.8301,
+      "step": 13881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7528934982948022e-06,
+      "loss": 0.8076,
+      "step": 13882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7522237830123237e-06,
+      "loss": 0.8804,
+      "step": 13883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.751554091601769e-06,
+      "loss": 0.8726,
+      "step": 13884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750884424075951e-06,
+      "loss": 0.8457,
+      "step": 13885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750214780447684e-06,
+      "loss": 0.8447,
+      "step": 13886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.749545160729777e-06,
+      "loss": 0.8643,
+      "step": 13887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7488755649350427e-06,
+      "loss": 0.8569,
+      "step": 13888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.748205993076288e-06,
+      "loss": 0.8096,
+      "step": 13889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7475364451663264e-06,
+      "loss": 0.896,
+      "step": 13890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7468669212179677e-06,
+      "loss": 0.8008,
+      "step": 13891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.746197421244017e-06,
+      "loss": 0.8706,
+      "step": 13892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7455279452572867e-06,
+      "loss": 0.9092,
+      "step": 13893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7448584932705843e-06,
+      "loss": 0.8789,
+      "step": 13894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.744189065296715e-06,
+      "loss": 0.917,
+      "step": 13895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.743519661348489e-06,
+      "loss": 0.811,
+      "step": 13896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.742850281438712e-06,
+      "loss": 0.8447,
+      "step": 13897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.742180925580189e-06,
+      "loss": 0.8032,
+      "step": 13898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.741511593785728e-06,
+      "loss": 0.8999,
+      "step": 13899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7408422860681333e-06,
+      "loss": 0.8779,
+      "step": 13900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7401730024402084e-06,
+      "loss": 0.8564,
+      "step": 13901
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7395037429147615e-06,
+      "loss": 0.8511,
+      "step": 13902
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.738834507504592e-06,
+      "loss": 0.8564,
+      "step": 13903
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.738165296222509e-06,
+      "loss": 0.7935,
+      "step": 13904
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7374961090813088e-06,
+      "loss": 0.8618,
+      "step": 13905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.736826946093798e-06,
+      "loss": 0.7993,
+      "step": 13906
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7361578072727813e-06,
+      "loss": 0.9131,
+      "step": 13907
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.735488692631053e-06,
+      "loss": 0.8296,
+      "step": 13908
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7348196021814196e-06,
+      "loss": 0.8716,
+      "step": 13909
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7341505359366838e-06,
+      "loss": 0.9312,
+      "step": 13910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.733481493909641e-06,
+      "loss": 0.8682,
+      "step": 13911
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7328124761130933e-06,
+      "loss": 0.895,
+      "step": 13912
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.732143482559839e-06,
+      "loss": 0.2827,
+      "step": 13913
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7314745132626784e-06,
+      "loss": 0.9038,
+      "step": 13914
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7308055682344102e-06,
+      "loss": 0.8218,
+      "step": 13915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7301366474878297e-06,
+      "loss": 0.8516,
+      "step": 13916
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.729467751035737e-06,
+      "loss": 0.8394,
+      "step": 13917
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7287988788909297e-06,
+      "loss": 0.7993,
+      "step": 13918
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7281300310662004e-06,
+      "loss": 0.8921,
+      "step": 13919
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.72746120757435e-06,
+      "loss": 0.8945,
+      "step": 13920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7267924084281714e-06,
+      "loss": 0.8555,
+      "step": 13921
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7261236336404593e-06,
+      "loss": 0.8843,
+      "step": 13922
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.725454883224011e-06,
+      "loss": 0.8521,
+      "step": 13923
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7247861571916183e-06,
+      "loss": 0.8374,
+      "step": 13924
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.724117455556077e-06,
+      "loss": 0.8999,
+      "step": 13925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7234487783301776e-06,
+      "loss": 0.8467,
+      "step": 13926
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7227801255267155e-06,
+      "loss": 0.8701,
+      "step": 13927
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.722111497158483e-06,
+      "loss": 0.8574,
+      "step": 13928
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7214428932382696e-06,
+      "loss": 0.8765,
+      "step": 13929
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7207743137788687e-06,
+      "loss": 0.8613,
+      "step": 13930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.720105758793073e-06,
+      "loss": 0.8838,
+      "step": 13931
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7194372282936687e-06,
+      "loss": 0.8281,
+      "step": 13932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.71876872229345e-06,
+      "loss": 0.7637,
+      "step": 13933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7181002408052035e-06,
+      "loss": 0.8896,
+      "step": 13934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7174317838417186e-06,
+      "loss": 0.8374,
+      "step": 13935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7167633514157863e-06,
+      "loss": 0.3612,
+      "step": 13936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7160949435401926e-06,
+      "loss": 0.7861,
+      "step": 13937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7154265602277246e-06,
+      "loss": 0.896,
+      "step": 13938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.714758201491173e-06,
+      "loss": 0.8765,
+      "step": 13939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.714089867343321e-06,
+      "loss": 0.8335,
+      "step": 13940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.713421557796958e-06,
+      "loss": 0.8472,
+      "step": 13941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7127532728648653e-06,
+      "loss": 0.8618,
+      "step": 13942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.712085012559832e-06,
+      "loss": 0.9062,
+      "step": 13943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.711416776894643e-06,
+      "loss": 0.8496,
+      "step": 13944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.710748565882079e-06,
+      "loss": 0.8486,
+      "step": 13945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7100803795349283e-06,
+      "loss": 0.8262,
+      "step": 13946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.709412217865973e-06,
+      "loss": 0.3082,
+      "step": 13947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708744080887994e-06,
+      "loss": 0.8149,
+      "step": 13948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708075968613778e-06,
+      "loss": 0.9058,
+      "step": 13949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.707407881056103e-06,
+      "loss": 0.8779,
+      "step": 13950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7067398182277513e-06,
+      "loss": 0.8833,
+      "step": 13951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.706071780141507e-06,
+      "loss": 0.8613,
+      "step": 13952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7054037668101472e-06,
+      "loss": 0.8711,
+      "step": 13953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.704735778246452e-06,
+      "loss": 0.9009,
+      "step": 13954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.704067814463205e-06,
+      "loss": 0.8496,
+      "step": 13955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7033998754731815e-06,
+      "loss": 0.8774,
+      "step": 13956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7027319612891634e-06,
+      "loss": 0.9072,
+      "step": 13957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7020640719239244e-06,
+      "loss": 0.8521,
+      "step": 13958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7013962073902456e-06,
+      "loss": 0.8047,
+      "step": 13959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7007283677009054e-06,
+      "loss": 0.8525,
+      "step": 13960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7000605528686763e-06,
+      "loss": 0.9009,
+      "step": 13961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6993927629063395e-06,
+      "loss": 0.7935,
+      "step": 13962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6987249978266684e-06,
+      "loss": 0.8623,
+      "step": 13963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.698057257642437e-06,
+      "loss": 0.7192,
+      "step": 13964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6973895423664243e-06,
+      "loss": 0.8516,
+      "step": 13965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.696721852011401e-06,
+      "loss": 0.7959,
+      "step": 13966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.696054186590141e-06,
+      "loss": 0.8545,
+      "step": 13967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6953865461154213e-06,
+      "loss": 0.8999,
+      "step": 13968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.694718930600012e-06,
+      "loss": 0.8667,
+      "step": 13969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.694051340056688e-06,
+      "loss": 0.812,
+      "step": 13970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.693383774498217e-06,
+      "loss": 0.8276,
+      "step": 13971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6927162339373747e-06,
+      "loss": 0.8169,
+      "step": 13972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6920487183869313e-06,
+      "loss": 0.8892,
+      "step": 13973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6913812278596556e-06,
+      "loss": 0.9033,
+      "step": 13974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.69071376236832e-06,
+      "loss": 0.7168,
+      "step": 13975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6900463219256944e-06,
+      "loss": 0.3669,
+      "step": 13976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6893789065445445e-06,
+      "loss": 0.9141,
+      "step": 13977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6887115162376438e-06,
+      "loss": 0.8848,
+      "step": 13978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6880441510177566e-06,
+      "loss": 0.8516,
+      "step": 13979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6873768108976516e-06,
+      "loss": 0.8823,
+      "step": 13980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.686709495890099e-06,
+      "loss": 0.8779,
+      "step": 13981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6860422060078617e-06,
+      "loss": 0.8389,
+      "step": 13982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6853749412637063e-06,
+      "loss": 0.8281,
+      "step": 13983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6847077016704024e-06,
+      "loss": 0.8745,
+      "step": 13984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.684040487240712e-06,
+      "loss": 0.7896,
+      "step": 13985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6833732979874013e-06,
+      "loss": 0.2437,
+      "step": 13986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6827061339232325e-06,
+      "loss": 0.8545,
+      "step": 13987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6820389950609713e-06,
+      "loss": 0.875,
+      "step": 13988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.681371881413383e-06,
+      "loss": 0.8794,
+      "step": 13989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.680704792993226e-06,
+      "loss": 0.8687,
+      "step": 13990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.680037729813267e-06,
+      "loss": 0.7715,
+      "step": 13991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6793706918862658e-06,
+      "loss": 0.7993,
+      "step": 13992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6787036792249837e-06,
+      "loss": 0.9033,
+      "step": 13993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.678036691842184e-06,
+      "loss": 0.8926,
+      "step": 13994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.677369729750624e-06,
+      "loss": 0.876,
+      "step": 13995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6767027929630644e-06,
+      "loss": 0.8818,
+      "step": 13996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.676035881492268e-06,
+      "loss": 0.8354,
+      "step": 13997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6753689953509904e-06,
+      "loss": 0.7871,
+      "step": 13998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.674702134551992e-06,
+      "loss": 0.8442,
+      "step": 13999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6740352991080286e-06,
+      "loss": 0.8604,
+      "step": 14000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.67336848903186e-06,
+      "loss": 0.9165,
+      "step": 14001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6727017043362435e-06,
+      "loss": 0.9136,
+      "step": 14002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.672034945033933e-06,
+      "loss": 0.8374,
+      "step": 14003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.671368211137688e-06,
+      "loss": 0.8545,
+      "step": 14004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6707015026602643e-06,
+      "loss": 0.8506,
+      "step": 14005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.670034819614413e-06,
+      "loss": 0.8096,
+      "step": 14006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.669368162012893e-06,
+      "loss": 0.8242,
+      "step": 14007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6687015298684566e-06,
+      "loss": 0.9048,
+      "step": 14008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6680349231938562e-06,
+      "loss": 0.8047,
+      "step": 14009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.667368342001849e-06,
+      "loss": 0.8018,
+      "step": 14010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.666701786305184e-06,
+      "loss": 0.9004,
+      "step": 14011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.666035256116614e-06,
+      "loss": 0.8042,
+      "step": 14012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.665368751448894e-06,
+      "loss": 0.833,
+      "step": 14013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664702272314772e-06,
+      "loss": 0.8486,
+      "step": 14014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6640358187269996e-06,
+      "loss": 0.9238,
+      "step": 14015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.663369390698326e-06,
+      "loss": 0.7998,
+      "step": 14016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.662702988241502e-06,
+      "loss": 0.8677,
+      "step": 14017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6620366113692807e-06,
+      "loss": 0.8462,
+      "step": 14018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6613702600944034e-06,
+      "loss": 0.8472,
+      "step": 14019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6607039344296236e-06,
+      "loss": 0.835,
+      "step": 14020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.660037634387689e-06,
+      "loss": 0.7896,
+      "step": 14021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6593713599813453e-06,
+      "loss": 0.8408,
+      "step": 14022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6587051112233416e-06,
+      "loss": 0.9043,
+      "step": 14023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.65803888812642e-06,
+      "loss": 0.8628,
+      "step": 14024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.65737269070333e-06,
+      "loss": 0.8677,
+      "step": 14025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6567065189668184e-06,
+      "loss": 0.8716,
+      "step": 14026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6560403729296247e-06,
+      "loss": 0.8208,
+      "step": 14027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.655374252604499e-06,
+      "loss": 0.8159,
+      "step": 14028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.654708158004181e-06,
+      "loss": 0.7603,
+      "step": 14029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6540420891414157e-06,
+      "loss": 0.8218,
+      "step": 14030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.653376046028948e-06,
+      "loss": 0.8545,
+      "step": 14031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.652710028679517e-06,
+      "loss": 0.832,
+      "step": 14032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6520440371058663e-06,
+      "loss": 0.8672,
+      "step": 14033
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.651378071320738e-06,
+      "loss": 0.8975,
+      "step": 14034
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.650712131336872e-06,
+      "loss": 0.8398,
+      "step": 14035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6500462171670104e-06,
+      "loss": 0.8813,
+      "step": 14036
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6493803288238894e-06,
+      "loss": 0.814,
+      "step": 14037
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6487144663202522e-06,
+      "loss": 0.8237,
+      "step": 14038
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6480486296688377e-06,
+      "loss": 0.8994,
+      "step": 14039
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.647382818882382e-06,
+      "loss": 0.8711,
+      "step": 14040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6467170339736246e-06,
+      "loss": 0.8774,
+      "step": 14041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.646051274955304e-06,
+      "loss": 0.8535,
+      "step": 14042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.645385541840154e-06,
+      "loss": 0.8887,
+      "step": 14043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6447198346409153e-06,
+      "loss": 0.8628,
+      "step": 14044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.644054153370321e-06,
+      "loss": 0.2986,
+      "step": 14045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.643388498041106e-06,
+      "loss": 0.8696,
+      "step": 14046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6427228686660087e-06,
+      "loss": 0.8335,
+      "step": 14047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.642057265257761e-06,
+      "loss": 0.8594,
+      "step": 14048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6413916878290966e-06,
+      "loss": 0.873,
+      "step": 14049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6407261363927527e-06,
+      "loss": 0.8501,
+      "step": 14050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6400606109614584e-06,
+      "loss": 0.8896,
+      "step": 14051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6393951115479486e-06,
+      "loss": 0.8467,
+      "step": 14052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6387296381649518e-06,
+      "loss": 0.8423,
+      "step": 14053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6380641908252035e-06,
+      "loss": 0.8428,
+      "step": 14054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6373987695414347e-06,
+      "loss": 0.8208,
+      "step": 14055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6367333743263724e-06,
+      "loss": 0.8638,
+      "step": 14056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.636068005192751e-06,
+      "loss": 0.8877,
+      "step": 14057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6354026621532965e-06,
+      "loss": 0.8882,
+      "step": 14058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.634737345220739e-06,
+      "loss": 0.8594,
+      "step": 14059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.634072054407809e-06,
+      "loss": 0.8779,
+      "step": 14060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.633406789727233e-06,
+      "loss": 0.8604,
+      "step": 14061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6327415511917365e-06,
+      "loss": 0.8564,
+      "step": 14062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.632076338814051e-06,
+      "loss": 0.9272,
+      "step": 14063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6314111526068994e-06,
+      "loss": 0.8208,
+      "step": 14064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.630745992583011e-06,
+      "loss": 0.8188,
+      "step": 14065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.630080858755106e-06,
+      "loss": 0.8149,
+      "step": 14066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6294157511359153e-06,
+      "loss": 0.8823,
+      "step": 14067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6287506697381612e-06,
+      "loss": 0.7949,
+      "step": 14068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.628085614574566e-06,
+      "loss": 0.8838,
+      "step": 14069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.627420585657856e-06,
+      "loss": 0.8779,
+      "step": 14070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.626755583000754e-06,
+      "loss": 0.8921,
+      "step": 14071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6260906066159796e-06,
+      "loss": 0.7939,
+      "step": 14072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.625425656516258e-06,
+      "loss": 0.8384,
+      "step": 14073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6247607327143095e-06,
+      "loss": 0.8413,
+      "step": 14074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.624095835222855e-06,
+      "loss": 0.8779,
+      "step": 14075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6234309640546163e-06,
+      "loss": 0.3616,
+      "step": 14076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6227661192223117e-06,
+      "loss": 0.8647,
+      "step": 14077
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6221013007386607e-06,
+      "loss": 0.8579,
+      "step": 14078
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.621436508616385e-06,
+      "loss": 0.9268,
+      "step": 14079
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6207717428681998e-06,
+      "loss": 0.8481,
+      "step": 14080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6201070035068264e-06,
+      "loss": 0.8735,
+      "step": 14081
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.619442290544978e-06,
+      "loss": 0.8169,
+      "step": 14082
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.618777603995375e-06,
+      "loss": 0.8481,
+      "step": 14083
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.618112943870734e-06,
+      "loss": 0.9199,
+      "step": 14084
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6174483101837676e-06,
+      "loss": 0.8521,
+      "step": 14085
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6167837029471952e-06,
+      "loss": 0.8584,
+      "step": 14086
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6161191221737312e-06,
+      "loss": 0.8823,
+      "step": 14087
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.615454567876087e-06,
+      "loss": 0.8193,
+      "step": 14088
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6147900400669812e-06,
+      "loss": 0.8657,
+      "step": 14089
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6141255387591236e-06,
+      "loss": 0.8965,
+      "step": 14090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6134610639652268e-06,
+      "loss": 0.8833,
+      "step": 14091
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6127966156980065e-06,
+      "loss": 0.835,
+      "step": 14092
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6121321939701724e-06,
+      "loss": 0.7812,
+      "step": 14093
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.611467798794438e-06,
+      "loss": 0.7189,
+      "step": 14094
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6108034301835094e-06,
+      "loss": 0.8677,
+      "step": 14095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6101390881501017e-06,
+      "loss": 0.8301,
+      "step": 14096
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.609474772706925e-06,
+      "loss": 0.877,
+      "step": 14097
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.608810483866685e-06,
+      "loss": 0.855,
+      "step": 14098
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.608146221642093e-06,
+      "loss": 0.8647,
+      "step": 14099
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6074819860458586e-06,
+      "loss": 0.8164,
+      "step": 14100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6068177770906867e-06,
+      "loss": 0.7784,
+      "step": 14101
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.606153594789287e-06,
+      "loss": 0.8481,
+      "step": 14102
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.605489439154365e-06,
+      "loss": 0.8691,
+      "step": 14103
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.604825310198626e-06,
+      "loss": 0.8911,
+      "step": 14104
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6041612079347798e-06,
+      "loss": 0.8628,
+      "step": 14105
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6034971323755287e-06,
+      "loss": 0.853,
+      "step": 14106
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.602833083533576e-06,
+      "loss": 0.8438,
+      "step": 14107
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.602169061421631e-06,
+      "loss": 0.8906,
+      "step": 14108
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6015050660523933e-06,
+      "loss": 0.8652,
+      "step": 14109
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.600841097438568e-06,
+      "loss": 0.7993,
+      "step": 14110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6001771555928554e-06,
+      "loss": 0.8501,
+      "step": 14111
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.599513240527961e-06,
+      "loss": 0.8882,
+      "step": 14112
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5988493522565867e-06,
+      "loss": 0.9077,
+      "step": 14113
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5981854907914293e-06,
+      "loss": 0.8418,
+      "step": 14114
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.597521656145194e-06,
+      "loss": 0.8618,
+      "step": 14115
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5968578483305807e-06,
+      "loss": 0.8359,
+      "step": 14116
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5961940673602848e-06,
+      "loss": 0.8359,
+      "step": 14117
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5955303132470108e-06,
+      "loss": 0.8394,
+      "step": 14118
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.594866586003455e-06,
+      "loss": 0.7603,
+      "step": 14119
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.594202885642314e-06,
+      "loss": 0.8511,
+      "step": 14120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5935392121762887e-06,
+      "loss": 0.8281,
+      "step": 14121
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5928755656180737e-06,
+      "loss": 0.8765,
+      "step": 14122
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.592211945980368e-06,
+      "loss": 0.8262,
+      "step": 14123
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.591548353275864e-06,
+      "loss": 0.8237,
+      "step": 14124
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5908847875172603e-06,
+      "loss": 0.873,
+      "step": 14125
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5902212487172518e-06,
+      "loss": 0.8496,
+      "step": 14126
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5895577368885303e-06,
+      "loss": 0.8291,
+      "step": 14127
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.588894252043793e-06,
+      "loss": 0.8608,
+      "step": 14128
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.588230794195734e-06,
+      "loss": 0.876,
+      "step": 14129
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5875673633570422e-06,
+      "loss": 0.9092,
+      "step": 14130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.586903959540414e-06,
+      "loss": 0.8062,
+      "step": 14131
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.586240582758539e-06,
+      "loss": 0.8364,
+      "step": 14132
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.585577233024109e-06,
+      "loss": 0.8286,
+      "step": 14133
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5849139103498174e-06,
+      "loss": 0.79,
+      "step": 14134
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.58425061474835e-06,
+      "loss": 0.8452,
+      "step": 14135
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.583587346232401e-06,
+      "loss": 0.8389,
+      "step": 14136
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5829241048146583e-06,
+      "loss": 0.8286,
+      "step": 14137
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5822608905078087e-06,
+      "loss": 0.8862,
+      "step": 14138
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.581597703324544e-06,
+      "loss": 0.3005,
+      "step": 14139
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5809345432775498e-06,
+      "loss": 0.8457,
+      "step": 14140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5802714103795132e-06,
+      "loss": 0.8667,
+      "step": 14141
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5796083046431242e-06,
+      "loss": 0.8525,
+      "step": 14142
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.578945226081064e-06,
+      "loss": 0.8267,
+      "step": 14143
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5782821747060215e-06,
+      "loss": 0.8823,
+      "step": 14144
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5776191505306823e-06,
+      "loss": 0.8188,
+      "step": 14145
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5769561535677292e-06,
+      "loss": 0.8193,
+      "step": 14146
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5762931838298487e-06,
+      "loss": 0.8149,
+      "step": 14147
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.575630241329721e-06,
+      "loss": 0.8145,
+      "step": 14148
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5749673260800323e-06,
+      "loss": 0.3217,
+      "step": 14149
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5743044380934655e-06,
+      "loss": 0.8491,
+      "step": 14150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.573641577382699e-06,
+      "loss": 0.8726,
+      "step": 14151
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.572978743960418e-06,
+      "loss": 0.8447,
+      "step": 14152
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5723159378393035e-06,
+      "loss": 0.8726,
+      "step": 14153
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5716531590320323e-06,
+      "loss": 0.8452,
+      "step": 14154
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.570990407551289e-06,
+      "loss": 0.8623,
+      "step": 14155
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5703276834097506e-06,
+      "loss": 0.8877,
+      "step": 14156
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.569664986620095e-06,
+      "loss": 0.7676,
+      "step": 14157
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5690023171950047e-06,
+      "loss": 0.3381,
+      "step": 14158
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.568339675147153e-06,
+      "loss": 0.8374,
+      "step": 14159
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.567677060489222e-06,
+      "loss": 0.8672,
+      "step": 14160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5670144732338825e-06,
+      "loss": 0.8364,
+      "step": 14161
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.566351913393815e-06,
+      "loss": 0.8154,
+      "step": 14162
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.565689380981697e-06,
+      "loss": 0.8105,
+      "step": 14163
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.565026876010198e-06,
+      "loss": 0.8379,
+      "step": 14164
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5643643984919978e-06,
+      "loss": 0.8125,
+      "step": 14165
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.56370194843977e-06,
+      "loss": 0.8916,
+      "step": 14166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.563039525866185e-06,
+      "loss": 0.8677,
+      "step": 14167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5623771307839208e-06,
+      "loss": 0.8921,
+      "step": 14168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.561714763205645e-06,
+      "loss": 0.8682,
+      "step": 14169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5610524231440324e-06,
+      "loss": 0.8784,
+      "step": 14170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5603901106117566e-06,
+      "loss": 0.8354,
+      "step": 14171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.559727825621485e-06,
+      "loss": 0.833,
+      "step": 14172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.559065568185889e-06,
+      "loss": 0.8369,
+      "step": 14173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.558403338317641e-06,
+      "loss": 0.8784,
+      "step": 14174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5577411360294077e-06,
+      "loss": 0.793,
+      "step": 14175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5570789613338606e-06,
+      "loss": 0.7393,
+      "step": 14176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5564168142436644e-06,
+      "loss": 0.8813,
+      "step": 14177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5557546947714904e-06,
+      "loss": 0.8931,
+      "step": 14178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5550926029300063e-06,
+      "loss": 0.8037,
+      "step": 14179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5544305387318757e-06,
+      "loss": 0.8418,
+      "step": 14180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.553768502189767e-06,
+      "loss": 0.8418,
+      "step": 14181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5531064933163474e-06,
+      "loss": 0.8149,
+      "step": 14182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.552444512124279e-06,
+      "loss": 0.8047,
+      "step": 14183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.551782558626229e-06,
+      "loss": 0.771,
+      "step": 14184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5511206328348607e-06,
+      "loss": 0.8447,
+      "step": 14185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5504587347628368e-06,
+      "loss": 0.8989,
+      "step": 14186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.549796864422823e-06,
+      "loss": 0.8457,
+      "step": 14187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.54913502182748e-06,
+      "loss": 0.8472,
+      "step": 14188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.548473206989469e-06,
+      "loss": 0.8887,
+      "step": 14189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5478114199214547e-06,
+      "loss": 0.812,
+      "step": 14190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5471496606360955e-06,
+      "loss": 0.8447,
+      "step": 14191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5464879291460535e-06,
+      "loss": 0.8535,
+      "step": 14192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.545826225463986e-06,
+      "loss": 0.8843,
+      "step": 14193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5451645496025555e-06,
+      "loss": 0.873,
+      "step": 14194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5445029015744204e-06,
+      "loss": 0.8071,
+      "step": 14195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5438412813922364e-06,
+      "loss": 0.8604,
+      "step": 14196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5431796890686653e-06,
+      "loss": 0.876,
+      "step": 14197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5425181246163622e-06,
+      "loss": 0.8716,
+      "step": 14198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5418565880479823e-06,
+      "loss": 0.8594,
+      "step": 14199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5411950793761852e-06,
+      "loss": 0.9126,
+      "step": 14200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5405335986136245e-06,
+      "loss": 0.8472,
+      "step": 14201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.539872145772955e-06,
+      "loss": 0.8838,
+      "step": 14202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5392107208668347e-06,
+      "loss": 0.8804,
+      "step": 14203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.538549323907914e-06,
+      "loss": 0.8579,
+      "step": 14204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.537887954908848e-06,
+      "loss": 0.8145,
+      "step": 14205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5372266138822874e-06,
+      "loss": 0.8491,
+      "step": 14206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5365653008408886e-06,
+      "loss": 0.853,
+      "step": 14207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5359040157973025e-06,
+      "loss": 0.3423,
+      "step": 14208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.535242758764177e-06,
+      "loss": 0.8315,
+      "step": 14209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5345815297541675e-06,
+      "loss": 0.8906,
+      "step": 14210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5339203287799227e-06,
+      "loss": 0.8589,
+      "step": 14211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5332591558540907e-06,
+      "loss": 0.8179,
+      "step": 14212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5325980109893244e-06,
+      "loss": 0.8657,
+      "step": 14213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5319368941982695e-06,
+      "loss": 0.8672,
+      "step": 14214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5312758054935743e-06,
+      "loss": 0.855,
+      "step": 14215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.530614744887889e-06,
+      "loss": 0.8633,
+      "step": 14216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5299537123938588e-06,
+      "loss": 0.8696,
+      "step": 14217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.529292708024129e-06,
+      "loss": 0.8086,
+      "step": 14218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5286317317913486e-06,
+      "loss": 0.8872,
+      "step": 14219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5279707837081623e-06,
+      "loss": 0.8779,
+      "step": 14220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5273098637872146e-06,
+      "loss": 0.8506,
+      "step": 14221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.526648972041148e-06,
+      "loss": 0.8618,
+      "step": 14222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.525988108482609e-06,
+      "loss": 0.8071,
+      "step": 14223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5253272731242416e-06,
+      "loss": 0.8916,
+      "step": 14224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5246664659786857e-06,
+      "loss": 0.8457,
+      "step": 14225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.524005687058586e-06,
+      "loss": 0.8472,
+      "step": 14226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.523344936376584e-06,
+      "loss": 0.835,
+      "step": 14227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5226842139453175e-06,
+      "loss": 0.8721,
+      "step": 14228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5220235197774333e-06,
+      "loss": 0.8594,
+      "step": 14229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.521362853885566e-06,
+      "loss": 0.8486,
+      "step": 14230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5207022162823568e-06,
+      "loss": 0.8218,
+      "step": 14231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5200416069804475e-06,
+      "loss": 0.8506,
+      "step": 14232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519381025992474e-06,
+      "loss": 0.8506,
+      "step": 14233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5187204733310746e-06,
+      "loss": 0.8789,
+      "step": 14234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.518059949008885e-06,
+      "loss": 0.7988,
+      "step": 14235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.517399453038545e-06,
+      "loss": 0.8877,
+      "step": 14236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5167389854326907e-06,
+      "loss": 0.8315,
+      "step": 14237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.516078546203955e-06,
+      "loss": 0.8081,
+      "step": 14238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.515418135364975e-06,
+      "loss": 0.8486,
+      "step": 14239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.514757752928388e-06,
+      "loss": 0.3569,
+      "step": 14240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5140973989068238e-06,
+      "loss": 0.8564,
+      "step": 14241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.513437073312919e-06,
+      "loss": 0.3755,
+      "step": 14242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5127767761593033e-06,
+      "loss": 0.811,
+      "step": 14243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5121165074586134e-06,
+      "loss": 0.8945,
+      "step": 14244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.51145626722348e-06,
+      "loss": 0.8154,
+      "step": 14245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.510796055466531e-06,
+      "loss": 0.8677,
+      "step": 14246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.510135872200402e-06,
+      "loss": 0.9214,
+      "step": 14247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5094757174377224e-06,
+      "loss": 0.8813,
+      "step": 14248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.508815591191119e-06,
+      "loss": 0.8608,
+      "step": 14249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5081554934732254e-06,
+      "loss": 0.8301,
+      "step": 14250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5074954242966663e-06,
+      "loss": 0.8354,
+      "step": 14251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5068353836740715e-06,
+      "loss": 0.8823,
+      "step": 14252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5061753716180703e-06,
+      "loss": 0.9038,
+      "step": 14253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5055153881412873e-06,
+      "loss": 0.8018,
+      "step": 14254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5048554332563487e-06,
+      "loss": 0.8887,
+      "step": 14255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5041955069758843e-06,
+      "loss": 0.8428,
+      "step": 14256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5035356093125157e-06,
+      "loss": 0.8843,
+      "step": 14257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.50287574027887e-06,
+      "loss": 0.8384,
+      "step": 14258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5022158998875684e-06,
+      "loss": 0.8599,
+      "step": 14259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5015560881512378e-06,
+      "loss": 0.835,
+      "step": 14260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5008963050825017e-06,
+      "loss": 0.9346,
+      "step": 14261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5002365506939794e-06,
+      "loss": 0.8081,
+      "step": 14262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4995768249982975e-06,
+      "loss": 0.8462,
+      "step": 14263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4989171280080743e-06,
+      "loss": 0.8506,
+      "step": 14264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49825745973593e-06,
+      "loss": 0.8613,
+      "step": 14265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4975978201944905e-06,
+      "loss": 0.833,
+      "step": 14266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49693820939637e-06,
+      "loss": 0.895,
+      "step": 14267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4962786273541893e-06,
+      "loss": 0.8296,
+      "step": 14268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49561907408057e-06,
+      "loss": 0.8481,
+      "step": 14269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4949595495881277e-06,
+      "loss": 0.8037,
+      "step": 14270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4943000538894822e-06,
+      "loss": 0.8623,
+      "step": 14271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4936405869972466e-06,
+      "loss": 0.8818,
+      "step": 14272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.492981148924042e-06,
+      "loss": 0.8457,
+      "step": 14273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4923217396824826e-06,
+      "loss": 0.832,
+      "step": 14274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4916623592851828e-06,
+      "loss": 0.8755,
+      "step": 14275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4910030077447606e-06,
+      "loss": 0.7979,
+      "step": 14276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4903436850738293e-06,
+      "loss": 0.8555,
+      "step": 14277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.489684391285e-06,
+      "loss": 0.8447,
+      "step": 14278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4890251263908903e-06,
+      "loss": 0.8315,
+      "step": 14279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4883658904041097e-06,
+      "loss": 0.3131,
+      "step": 14280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4877066833372706e-06,
+      "loss": 0.9009,
+      "step": 14281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4870475052029886e-06,
+      "loss": 0.8237,
+      "step": 14282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.48638835601387e-06,
+      "loss": 0.8237,
+      "step": 14283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.485729235782527e-06,
+      "loss": 0.8486,
+      "step": 14284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.485070144521572e-06,
+      "loss": 0.8994,
+      "step": 14285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4844110822436107e-06,
+      "loss": 0.8428,
+      "step": 14286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4837520489612546e-06,
+      "loss": 0.9028,
+      "step": 14287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4830930446871097e-06,
+      "loss": 0.8291,
+      "step": 14288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4824340694337856e-06,
+      "loss": 0.8232,
+      "step": 14289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4817751232138906e-06,
+      "loss": 0.7811,
+      "step": 14290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4811162060400284e-06,
+      "loss": 0.8696,
+      "step": 14291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.480457317924808e-06,
+      "loss": 0.8857,
+      "step": 14292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.479798458880832e-06,
+      "loss": 0.8618,
+      "step": 14293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.479139628920707e-06,
+      "loss": 0.8623,
+      "step": 14294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4784808280570393e-06,
+      "loss": 0.7944,
+      "step": 14295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.47782205630243e-06,
+      "loss": 0.8179,
+      "step": 14296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4771633136694823e-06,
+      "loss": 0.7822,
+      "step": 14297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.476504600170802e-06,
+      "loss": 0.8477,
+      "step": 14298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.475845915818988e-06,
+      "loss": 0.8638,
+      "step": 14299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4751872606266453e-06,
+      "loss": 0.876,
+      "step": 14300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4745286346063704e-06,
+      "loss": 0.8086,
+      "step": 14301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4738700377707677e-06,
+      "loss": 0.8374,
+      "step": 14302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4732114701324378e-06,
+      "loss": 0.8418,
+      "step": 14303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.472552931703975e-06,
+      "loss": 0.8096,
+      "step": 14304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.471894422497984e-06,
+      "loss": 0.8857,
+      "step": 14305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4712359425270612e-06,
+      "loss": 0.8169,
+      "step": 14306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4705774918038017e-06,
+      "loss": 0.8091,
+      "step": 14307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.469919070340807e-06,
+      "loss": 0.8345,
+      "step": 14308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4692606781506705e-06,
+      "loss": 0.9253,
+      "step": 14309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.468602315245988e-06,
+      "loss": 0.8364,
+      "step": 14310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.467943981639359e-06,
+      "loss": 0.9023,
+      "step": 14311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.467285677343375e-06,
+      "loss": 0.8994,
+      "step": 14312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.466627402370629e-06,
+      "loss": 0.8169,
+      "step": 14313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4659691567337205e-06,
+      "loss": 0.8398,
+      "step": 14314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4653109404452378e-06,
+      "loss": 0.8599,
+      "step": 14315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.464652753517776e-06,
+      "loss": 0.8887,
+      "step": 14316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.463994595963924e-06,
+      "loss": 0.8633,
+      "step": 14317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4633364677962777e-06,
+      "loss": 0.8545,
+      "step": 14318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4626783690274275e-06,
+      "loss": 0.2847,
+      "step": 14319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4620202996699604e-06,
+      "loss": 0.873,
+      "step": 14320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.461362259736469e-06,
+      "loss": 0.3904,
+      "step": 14321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4607042492395433e-06,
+      "loss": 0.8423,
+      "step": 14322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.460046268191769e-06,
+      "loss": 0.7563,
+      "step": 14323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4593883166057385e-06,
+      "loss": 0.8682,
+      "step": 14324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4587303944940353e-06,
+      "loss": 0.7896,
+      "step": 14325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4580725018692486e-06,
+      "loss": 0.8716,
+      "step": 14326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.457414638743966e-06,
+      "loss": 0.8711,
+      "step": 14327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4567568051307713e-06,
+      "loss": 0.8462,
+      "step": 14328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4560990010422525e-06,
+      "loss": 0.855,
+      "step": 14329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4554412264909897e-06,
+      "loss": 0.856,
+      "step": 14330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4547834814895715e-06,
+      "loss": 0.874,
+      "step": 14331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.454125766050581e-06,
+      "loss": 0.9028,
+      "step": 14332
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4534680801865987e-06,
+      "loss": 0.8472,
+      "step": 14333
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.45281042391021e-06,
+      "loss": 0.8911,
+      "step": 14334
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4521527972339974e-06,
+      "loss": 0.8604,
+      "step": 14335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.451495200170539e-06,
+      "loss": 0.8574,
+      "step": 14336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.450837632732419e-06,
+      "loss": 0.8921,
+      "step": 14337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4501800949322157e-06,
+      "loss": 0.8613,
+      "step": 14338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.449522586782508e-06,
+      "loss": 0.8545,
+      "step": 14339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.448865108295879e-06,
+      "loss": 0.8608,
+      "step": 14340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4482076594849033e-06,
+      "loss": 0.8623,
+      "step": 14341
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4475502403621593e-06,
+      "loss": 0.8779,
+      "step": 14342
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.446892850940228e-06,
+      "loss": 0.8428,
+      "step": 14343
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.446235491231682e-06,
+      "loss": 0.835,
+      "step": 14344
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4455781612491013e-06,
+      "loss": 0.8848,
+      "step": 14345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4449208610050576e-06,
+      "loss": 0.9111,
+      "step": 14346
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4442635905121293e-06,
+      "loss": 0.8896,
+      "step": 14347
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4436063497828914e-06,
+      "loss": 0.8853,
+      "step": 14348
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4429491388299143e-06,
+      "loss": 0.8403,
+      "step": 14349
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.442291957665773e-06,
+      "loss": 0.3003,
+      "step": 14350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.441634806303046e-06,
+      "loss": 0.8735,
+      "step": 14351
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4409776847542965e-06,
+      "loss": 0.8613,
+      "step": 14352
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4403205930321025e-06,
+      "loss": 0.8276,
+      "step": 14353
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4396635311490323e-06,
+      "loss": 0.7827,
+      "step": 14354
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4390064991176575e-06,
+      "loss": 0.8647,
+      "step": 14355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.43834949695055e-06,
+      "loss": 0.8745,
+      "step": 14356
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.437692524660275e-06,
+      "loss": 0.9087,
+      "step": 14357
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4370355822594072e-06,
+      "loss": 0.8457,
+      "step": 14358
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4363786697605084e-06,
+      "loss": 0.8701,
+      "step": 14359
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4357217871761493e-06,
+      "loss": 0.8481,
+      "step": 14360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4350649345189e-06,
+      "loss": 0.8799,
+      "step": 14361
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4344081118013232e-06,
+      "loss": 0.8252,
+      "step": 14362
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4337513190359855e-06,
+      "loss": 0.8257,
+      "step": 14363
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.433094556235455e-06,
+      "loss": 0.3409,
+      "step": 14364
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4324378234122933e-06,
+      "loss": 0.8726,
+      "step": 14365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4317811205790686e-06,
+      "loss": 0.8364,
+      "step": 14366
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.431124447748339e-06,
+      "loss": 0.8462,
+      "step": 14367
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4304678049326723e-06,
+      "loss": 0.8154,
+      "step": 14368
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.429811192144631e-06,
+      "loss": 0.9062,
+      "step": 14369
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4291546093967736e-06,
+      "loss": 0.8306,
+      "step": 14370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.428498056701665e-06,
+      "loss": 0.8203,
+      "step": 14371
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4278415340718662e-06,
+      "loss": 0.8438,
+      "step": 14372
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4271850415199336e-06,
+      "loss": 0.8701,
+      "step": 14373
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4265285790584323e-06,
+      "loss": 0.8394,
+      "step": 14374
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.425872146699918e-06,
+      "loss": 0.8872,
+      "step": 14375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4252157444569478e-06,
+      "loss": 0.9004,
+      "step": 14376
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.424559372342085e-06,
+      "loss": 0.8247,
+      "step": 14377
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4239030303678826e-06,
+      "loss": 0.8477,
+      "step": 14378
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4232467185468982e-06,
+      "loss": 0.8076,
+      "step": 14379
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4225904368916907e-06,
+      "loss": 0.8872,
+      "step": 14380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4219341854148124e-06,
+      "loss": 0.8247,
+      "step": 14381
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.421277964128821e-06,
+      "loss": 0.8774,
+      "step": 14382
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4206217730462675e-06,
+      "loss": 0.8213,
+      "step": 14383
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.41996561217971e-06,
+      "loss": 0.8965,
+      "step": 14384
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.419309481541701e-06,
+      "loss": 0.8291,
+      "step": 14385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4186533811447903e-06,
+      "loss": 0.7979,
+      "step": 14386
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4179973110015334e-06,
+      "loss": 0.8359,
+      "step": 14387
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.417341271124482e-06,
+      "loss": 0.8315,
+      "step": 14388
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4166852615261837e-06,
+      "loss": 0.8711,
+      "step": 14389
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4160292822191936e-06,
+      "loss": 0.8843,
+      "step": 14390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4153733332160587e-06,
+      "loss": 0.8789,
+      "step": 14391
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.414717414529327e-06,
+      "loss": 0.7954,
+      "step": 14392
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.414061526171552e-06,
+      "loss": 0.8516,
+      "step": 14393
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4134056681552785e-06,
+      "loss": 0.3416,
+      "step": 14394
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.412749840493055e-06,
+      "loss": 0.8662,
+      "step": 14395
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4120940431974264e-06,
+      "loss": 0.8491,
+      "step": 14396
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4114382762809418e-06,
+      "loss": 0.8564,
+      "step": 14397
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.410782539756148e-06,
+      "loss": 0.4143,
+      "step": 14398
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.410126833635585e-06,
+      "loss": 0.895,
+      "step": 14399
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4094711579318036e-06,
+      "loss": 0.8213,
+      "step": 14400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4088155126573457e-06,
+      "loss": 0.8633,
+      "step": 14401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.408159897824752e-06,
+      "loss": 0.3162,
+      "step": 14402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.407504313446571e-06,
+      "loss": 0.7905,
+      "step": 14403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.40684875953534e-06,
+      "loss": 0.9053,
+      "step": 14404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.406193236103602e-06,
+      "loss": 0.9111,
+      "step": 14405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.405537743163901e-06,
+      "loss": 0.874,
+      "step": 14406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.404882280728774e-06,
+      "loss": 0.8682,
+      "step": 14407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.404226848810762e-06,
+      "loss": 0.8438,
+      "step": 14408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4035714474224075e-06,
+      "loss": 0.8491,
+      "step": 14409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4029160765762453e-06,
+      "loss": 0.8594,
+      "step": 14410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4022607362848164e-06,
+      "loss": 0.8154,
+      "step": 14411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.401605426560655e-06,
+      "loss": 0.9131,
+      "step": 14412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.400950147416302e-06,
+      "loss": 0.894,
+      "step": 14413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4002948988642936e-06,
+      "loss": 0.8735,
+      "step": 14414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3996396809171617e-06,
+      "loss": 0.8638,
+      "step": 14415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.398984493587446e-06,
+      "loss": 0.8696,
+      "step": 14416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3983293368876815e-06,
+      "loss": 0.8613,
+      "step": 14417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3976742108303984e-06,
+      "loss": 0.833,
+      "step": 14418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.397019115428134e-06,
+      "loss": 0.8745,
+      "step": 14419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3963640506934193e-06,
+      "loss": 0.8955,
+      "step": 14420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.395709016638786e-06,
+      "loss": 0.8345,
+      "step": 14421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3950540132767696e-06,
+      "loss": 0.8379,
+      "step": 14422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3943990406198985e-06,
+      "loss": 0.856,
+      "step": 14423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3937440986807022e-06,
+      "loss": 0.8716,
+      "step": 14424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3930891874717144e-06,
+      "loss": 0.8784,
+      "step": 14425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.392434307005462e-06,
+      "loss": 0.8604,
+      "step": 14426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.391779457294475e-06,
+      "loss": 0.8291,
+      "step": 14427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3911246383512796e-06,
+      "loss": 0.8135,
+      "step": 14428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.390469850188406e-06,
+      "loss": 0.8945,
+      "step": 14429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3898150928183815e-06,
+      "loss": 0.8584,
+      "step": 14430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3891603662537297e-06,
+      "loss": 0.8169,
+      "step": 14431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.388505670506981e-06,
+      "loss": 0.8804,
+      "step": 14432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3878510055906567e-06,
+      "loss": 0.7988,
+      "step": 14433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3871963715172818e-06,
+      "loss": 0.7769,
+      "step": 14434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.386541768299384e-06,
+      "loss": 0.8218,
+      "step": 14435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.385887195949483e-06,
+      "loss": 0.3184,
+      "step": 14436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.385232654480104e-06,
+      "loss": 0.8599,
+      "step": 14437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3845781439037695e-06,
+      "loss": 0.8555,
+      "step": 14438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3839236642330003e-06,
+      "loss": 0.3221,
+      "step": 14439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3832692154803195e-06,
+      "loss": 0.813,
+      "step": 14440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3826147976582435e-06,
+      "loss": 0.8218,
+      "step": 14441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.381960410779296e-06,
+      "loss": 0.8745,
+      "step": 14442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3813060548559974e-06,
+      "loss": 0.8271,
+      "step": 14443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3806517299008617e-06,
+      "loss": 0.8979,
+      "step": 14444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3799974359264113e-06,
+      "loss": 0.8608,
+      "step": 14445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.379343172945164e-06,
+      "loss": 0.875,
+      "step": 14446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3786889409696342e-06,
+      "loss": 0.8228,
+      "step": 14447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3780347400123416e-06,
+      "loss": 0.855,
+      "step": 14448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3773805700857986e-06,
+      "loss": 0.8149,
+      "step": 14449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.376726431202522e-06,
+      "loss": 0.9087,
+      "step": 14450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.376072323375028e-06,
+      "loss": 0.8652,
+      "step": 14451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.375418246615829e-06,
+      "loss": 0.8682,
+      "step": 14452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.374764200937438e-06,
+      "loss": 0.873,
+      "step": 14453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3741101863523707e-06,
+      "loss": 0.8525,
+      "step": 14454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3734562028731365e-06,
+      "loss": 0.8403,
+      "step": 14455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.37280225051225e-06,
+      "loss": 0.8135,
+      "step": 14456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3721483292822176e-06,
+      "loss": 0.8657,
+      "step": 14457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3714944391955546e-06,
+      "loss": 0.8516,
+      "step": 14458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3708405802647696e-06,
+      "loss": 0.8706,
+      "step": 14459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3701867525023703e-06,
+      "loss": 0.897,
+      "step": 14460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369532955920868e-06,
+      "loss": 0.875,
+      "step": 14461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.368879190532769e-06,
+      "loss": 0.8472,
+      "step": 14462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.36822545635058e-06,
+      "loss": 0.8071,
+      "step": 14463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3675717533868112e-06,
+      "loss": 0.8462,
+      "step": 14464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3669180816539655e-06,
+      "loss": 0.8384,
+      "step": 14465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3662644411645513e-06,
+      "loss": 0.873,
+      "step": 14466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.365610831931074e-06,
+      "loss": 0.8721,
+      "step": 14467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3649572539660345e-06,
+      "loss": 0.856,
+      "step": 14468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3643037072819422e-06,
+      "loss": 0.8564,
+      "step": 14469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.363650191891295e-06,
+      "loss": 0.8662,
+      "step": 14470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.362996707806598e-06,
+      "loss": 0.8789,
+      "step": 14471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3623432550403555e-06,
+      "loss": 0.8433,
+      "step": 14472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3616898336050663e-06,
+      "loss": 0.9092,
+      "step": 14473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.361036443513231e-06,
+      "loss": 0.8013,
+      "step": 14474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.360383084777353e-06,
+      "loss": 0.8672,
+      "step": 14475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3597297574099297e-06,
+      "loss": 0.8687,
+      "step": 14476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.359076461423462e-06,
+      "loss": 0.8696,
+      "step": 14477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3584231968304453e-06,
+      "loss": 0.8823,
+      "step": 14478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3577699636433802e-06,
+      "loss": 0.8398,
+      "step": 14479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.357116761874766e-06,
+      "loss": 0.8643,
+      "step": 14480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3564635915370936e-06,
+      "loss": 0.7495,
+      "step": 14481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3558104526428636e-06,
+      "loss": 0.8091,
+      "step": 14482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3551573452045714e-06,
+      "loss": 0.8408,
+      "step": 14483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.35450426923471e-06,
+      "loss": 0.7642,
+      "step": 14484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.353851224745777e-06,
+      "loss": 0.8652,
+      "step": 14485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.353198211750262e-06,
+      "loss": 0.8062,
+      "step": 14486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3525452302606596e-06,
+      "loss": 0.8213,
+      "step": 14487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3518922802894655e-06,
+      "loss": 0.8472,
+      "step": 14488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.351239361849168e-06,
+      "loss": 0.9092,
+      "step": 14489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3505864749522586e-06,
+      "loss": 0.8345,
+      "step": 14490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.349933619611231e-06,
+      "loss": 0.8789,
+      "step": 14491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.349280795838572e-06,
+      "loss": 0.8691,
+      "step": 14492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3486280036467745e-06,
+      "loss": 0.8311,
+      "step": 14493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3479752430483224e-06,
+      "loss": 0.8296,
+      "step": 14494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3473225140557096e-06,
+      "loss": 0.8872,
+      "step": 14495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3466698166814213e-06,
+      "loss": 0.8711,
+      "step": 14496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3460171509379437e-06,
+      "loss": 0.8682,
+      "step": 14497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3453645168377654e-06,
+      "loss": 0.9082,
+      "step": 14498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3447119143933703e-06,
+      "loss": 0.8853,
+      "step": 14499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3440593436172442e-06,
+      "loss": 0.8325,
+      "step": 14500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3434068045218737e-06,
+      "loss": 0.9155,
+      "step": 14501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3427542971197414e-06,
+      "loss": 0.8452,
+      "step": 14502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.342101821423329e-06,
+      "loss": 0.8887,
+      "step": 14503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.341449377445124e-06,
+      "loss": 0.8276,
+      "step": 14504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3407969651976045e-06,
+      "loss": 0.2905,
+      "step": 14505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3401445846932545e-06,
+      "loss": 0.8008,
+      "step": 14506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3394922359445525e-06,
+      "loss": 0.8296,
+      "step": 14507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.338839918963982e-06,
+      "loss": 0.7788,
+      "step": 14508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3381876337640216e-06,
+      "loss": 0.835,
+      "step": 14509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.337535380357149e-06,
+      "loss": 0.8169,
+      "step": 14510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3368831587558455e-06,
+      "loss": 0.8462,
+      "step": 14511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3362309689725887e-06,
+      "loss": 0.8662,
+      "step": 14512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.335578811019853e-06,
+      "loss": 0.8184,
+      "step": 14513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.33492668491012e-06,
+      "loss": 0.9214,
+      "step": 14514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3342745906558614e-06,
+      "loss": 0.8843,
+      "step": 14515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3336225282695544e-06,
+      "loss": 0.897,
+      "step": 14516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3329704977636766e-06,
+      "loss": 0.8262,
+      "step": 14517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3323184991506983e-06,
+      "loss": 0.8862,
+      "step": 14518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.331666532443095e-06,
+      "loss": 0.8867,
+      "step": 14519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3310145976533414e-06,
+      "loss": 0.8008,
+      "step": 14520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3303626947939078e-06,
+      "loss": 0.8506,
+      "step": 14521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3297108238772674e-06,
+      "loss": 0.9116,
+      "step": 14522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3290589849158894e-06,
+      "loss": 0.8145,
+      "step": 14523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.328407177922247e-06,
+      "loss": 0.8481,
+      "step": 14524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.32775540290881e-06,
+      "loss": 0.8608,
+      "step": 14525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3271036598880456e-06,
+      "loss": 0.896,
+      "step": 14526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3264519488724266e-06,
+      "loss": 0.9097,
+      "step": 14527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3258002698744175e-06,
+      "loss": 0.8184,
+      "step": 14528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3251486229064867e-06,
+      "loss": 0.8293,
+      "step": 14529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.324497007981104e-06,
+      "loss": 0.876,
+      "step": 14530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.323845425110733e-06,
+      "loss": 0.8936,
+      "step": 14531
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3231938743078383e-06,
+      "loss": 0.8643,
+      "step": 14532
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.32254235558489e-06,
+      "loss": 0.8296,
+      "step": 14533
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.321890868954349e-06,
+      "loss": 0.3123,
+      "step": 14534
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.321239414428681e-06,
+      "loss": 0.7842,
+      "step": 14535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3205879920203454e-06,
+      "loss": 0.8501,
+      "step": 14536
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.31993660174181e-06,
+      "loss": 0.8491,
+      "step": 14537
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3192852436055355e-06,
+      "loss": 0.8167,
+      "step": 14538
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3186339176239808e-06,
+      "loss": 0.814,
+      "step": 14539
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3179826238096098e-06,
+      "loss": 0.856,
+      "step": 14540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3173313621748827e-06,
+      "loss": 0.8174,
+      "step": 14541
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.316680132732256e-06,
+      "loss": 0.8696,
+      "step": 14542
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.316028935494193e-06,
+      "loss": 0.8867,
+      "step": 14543
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3153777704731493e-06,
+      "loss": 0.8027,
+      "step": 14544
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.314726637681582e-06,
+      "loss": 0.8086,
+      "step": 14545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3140755371319523e-06,
+      "loss": 0.8335,
+      "step": 14546
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3134244688367133e-06,
+      "loss": 0.7671,
+      "step": 14547
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3127734328083206e-06,
+      "loss": 0.8027,
+      "step": 14548
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.312122429059233e-06,
+      "loss": 0.873,
+      "step": 14549
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.311471457601902e-06,
+      "loss": 0.8628,
+      "step": 14550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.310820518448784e-06,
+      "loss": 0.8799,
+      "step": 14551
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3101696116123295e-06,
+      "loss": 0.8716,
+      "step": 14552
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.309518737104993e-06,
+      "loss": 0.8838,
+      "step": 14553
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3088678949392295e-06,
+      "loss": 0.8262,
+      "step": 14554
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.308217085127485e-06,
+      "loss": 0.8188,
+      "step": 14555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.307566307682215e-06,
+      "loss": 0.3336,
+      "step": 14556
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3069155626158696e-06,
+      "loss": 0.8774,
+      "step": 14557
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.306264849940895e-06,
+      "loss": 0.9058,
+      "step": 14558
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3056141696697454e-06,
+      "loss": 0.3447,
+      "step": 14559
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.304963521814865e-06,
+      "loss": 0.8301,
+      "step": 14560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.304312906388703e-06,
+      "loss": 0.8608,
+      "step": 14561
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.303662323403709e-06,
+      "loss": 0.7939,
+      "step": 14562
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.303011772872327e-06,
+      "loss": 0.8486,
+      "step": 14563
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.302361254807005e-06,
+      "loss": 0.8281,
+      "step": 14564
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.301710769220185e-06,
+      "loss": 0.845,
+      "step": 14565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3010603161243153e-06,
+      "loss": 0.8877,
+      "step": 14566
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.30040989553184e-06,
+      "loss": 0.8906,
+      "step": 14567
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2997595074551996e-06,
+      "loss": 0.8506,
+      "step": 14568
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2991091519068396e-06,
+      "loss": 0.8501,
+      "step": 14569
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.298458828899204e-06,
+      "loss": 0.8257,
+      "step": 14570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.297808538444729e-06,
+      "loss": 0.8433,
+      "step": 14571
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2971582805558622e-06,
+      "loss": 0.873,
+      "step": 14572
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2965080552450392e-06,
+      "loss": 0.7471,
+      "step": 14573
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2958578625247e-06,
+      "loss": 0.8384,
+      "step": 14574
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.295207702407288e-06,
+      "loss": 0.8359,
+      "step": 14575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2945575749052366e-06,
+      "loss": 0.8477,
+      "step": 14576
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2939074800309874e-06,
+      "loss": 0.876,
+      "step": 14577
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.293257417796978e-06,
+      "loss": 0.7559,
+      "step": 14578
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2926073882156408e-06,
+      "loss": 0.8687,
+      "step": 14579
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2919573912994186e-06,
+      "loss": 0.8672,
+      "step": 14580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.291307427060739e-06,
+      "loss": 0.8857,
+      "step": 14581
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2906574955120418e-06,
+      "loss": 0.8311,
+      "step": 14582
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2900075966657623e-06,
+      "loss": 0.834,
+      "step": 14583
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2893577305343312e-06,
+      "loss": 0.8184,
+      "step": 14584
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.288707897130181e-06,
+      "loss": 0.915,
+      "step": 14585
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.288058096465747e-06,
+      "loss": 0.8564,
+      "step": 14586
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.287408328553459e-06,
+      "loss": 0.8423,
+      "step": 14587
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2867585934057487e-06,
+      "loss": 0.8467,
+      "step": 14588
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.286108891035045e-06,
+      "loss": 0.8872,
+      "step": 14589
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2854592214537796e-06,
+      "loss": 0.8687,
+      "step": 14590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.284809584674382e-06,
+      "loss": 0.8516,
+      "step": 14591
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.284159980709278e-06,
+      "loss": 0.8223,
+      "step": 14592
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2835104095708997e-06,
+      "loss": 0.7744,
+      "step": 14593
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.28286087127167e-06,
+      "loss": 0.8535,
+      "step": 14594
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2822113658240184e-06,
+      "loss": 0.8267,
+      "step": 14595
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2815618932403716e-06,
+      "loss": 0.8179,
+      "step": 14596
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2809124535331517e-06,
+      "loss": 0.8496,
+      "step": 14597
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2802630467147854e-06,
+      "loss": 0.8599,
+      "step": 14598
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.279613672797699e-06,
+      "loss": 0.8213,
+      "step": 14599
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2789643317943134e-06,
+      "loss": 0.9043,
+      "step": 14600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2783150237170523e-06,
+      "loss": 0.855,
+      "step": 14601
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2776657485783357e-06,
+      "loss": 0.835,
+      "step": 14602
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.277016506390589e-06,
+      "loss": 0.8408,
+      "step": 14603
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2763672971662323e-06,
+      "loss": 0.8418,
+      "step": 14604
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2757181209176835e-06,
+      "loss": 0.8755,
+      "step": 14605
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.275068977657365e-06,
+      "loss": 0.8989,
+      "step": 14606
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.274419867397697e-06,
+      "loss": 0.8511,
+      "step": 14607
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.273770790151093e-06,
+      "loss": 0.8706,
+      "step": 14608
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.273121745929976e-06,
+      "loss": 0.9146,
+      "step": 14609
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.272472734746761e-06,
+      "loss": 0.7954,
+      "step": 14610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2718237566138634e-06,
+      "loss": 0.8486,
+      "step": 14611
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2711748115437027e-06,
+      "loss": 0.8848,
+      "step": 14612
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.270525899548691e-06,
+      "loss": 0.8555,
+      "step": 14613
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2698770206412432e-06,
+      "loss": 0.8613,
+      "step": 14614
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2692281748337764e-06,
+      "loss": 0.8926,
+      "step": 14615
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2685793621387006e-06,
+      "loss": 0.856,
+      "step": 14616
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2679305825684314e-06,
+      "loss": 0.7578,
+      "step": 14617
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2672818361353777e-06,
+      "loss": 0.8613,
+      "step": 14618
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2666331228519532e-06,
+      "loss": 0.8706,
+      "step": 14619
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2659844427305704e-06,
+      "loss": 0.8418,
+      "step": 14620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2653357957836353e-06,
+      "loss": 0.8462,
+      "step": 14621
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2646871820235604e-06,
+      "loss": 0.8862,
+      "step": 14622
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2640386014627563e-06,
+      "loss": 0.7593,
+      "step": 14623
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2633900541136266e-06,
+      "loss": 0.8877,
+      "step": 14624
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2627415399885837e-06,
+      "loss": 0.918,
+      "step": 14625
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2620930591000316e-06,
+      "loss": 0.8369,
+      "step": 14626
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.261444611460377e-06,
+      "loss": 0.9141,
+      "step": 14627
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.260796197082029e-06,
+      "loss": 0.8062,
+      "step": 14628
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.260147815977389e-06,
+      "loss": 0.8853,
+      "step": 14629
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2594994681588645e-06,
+      "loss": 0.8716,
+      "step": 14630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.258851153638855e-06,
+      "loss": 0.8652,
+      "step": 14631
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.258202872429768e-06,
+      "loss": 0.9077,
+      "step": 14632
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.257554624544006e-06,
+      "loss": 0.7861,
+      "step": 14633
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2569064099939673e-06,
+      "loss": 0.8311,
+      "step": 14634
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2562582287920573e-06,
+      "loss": 0.8345,
+      "step": 14635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.255610080950676e-06,
+      "loss": 0.814,
+      "step": 14636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2549619664822206e-06,
+      "loss": 0.814,
+      "step": 14637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.254313885399094e-06,
+      "loss": 0.8584,
+      "step": 14638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.253665837713694e-06,
+      "loss": 0.8467,
+      "step": 14639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.253017823438416e-06,
+      "loss": 0.8262,
+      "step": 14640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.252369842585663e-06,
+      "loss": 0.8628,
+      "step": 14641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2517218951678273e-06,
+      "loss": 0.8345,
+      "step": 14642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2510739811973058e-06,
+      "loss": 0.9087,
+      "step": 14643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.250426100686497e-06,
+      "loss": 0.8877,
+      "step": 14644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.249778253647793e-06,
+      "loss": 0.8608,
+      "step": 14645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2491304400935898e-06,
+      "loss": 0.8257,
+      "step": 14646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2484826600362785e-06,
+      "loss": 0.8013,
+      "step": 14647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2478349134882547e-06,
+      "loss": 0.8647,
+      "step": 14648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2471872004619114e-06,
+      "loss": 0.8564,
+      "step": 14649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.246539520969637e-06,
+      "loss": 0.8267,
+      "step": 14650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2458918750238253e-06,
+      "loss": 0.9087,
+      "step": 14651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2452442626368673e-06,
+      "loss": 0.792,
+      "step": 14652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2445966838211508e-06,
+      "loss": 0.8501,
+      "step": 14653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2439491385890667e-06,
+      "loss": 0.9438,
+      "step": 14654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2433016269530015e-06,
+      "loss": 0.8789,
+      "step": 14655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2426541489253445e-06,
+      "loss": 0.814,
+      "step": 14656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2420067045184843e-06,
+      "loss": 0.8843,
+      "step": 14657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2413592937448046e-06,
+      "loss": 0.9082,
+      "step": 14658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2407119166166923e-06,
+      "loss": 0.7954,
+      "step": 14659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2400645731465357e-06,
+      "loss": 0.8545,
+      "step": 14660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.239417263346716e-06,
+      "loss": 0.8545,
+      "step": 14661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.238769987229619e-06,
+      "loss": 0.8911,
+      "step": 14662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.238122744807626e-06,
+      "loss": 0.8101,
+      "step": 14663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.237475536093122e-06,
+      "loss": 0.8545,
+      "step": 14664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.236828361098489e-06,
+      "loss": 0.8262,
+      "step": 14665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2361812198361064e-06,
+      "loss": 0.8589,
+      "step": 14666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.235534112318358e-06,
+      "loss": 0.8101,
+      "step": 14667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.234887038557622e-06,
+      "loss": 0.8965,
+      "step": 14668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2342399985662774e-06,
+      "loss": 0.8384,
+      "step": 14669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2335929923567057e-06,
+      "loss": 0.9136,
+      "step": 14670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2329460199412833e-06,
+      "loss": 0.8613,
+      "step": 14671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2322990813323862e-06,
+      "loss": 0.8188,
+      "step": 14672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.231652176542396e-06,
+      "loss": 0.8408,
+      "step": 14673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.231005305583684e-06,
+      "loss": 0.8604,
+      "step": 14674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.230358468468631e-06,
+      "loss": 0.8608,
+      "step": 14675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2297116652096055e-06,
+      "loss": 0.8984,
+      "step": 14676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.229064895818987e-06,
+      "loss": 0.8535,
+      "step": 14677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.228418160309149e-06,
+      "loss": 0.8218,
+      "step": 14678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2277714586924603e-06,
+      "loss": 0.8569,
+      "step": 14679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2271247909812986e-06,
+      "loss": 0.9268,
+      "step": 14680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.226478157188033e-06,
+      "loss": 0.8579,
+      "step": 14681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225831557325034e-06,
+      "loss": 0.7739,
+      "step": 14682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225184991404675e-06,
+      "loss": 0.8159,
+      "step": 14683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2245384594393226e-06,
+      "loss": 0.8809,
+      "step": 14684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2238919614413462e-06,
+      "loss": 0.8335,
+      "step": 14685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2232454974231176e-06,
+      "loss": 0.7661,
+      "step": 14686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.222599067397001e-06,
+      "loss": 0.9238,
+      "step": 14687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2219526713753645e-06,
+      "loss": 0.873,
+      "step": 14688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2213063093705764e-06,
+      "loss": 0.8184,
+      "step": 14689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2206599813950006e-06,
+      "loss": 0.8125,
+      "step": 14690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2200136874610056e-06,
+      "loss": 0.3804,
+      "step": 14691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2193674275809513e-06,
+      "loss": 0.8267,
+      "step": 14692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.218721201767203e-06,
+      "loss": 0.9092,
+      "step": 14693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.218075010032128e-06,
+      "loss": 0.813,
+      "step": 14694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2174288523880846e-06,
+      "loss": 0.9009,
+      "step": 14695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2167827288474373e-06,
+      "loss": 0.8745,
+      "step": 14696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2161366394225445e-06,
+      "loss": 0.8716,
+      "step": 14697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.21549058412577e-06,
+      "loss": 0.7666,
+      "step": 14698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2148445629694735e-06,
+      "loss": 0.8311,
+      "step": 14699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.214198575966011e-06,
+      "loss": 0.7925,
+      "step": 14700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.213552623127746e-06,
+      "loss": 0.8057,
+      "step": 14701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.212906704467035e-06,
+      "loss": 0.8662,
+      "step": 14702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2122608199962323e-06,
+      "loss": 0.8442,
+      "step": 14703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.211614969727699e-06,
+      "loss": 0.8633,
+      "step": 14704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.210969153673789e-06,
+      "loss": 0.9009,
+      "step": 14705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2103233718468574e-06,
+      "loss": 0.874,
+      "step": 14706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.209677624259262e-06,
+      "loss": 0.8687,
+      "step": 14707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2090319109233536e-06,
+      "loss": 0.8555,
+      "step": 14708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.208386231851486e-06,
+      "loss": 0.875,
+      "step": 14709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.207740587056014e-06,
+      "loss": 0.8701,
+      "step": 14710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.207094976549289e-06,
+      "loss": 0.8271,
+      "step": 14711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.206449400343663e-06,
+      "loss": 0.8506,
+      "step": 14712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.205803858451485e-06,
+      "loss": 0.8262,
+      "step": 14713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.205158350885107e-06,
+      "loss": 0.3892,
+      "step": 14714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2045128776568783e-06,
+      "loss": 0.894,
+      "step": 14715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2038674387791464e-06,
+      "loss": 0.8325,
+      "step": 14716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2032220342642618e-06,
+      "loss": 0.8477,
+      "step": 14717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2025766641245715e-06,
+      "loss": 0.8291,
+      "step": 14718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.201931328372421e-06,
+      "loss": 0.9165,
+      "step": 14719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2012860270201585e-06,
+      "loss": 0.8613,
+      "step": 14720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.200640760080128e-06,
+      "loss": 0.9043,
+      "step": 14721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1999955275646743e-06,
+      "loss": 0.915,
+      "step": 14722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1993503294861443e-06,
+      "loss": 0.9004,
+      "step": 14723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1987051658568797e-06,
+      "loss": 0.8057,
+      "step": 14724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1980600366892225e-06,
+      "loss": 0.9092,
+      "step": 14725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1974149419955182e-06,
+      "loss": 0.9048,
+      "step": 14726
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1967698817881052e-06,
+      "loss": 0.8555,
+      "step": 14727
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1961248560793273e-06,
+      "loss": 0.8442,
+      "step": 14728
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.195479864881521e-06,
+      "loss": 0.8667,
+      "step": 14729
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.19483490820703e-06,
+      "loss": 0.876,
+      "step": 14730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.194189986068193e-06,
+      "loss": 0.9136,
+      "step": 14731
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.193545098477345e-06,
+      "loss": 0.834,
+      "step": 14732
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.192900245446827e-06,
+      "loss": 0.856,
+      "step": 14733
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.192255426988975e-06,
+      "loss": 0.8403,
+      "step": 14734
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1916106431161242e-06,
+      "loss": 0.8423,
+      "step": 14735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.190965893840613e-06,
+      "loss": 0.8496,
+      "step": 14736
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1903211791747747e-06,
+      "loss": 0.7803,
+      "step": 14737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.189676499130942e-06,
+      "loss": 0.8647,
+      "step": 14738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1890318537214532e-06,
+      "loss": 0.8721,
+      "step": 14739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.188387242958637e-06,
+      "loss": 0.9302,
+      "step": 14740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1877426668548294e-06,
+      "loss": 0.3816,
+      "step": 14741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1870981254223584e-06,
+      "loss": 0.8745,
+      "step": 14742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.186453618673557e-06,
+      "loss": 0.8804,
+      "step": 14743
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.185809146620758e-06,
+      "loss": 0.8154,
+      "step": 14744
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1851647092762863e-06,
+      "loss": 0.9014,
+      "step": 14745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1845203066524745e-06,
+      "loss": 0.832,
+      "step": 14746
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1838759387616515e-06,
+      "loss": 0.7751,
+      "step": 14747
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1832316056161407e-06,
+      "loss": 0.9111,
+      "step": 14748
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.182587307228275e-06,
+      "loss": 0.3093,
+      "step": 14749
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1819430436103765e-06,
+      "loss": 0.834,
+      "step": 14750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1812988147747714e-06,
+      "loss": 0.8315,
+      "step": 14751
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1806546207337885e-06,
+      "loss": 0.835,
+      "step": 14752
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1800104614997484e-06,
+      "loss": 0.8555,
+      "step": 14753
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.179366337084975e-06,
+      "loss": 0.3416,
+      "step": 14754
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1787222475017943e-06,
+      "loss": 0.8569,
+      "step": 14755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.178078192762526e-06,
+      "loss": 0.8193,
+      "step": 14756
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1774341728794944e-06,
+      "loss": 0.8394,
+      "step": 14757
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1767901878650166e-06,
+      "loss": 0.8413,
+      "step": 14758
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1761462377314166e-06,
+      "loss": 0.8203,
+      "step": 14759
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.175502322491014e-06,
+      "loss": 0.8188,
+      "step": 14760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1748584421561256e-06,
+      "loss": 0.3308,
+      "step": 14761
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1742145967390727e-06,
+      "loss": 0.8296,
+      "step": 14762
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1735707862521707e-06,
+      "loss": 0.8623,
+      "step": 14763
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.172927010707737e-06,
+      "loss": 0.8765,
+      "step": 14764
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1722832701180906e-06,
+      "loss": 0.8696,
+      "step": 14765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1716395644955437e-06,
+      "loss": 0.8892,
+      "step": 14766
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1709958938524133e-06,
+      "loss": 0.8662,
+      "step": 14767
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.170352258201015e-06,
+      "loss": 0.9409,
+      "step": 14768
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1697086575536596e-06,
+      "loss": 0.8628,
+      "step": 14769
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.169065091922663e-06,
+      "loss": 0.8354,
+      "step": 14770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.168421561320335e-06,
+      "loss": 0.8535,
+      "step": 14771
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1677780657589897e-06,
+      "loss": 0.9106,
+      "step": 14772
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.167134605250938e-06,
+      "loss": 0.8276,
+      "step": 14773
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.166491179808487e-06,
+      "loss": 0.897,
+      "step": 14774
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1658477894439497e-06,
+      "loss": 0.8857,
+      "step": 14775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1652044341696365e-06,
+      "loss": 0.8643,
+      "step": 14776
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1645611139978514e-06,
+      "loss": 0.8877,
+      "step": 14777
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.163917828940907e-06,
+      "loss": 0.7998,
+      "step": 14778
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1632745790111064e-06,
+      "loss": 0.8784,
+      "step": 14779
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.162631364220756e-06,
+      "loss": 0.8066,
+      "step": 14780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.161988184582165e-06,
+      "loss": 0.7554,
+      "step": 14781
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1613450401076363e-06,
+      "loss": 0.7974,
+      "step": 14782
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.160701930809472e-06,
+      "loss": 0.8237,
+      "step": 14783
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.160058856699981e-06,
+      "loss": 0.8257,
+      "step": 14784
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1594158177914624e-06,
+      "loss": 0.896,
+      "step": 14785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1587728140962204e-06,
+      "loss": 0.9023,
+      "step": 14786
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.158129845626554e-06,
+      "loss": 0.8486,
+      "step": 14787
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.157486912394767e-06,
+      "loss": 0.8896,
+      "step": 14788
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.15684401441316e-06,
+      "loss": 0.7666,
+      "step": 14789
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1562011516940296e-06,
+      "loss": 0.9272,
+      "step": 14790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.155558324249677e-06,
+      "loss": 0.8555,
+      "step": 14791
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.154915532092402e-06,
+      "loss": 0.8301,
+      "step": 14792
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1542727752344977e-06,
+      "loss": 0.8091,
+      "step": 14793
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1536300536882658e-06,
+      "loss": 0.812,
+      "step": 14794
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1529873674659996e-06,
+      "loss": 0.8359,
+      "step": 14795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1523447165799947e-06,
+      "loss": 0.8745,
+      "step": 14796
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1517021010425487e-06,
+      "loss": 0.8184,
+      "step": 14797
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1510595208659526e-06,
+      "loss": 0.9185,
+      "step": 14798
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1504169760625035e-06,
+      "loss": 0.3459,
+      "step": 14799
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.14977446664449e-06,
+      "loss": 0.875,
+      "step": 14800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1491319926242057e-06,
+      "loss": 0.8711,
+      "step": 14801
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1484895540139464e-06,
+      "loss": 0.8711,
+      "step": 14802
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1478471508259968e-06,
+      "loss": 0.7756,
+      "step": 14803
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1472047830726494e-06,
+      "loss": 0.8345,
+      "step": 14804
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.146562450766197e-06,
+      "loss": 0.9165,
+      "step": 14805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.145920153918924e-06,
+      "loss": 0.8599,
+      "step": 14806
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.145277892543121e-06,
+      "loss": 0.7749,
+      "step": 14807
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1446356666510726e-06,
+      "loss": 0.8076,
+      "step": 14808
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.143993476255069e-06,
+      "loss": 0.9004,
+      "step": 14809
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1433513213673956e-06,
+      "loss": 0.8735,
+      "step": 14810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1427092020003347e-06,
+      "loss": 0.8894,
+      "step": 14811
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1420671181661755e-06,
+      "loss": 0.8618,
+      "step": 14812
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1414250698772003e-06,
+      "loss": 0.8433,
+      "step": 14813
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1407830571456904e-06,
+      "loss": 0.8994,
+      "step": 14814
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.140141079983933e-06,
+      "loss": 0.7935,
+      "step": 14815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1394991384042056e-06,
+      "loss": 0.8691,
+      "step": 14816
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1388572324187905e-06,
+      "loss": 0.3921,
+      "step": 14817
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.138215362039972e-06,
+      "loss": 0.7998,
+      "step": 14818
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.137573527280026e-06,
+      "loss": 0.8594,
+      "step": 14819
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1369317281512325e-06,
+      "loss": 0.8218,
+      "step": 14820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.136289964665873e-06,
+      "loss": 0.8804,
+      "step": 14821
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1356482368362216e-06,
+      "loss": 0.7755,
+      "step": 14822
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1350065446745593e-06,
+      "loss": 0.7729,
+      "step": 14823
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1343648881931575e-06,
+      "loss": 0.8911,
+      "step": 14824
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1337232674042973e-06,
+      "loss": 0.9033,
+      "step": 14825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1330816823202525e-06,
+      "loss": 0.832,
+      "step": 14826
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.132440132953295e-06,
+      "loss": 0.7847,
+      "step": 14827
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1317986193157023e-06,
+      "loss": 0.9053,
+      "step": 14828
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131157141419745e-06,
+      "loss": 0.7529,
+      "step": 14829
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1305156992776966e-06,
+      "loss": 0.8223,
+      "step": 14830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1298742929018294e-06,
+      "loss": 0.8623,
+      "step": 14831
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1292329223044137e-06,
+      "loss": 0.8384,
+      "step": 14832
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.128591587497719e-06,
+      "loss": 0.7803,
+      "step": 14833
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1279502884940185e-06,
+      "loss": 0.7803,
+      "step": 14834
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.127309025305577e-06,
+      "loss": 0.8364,
+      "step": 14835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.126667797944667e-06,
+      "loss": 0.8296,
+      "step": 14836
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.126026606423551e-06,
+      "loss": 0.8657,
+      "step": 14837
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1253854507545007e-06,
+      "loss": 0.8223,
+      "step": 14838
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1247443309497814e-06,
+      "loss": 0.8926,
+      "step": 14839
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1241032470216564e-06,
+      "loss": 0.8594,
+      "step": 14840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.123462198982393e-06,
+      "loss": 0.8555,
+      "step": 14841
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.122821186844256e-06,
+      "loss": 0.8682,
+      "step": 14842
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.122180210619506e-06,
+      "loss": 0.8301,
+      "step": 14843
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1215392703204093e-06,
+      "loss": 0.8403,
+      "step": 14844
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.120898365959226e-06,
+      "loss": 0.8569,
+      "step": 14845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1202574975482157e-06,
+      "loss": 0.8447,
+      "step": 14846
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.119616665099645e-06,
+      "loss": 0.739,
+      "step": 14847
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1189758686257687e-06,
+      "loss": 0.8101,
+      "step": 14848
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.118335108138848e-06,
+      "loss": 0.793,
+      "step": 14849
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.117694383651143e-06,
+      "loss": 0.8438,
+      "step": 14850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1170536951749102e-06,
+      "loss": 0.792,
+      "step": 14851
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.116413042722408e-06,
+      "loss": 0.8569,
+      "step": 14852
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1157724263058907e-06,
+      "loss": 0.8916,
+      "step": 14853
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1151318459376166e-06,
+      "loss": 0.8491,
+      "step": 14854
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1144913016298424e-06,
+      "loss": 0.854,
+      "step": 14855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1138507933948182e-06,
+      "loss": 0.8472,
+      "step": 14856
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1132103212448016e-06,
+      "loss": 0.8076,
+      "step": 14857
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1125698851920456e-06,
+      "loss": 0.8501,
+      "step": 14858
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1119294852488e-06,
+      "loss": 0.8774,
+      "step": 14859
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1112891214273205e-06,
+      "loss": 0.3314,
+      "step": 14860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1106487937398555e-06,
+      "loss": 0.3257,
+      "step": 14861
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1100085021986546e-06,
+      "loss": 0.8369,
+      "step": 14862
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1093682468159723e-06,
+      "loss": 0.8657,
+      "step": 14863
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.108728027604053e-06,
+      "loss": 0.8442,
+      "step": 14864
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.108087844575149e-06,
+      "loss": 0.791,
+      "step": 14865
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1074476977415026e-06,
+      "loss": 0.853,
+      "step": 14866
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1068075871153656e-06,
+      "loss": 0.8198,
+      "step": 14867
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1061675127089834e-06,
+      "loss": 0.8696,
+      "step": 14868
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1055274745345994e-06,
+      "loss": 0.812,
+      "step": 14869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.104887472604461e-06,
+      "loss": 0.7959,
+      "step": 14870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.104247506930812e-06,
+      "loss": 0.8252,
+      "step": 14871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1036075775258945e-06,
+      "loss": 0.8501,
+      "step": 14872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1029676844019543e-06,
+      "loss": 0.8311,
+      "step": 14873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.10232782757123e-06,
+      "loss": 0.8213,
+      "step": 14874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.101688007045963e-06,
+      "loss": 0.8135,
+      "step": 14875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1010482228383994e-06,
+      "loss": 0.8359,
+      "step": 14876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1004084749607743e-06,
+      "loss": 0.8433,
+      "step": 14877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0997687634253273e-06,
+      "loss": 0.8369,
+      "step": 14878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0991290882442997e-06,
+      "loss": 0.8936,
+      "step": 14879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.098489449429928e-06,
+      "loss": 0.895,
+      "step": 14880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0978498469944497e-06,
+      "loss": 0.8267,
+      "step": 14881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0972102809500993e-06,
+      "loss": 0.9102,
+      "step": 14882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.096570751309116e-06,
+      "loss": 0.8296,
+      "step": 14883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0959312580837343e-06,
+      "loss": 0.8252,
+      "step": 14884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0952918012861866e-06,
+      "loss": 0.8398,
+      "step": 14885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0946523809287088e-06,
+      "loss": 0.897,
+      "step": 14886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0940129970235344e-06,
+      "loss": 0.8486,
+      "step": 14887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0933736495828933e-06,
+      "loss": 0.8047,
+      "step": 14888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09273433861902e-06,
+      "loss": 0.9038,
+      "step": 14889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0920950641441437e-06,
+      "loss": 0.8604,
+      "step": 14890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.091455826170494e-06,
+      "loss": 0.8486,
+      "step": 14891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0908166247103043e-06,
+      "loss": 0.8438,
+      "step": 14892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0901774597758006e-06,
+      "loss": 0.8379,
+      "step": 14893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0895383313792095e-06,
+      "loss": 0.8784,
+      "step": 14894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0888992395327635e-06,
+      "loss": 0.8115,
+      "step": 14895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0882601842486848e-06,
+      "loss": 0.8599,
+      "step": 14896
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0876211655392024e-06,
+      "loss": 0.8457,
+      "step": 14897
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.086982183416539e-06,
+      "loss": 0.8389,
+      "step": 14898
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0863432378929214e-06,
+      "loss": 0.8125,
+      "step": 14899
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.085704328980575e-06,
+      "loss": 0.8115,
+      "step": 14900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0850654566917192e-06,
+      "loss": 0.8022,
+      "step": 14901
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.08442662103858e-06,
+      "loss": 0.8267,
+      "step": 14902
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0837878220333783e-06,
+      "loss": 0.9058,
+      "step": 14903
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.083149059688333e-06,
+      "loss": 0.832,
+      "step": 14904
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0825103340156694e-06,
+      "loss": 0.8389,
+      "step": 14905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.081871645027603e-06,
+      "loss": 0.8442,
+      "step": 14906
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.081232992736355e-06,
+      "loss": 0.8428,
+      "step": 14907
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0805943771541448e-06,
+      "loss": 0.9014,
+      "step": 14908
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.079955798293187e-06,
+      "loss": 0.8418,
+      "step": 14909
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0793172561657027e-06,
+      "loss": 0.7998,
+      "step": 14910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.078678750783903e-06,
+      "loss": 0.408,
+      "step": 14911
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0780402821600064e-06,
+      "loss": 0.8818,
+      "step": 14912
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.077401850306231e-06,
+      "loss": 0.877,
+      "step": 14913
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.076763455234785e-06,
+      "loss": 0.8486,
+      "step": 14914
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.076125096957885e-06,
+      "loss": 0.8618,
+      "step": 14915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.075486775487745e-06,
+      "loss": 0.8369,
+      "step": 14916
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0748484908365743e-06,
+      "loss": 0.8838,
+      "step": 14917
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0742102430165876e-06,
+      "loss": 0.855,
+      "step": 14918
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.073572032039991e-06,
+      "loss": 0.8657,
+      "step": 14919
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0729338579189983e-06,
+      "loss": 0.8369,
+      "step": 14920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0722957206658187e-06,
+      "loss": 0.8721,
+      "step": 14921
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0716576202926573e-06,
+      "loss": 0.9214,
+      "step": 14922
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0710195568117258e-06,
+      "loss": 0.8311,
+      "step": 14923
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0703815302352304e-06,
+      "loss": 0.8706,
+      "step": 14924
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.069743540575375e-06,
+      "loss": 0.8486,
+      "step": 14925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06910558784437e-06,
+      "loss": 0.8491,
+      "step": 14926
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.068467672054416e-06,
+      "loss": 0.8423,
+      "step": 14927
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.067829793217719e-06,
+      "loss": 0.8457,
+      "step": 14928
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.067191951346484e-06,
+      "loss": 0.3772,
+      "step": 14929
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0665541464529126e-06,
+      "loss": 0.8315,
+      "step": 14930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0659163785492073e-06,
+      "loss": 0.8901,
+      "step": 14931
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.065278647647567e-06,
+      "loss": 0.2968,
+      "step": 14932
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0646409537601973e-06,
+      "loss": 0.8599,
+      "step": 14933
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.064003296899296e-06,
+      "loss": 0.3982,
+      "step": 14934
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06336567707706e-06,
+      "loss": 0.8457,
+      "step": 14935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0627280943056925e-06,
+      "loss": 0.8647,
+      "step": 14936
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0620905485973894e-06,
+      "loss": 0.8052,
+      "step": 14937
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.061453039964346e-06,
+      "loss": 0.8652,
+      "step": 14938
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0608155684187635e-06,
+      "loss": 0.7886,
+      "step": 14939
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.060178133972833e-06,
+      "loss": 0.8667,
+      "step": 14940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.059540736638751e-06,
+      "loss": 0.915,
+      "step": 14941
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0589033764287147e-06,
+      "loss": 0.8105,
+      "step": 14942
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.058266053354915e-06,
+      "loss": 0.8423,
+      "step": 14943
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.057628767429543e-06,
+      "loss": 0.8799,
+      "step": 14944
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0569915186647964e-06,
+      "loss": 0.3558,
+      "step": 14945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.056354307072863e-06,
+      "loss": 0.8325,
+      "step": 14946
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0557171326659353e-06,
+      "loss": 0.8438,
+      "step": 14947
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0550799954562007e-06,
+      "loss": 0.8374,
+      "step": 14948
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0544428954558513e-06,
+      "loss": 0.8301,
+      "step": 14949
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.053805832677076e-06,
+      "loss": 0.8721,
+      "step": 14950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0531688071320608e-06,
+      "loss": 0.8433,
+      "step": 14951
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0525318188329942e-06,
+      "loss": 0.3108,
+      "step": 14952
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0518948677920644e-06,
+      "loss": 0.8345,
+      "step": 14953
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0512579540214533e-06,
+      "loss": 0.9004,
+      "step": 14954
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.05062107753335e-06,
+      "loss": 0.3765,
+      "step": 14955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.049984238339937e-06,
+      "loss": 0.918,
+      "step": 14956
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.049347436453397e-06,
+      "loss": 0.8965,
+      "step": 14957
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0487106718859173e-06,
+      "loss": 0.8853,
+      "step": 14958
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0480739446496756e-06,
+      "loss": 0.9087,
+      "step": 14959
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0474372547568537e-06,
+      "loss": 0.8438,
+      "step": 14960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.046800602219636e-06,
+      "loss": 0.8013,
+      "step": 14961
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0461639870502003e-06,
+      "loss": 0.8237,
+      "step": 14962
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0455274092607266e-06,
+      "loss": 0.8184,
+      "step": 14963
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0448908688633927e-06,
+      "loss": 0.9336,
+      "step": 14964
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0442543658703774e-06,
+      "loss": 0.854,
+      "step": 14965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.043617900293859e-06,
+      "loss": 0.8394,
+      "step": 14966
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0429814721460116e-06,
+      "loss": 0.8999,
+      "step": 14967
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.042345081439014e-06,
+      "loss": 0.7949,
+      "step": 14968
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.041708728185039e-06,
+      "loss": 0.8818,
+      "step": 14969
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.041072412396261e-06,
+      "loss": 0.8643,
+      "step": 14970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0404361340848563e-06,
+      "loss": 0.8555,
+      "step": 14971
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0397998932629957e-06,
+      "loss": 0.8691,
+      "step": 14972
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0391636899428505e-06,
+      "loss": 0.8003,
+      "step": 14973
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0385275241365965e-06,
+      "loss": 0.3218,
+      "step": 14974
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0378913958564003e-06,
+      "loss": 0.8149,
+      "step": 14975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.037255305114435e-06,
+      "loss": 0.8315,
+      "step": 14976
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.036619251922866e-06,
+      "loss": 0.8545,
+      "step": 14977
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.035983236293866e-06,
+      "loss": 0.8774,
+      "step": 14978
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.035347258239603e-06,
+      "loss": 0.8643,
+      "step": 14979
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0347113177722397e-06,
+      "loss": 0.8735,
+      "step": 14980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0340754149039475e-06,
+      "loss": 0.8491,
+      "step": 14981
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0334395496468916e-06,
+      "loss": 0.8169,
+      "step": 14982
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0328037220132346e-06,
+      "loss": 0.8774,
+      "step": 14983
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0321679320151433e-06,
+      "loss": 0.8472,
+      "step": 14984
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.03153217966478e-06,
+      "loss": 0.3904,
+      "step": 14985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.030896464974307e-06,
+      "loss": 0.8574,
+      "step": 14986
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0302607879558887e-06,
+      "loss": 0.8184,
+      "step": 14987
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0296251486216853e-06,
+      "loss": 0.8506,
+      "step": 14988
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0289895469838573e-06,
+      "loss": 0.8462,
+      "step": 14989
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.028353983054566e-06,
+      "loss": 0.8594,
+      "step": 14990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0277184568459694e-06,
+      "loss": 0.897,
+      "step": 14991
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0270829683702273e-06,
+      "loss": 0.8301,
+      "step": 14992
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0264475176394957e-06,
+      "loss": 0.8125,
+      "step": 14993
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0258121046659332e-06,
+      "loss": 0.8247,
+      "step": 14994
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0251767294616974e-06,
+      "loss": 0.8384,
+      "step": 14995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.024541392038941e-06,
+      "loss": 0.8076,
+      "step": 14996
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0239060924098225e-06,
+      "loss": 0.811,
+      "step": 14997
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0232708305864934e-06,
+      "loss": 0.8467,
+      "step": 14998
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.022635606581107e-06,
+      "loss": 0.894,
+      "step": 14999
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0220004204058197e-06,
+      "loss": 0.8286,
+      "step": 15000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0213652720727803e-06,
+      "loss": 0.8179,
+      "step": 15001
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0207301615941396e-06,
+      "loss": 0.897,
+      "step": 15002
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.020095088982052e-06,
+      "loss": 0.8472,
+      "step": 15003
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0194600542486654e-06,
+      "loss": 0.8228,
+      "step": 15004
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.018825057406129e-06,
+      "loss": 0.8428,
+      "step": 15005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.018190098466589e-06,
+      "loss": 0.8164,
+      "step": 15006
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.017555177442197e-06,
+      "loss": 0.814,
+      "step": 15007
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0169202943450994e-06,
+      "loss": 0.8579,
+      "step": 15008
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.016285449187439e-06,
+      "loss": 0.8267,
+      "step": 15009
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0156506419813645e-06,
+      "loss": 0.8164,
+      "step": 15010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.015015872739022e-06,
+      "loss": 0.8447,
+      "step": 15011
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0143811414725516e-06,
+      "loss": 0.8564,
+      "step": 15012
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.013746448194101e-06,
+      "loss": 0.7959,
+      "step": 15013
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0131117929158095e-06,
+      "loss": 0.8428,
+      "step": 15014
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0124771756498194e-06,
+      "loss": 0.8472,
+      "step": 15015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0118425964082754e-06,
+      "loss": 0.8955,
+      "step": 15016
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.011208055203314e-06,
+      "loss": 0.8584,
+      "step": 15017
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0105735520470754e-06,
+      "loss": 0.8926,
+      "step": 15018
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.009939086951702e-06,
+      "loss": 0.8823,
+      "step": 15019
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0093046599293283e-06,
+      "loss": 0.8706,
+      "step": 15020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0086702709920946e-06,
+      "loss": 0.7979,
+      "step": 15021
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.008035920152135e-06,
+      "loss": 0.8228,
+      "step": 15022
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0074016074215867e-06,
+      "loss": 0.8335,
+      "step": 15023
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.006767332812589e-06,
+      "loss": 0.9097,
+      "step": 15024
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.006133096337269e-06,
+      "loss": 0.8477,
+      "step": 15025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.005498898007766e-06,
+      "loss": 0.8911,
+      "step": 15026
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.004864737836214e-06,
+      "loss": 0.3137,
+      "step": 15027
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0042306158347422e-06,
+      "loss": 0.8027,
+      "step": 15028
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0035965320154847e-06,
+      "loss": 0.8438,
+      "step": 15029
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.00296248639057e-06,
+      "loss": 0.874,
+      "step": 15030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.00232847897213e-06,
+      "loss": 0.8281,
+      "step": 15031
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.001694509772296e-06,
+      "loss": 0.8779,
+      "step": 15032
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.001060578803192e-06,
+      "loss": 0.8604,
+      "step": 15033
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0004266860769514e-06,
+      "loss": 0.8643,
+      "step": 15034
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.999792831605699e-06,
+      "loss": 0.8823,
+      "step": 15035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.99915901540156e-06,
+      "loss": 0.8882,
+      "step": 15036
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9985252374766637e-06,
+      "loss": 0.8745,
+      "step": 15037
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.997891497843133e-06,
+      "loss": 0.8589,
+      "step": 15038
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9972577965130918e-06,
+      "loss": 0.7461,
+      "step": 15039
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.996624133498667e-06,
+      "loss": 0.8574,
+      "step": 15040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9959905088119777e-06,
+      "loss": 0.8306,
+      "step": 15041
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9953569224651502e-06,
+      "loss": 0.8701,
+      "step": 15042
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.994723374470301e-06,
+      "loss": 0.8496,
+      "step": 15043
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9940898648395544e-06,
+      "loss": 0.7769,
+      "step": 15044
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9934563935850303e-06,
+      "loss": 0.8789,
+      "step": 15045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9928229607188458e-06,
+      "loss": 0.8774,
+      "step": 15046
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.992189566253122e-06,
+      "loss": 0.8081,
+      "step": 15047
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.991556210199977e-06,
+      "loss": 0.8169,
+      "step": 15048
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.990922892571524e-06,
+      "loss": 0.8633,
+      "step": 15049
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9902896133798844e-06,
+      "loss": 0.8291,
+      "step": 15050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9896563726371696e-06,
+      "loss": 0.3534,
+      "step": 15051
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.989023170355496e-06,
+      "loss": 0.8325,
+      "step": 15052
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9883900065469804e-06,
+      "loss": 0.8716,
+      "step": 15053
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9877568812237325e-06,
+      "loss": 0.8335,
+      "step": 15054
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9871237943978652e-06,
+      "loss": 0.8813,
+      "step": 15055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.986490746081494e-06,
+      "loss": 0.8135,
+      "step": 15056
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9858577362867262e-06,
+      "loss": 0.8096,
+      "step": 15057
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.985224765025675e-06,
+      "loss": 0.3472,
+      "step": 15058
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.984591832310447e-06,
+      "loss": 0.8774,
+      "step": 15059
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.983958938153154e-06,
+      "loss": 0.3745,
+      "step": 15060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.983326082565905e-06,
+      "loss": 0.8623,
+      "step": 15061
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.982693265560803e-06,
+      "loss": 0.7637,
+      "step": 15062
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.98206048714996e-06,
+      "loss": 0.855,
+      "step": 15063
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9814277473454797e-06,
+      "loss": 0.8442,
+      "step": 15064
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9807950461594653e-06,
+      "loss": 0.4385,
+      "step": 15065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9801623836040263e-06,
+      "loss": 0.8608,
+      "step": 15066
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.979529759691262e-06,
+      "loss": 0.853,
+      "step": 15067
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.978897174433277e-06,
+      "loss": 0.7949,
+      "step": 15068
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9782646278421757e-06,
+      "loss": 0.8037,
+      "step": 15069
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9776321199300564e-06,
+      "loss": 0.8242,
+      "step": 15070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.976999650709023e-06,
+      "loss": 0.8228,
+      "step": 15071
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.976367220191172e-06,
+      "loss": 0.8823,
+      "step": 15072
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9757348283886063e-06,
+      "loss": 0.8301,
+      "step": 15073
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.975102475313424e-06,
+      "loss": 0.7666,
+      "step": 15074
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9744701609777194e-06,
+      "loss": 0.8394,
+      "step": 15075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9738378853935945e-06,
+      "loss": 0.8389,
+      "step": 15076
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9732056485731443e-06,
+      "loss": 0.7979,
+      "step": 15077
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.972573450528462e-06,
+      "loss": 0.8428,
+      "step": 15078
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9719412912716463e-06,
+      "loss": 0.8564,
+      "step": 15079
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.971309170814788e-06,
+      "loss": 0.8716,
+      "step": 15080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.970677089169982e-06,
+      "loss": 0.814,
+      "step": 15081
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9700450463493224e-06,
+      "loss": 0.8726,
+      "step": 15082
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9694130423648992e-06,
+      "loss": 0.8848,
+      "step": 15083
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9687810772288032e-06,
+      "loss": 0.4399,
+      "step": 15084
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.968149150953128e-06,
+      "loss": 0.8569,
+      "step": 15085
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9675172635499603e-06,
+      "loss": 0.8711,
+      "step": 15086
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.966885415031392e-06,
+      "loss": 0.8242,
+      "step": 15087
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9662536054095057e-06,
+      "loss": 0.8125,
+      "step": 15088
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.965621834696395e-06,
+      "loss": 0.9023,
+      "step": 15089
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9649901029041457e-06,
+      "loss": 0.877,
+      "step": 15090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.96435841004484e-06,
+      "loss": 0.791,
+      "step": 15091
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.963726756130566e-06,
+      "loss": 0.7808,
+      "step": 15092
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9630951411734098e-06,
+      "loss": 0.8237,
+      "step": 15093
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962463565185452e-06,
+      "loss": 0.8638,
+      "step": 15094
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.961832028178777e-06,
+      "loss": 0.8774,
+      "step": 15095
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.961200530165467e-06,
+      "loss": 0.9175,
+      "step": 15096
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.960569071157603e-06,
+      "loss": 0.8184,
+      "step": 15097
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9599376511672684e-06,
+      "loss": 0.8374,
+      "step": 15098
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.95930627020654e-06,
+      "loss": 0.811,
+      "step": 15099
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9586749282874993e-06,
+      "loss": 0.8716,
+      "step": 15100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9580436254222223e-06,
+      "loss": 0.7805,
+      "step": 15101
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.957412361622789e-06,
+      "loss": 0.3633,
+      "step": 15102
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9567811369012765e-06,
+      "loss": 0.8535,
+      "step": 15103
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9561499512697597e-06,
+      "loss": 0.8804,
+      "step": 15104
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9555188047403156e-06,
+      "loss": 0.8818,
+      "step": 15105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9548876973250197e-06,
+      "loss": 0.8599,
+      "step": 15106
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.954256629035942e-06,
+      "loss": 0.8271,
+      "step": 15107
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9536255998851615e-06,
+      "loss": 0.9009,
+      "step": 15108
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.952994609884746e-06,
+      "loss": 0.8462,
+      "step": 15109
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9523636590467697e-06,
+      "loss": 0.8062,
+      "step": 15110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.951732747383304e-06,
+      "loss": 0.8394,
+      "step": 15111
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.951101874906418e-06,
+      "loss": 0.9233,
+      "step": 15112
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.950471041628181e-06,
+      "loss": 0.9062,
+      "step": 15113
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9498402475606648e-06,
+      "loss": 0.8232,
+      "step": 15114
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9492094927159336e-06,
+      "loss": 0.8818,
+      "step": 15115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9485787771060583e-06,
+      "loss": 0.8154,
+      "step": 15116
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.947948100743101e-06,
+      "loss": 0.8311,
+      "step": 15117
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.947317463639131e-06,
+      "loss": 0.8555,
+      "step": 15118
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.946686865806214e-06,
+      "loss": 0.8154,
+      "step": 15119
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.946056307256411e-06,
+      "loss": 0.856,
+      "step": 15120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.945425788001788e-06,
+      "loss": 0.8501,
+      "step": 15121
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.944795308054409e-06,
+      "loss": 0.8696,
+      "step": 15122
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9441648674263314e-06,
+      "loss": 0.8794,
+      "step": 15123
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9435344661296216e-06,
+      "loss": 0.8413,
+      "step": 15124
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9429041041763377e-06,
+      "loss": 0.873,
+      "step": 15125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9422737815785386e-06,
+      "loss": 0.8696,
+      "step": 15126
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9416434983482868e-06,
+      "loss": 0.8442,
+      "step": 15127
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.941013254497637e-06,
+      "loss": 0.8462,
+      "step": 15128
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9403830500386475e-06,
+      "loss": 0.8701,
+      "step": 15129
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9397528849833777e-06,
+      "loss": 0.7925,
+      "step": 15130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.939122759343881e-06,
+      "loss": 0.7788,
+      "step": 15131
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9384926731322137e-06,
+      "loss": 0.8174,
+      "step": 15132
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9378626263604295e-06,
+      "loss": 0.811,
+      "step": 15133
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9372326190405816e-06,
+      "loss": 0.8359,
+      "step": 15134
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9366026511847273e-06,
+      "loss": 0.8223,
+      "step": 15135
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.935972722804913e-06,
+      "loss": 0.8276,
+      "step": 15136
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.935342833913195e-06,
+      "loss": 0.8462,
+      "step": 15137
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9347129845216204e-06,
+      "loss": 0.7915,
+      "step": 15138
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9340831746422417e-06,
+      "loss": 0.8354,
+      "step": 15139
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9334534042871077e-06,
+      "loss": 0.8374,
+      "step": 15140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.932823673468265e-06,
+      "loss": 0.7983,
+      "step": 15141
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.932193982197764e-06,
+      "loss": 0.8408,
+      "step": 15142
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9315643304876517e-06,
+      "loss": 0.8232,
+      "step": 15143
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9309347183499713e-06,
+      "loss": 0.8257,
+      "step": 15144
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.930305145796772e-06,
+      "loss": 0.7891,
+      "step": 15145
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.929675612840096e-06,
+      "loss": 0.8574,
+      "step": 15146
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9290461194919863e-06,
+      "loss": 0.7983,
+      "step": 15147
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9284166657644896e-06,
+      "loss": 0.8015,
+      "step": 15148
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9277872516696466e-06,
+      "loss": 0.9028,
+      "step": 15149
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.927157877219497e-06,
+      "loss": 0.7583,
+      "step": 15150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9265285424260858e-06,
+      "loss": 0.8354,
+      "step": 15151
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9258992473014503e-06,
+      "loss": 0.8462,
+      "step": 15152
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.925269991857631e-06,
+      "loss": 0.8452,
+      "step": 15153
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9246407761066643e-06,
+      "loss": 0.8506,
+      "step": 15154
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9240116000605913e-06,
+      "loss": 0.8784,
+      "step": 15155
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9233824637314484e-06,
+      "loss": 0.8564,
+      "step": 15156
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.922753367131269e-06,
+      "loss": 0.4635,
+      "step": 15157
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.922124310272093e-06,
+      "loss": 0.8535,
+      "step": 15158
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9214952931659533e-06,
+      "loss": 0.8657,
+      "step": 15159
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9208663158248823e-06,
+      "loss": 0.8359,
+      "step": 15160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9202373782609173e-06,
+      "loss": 0.8457,
+      "step": 15161
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.919608480486087e-06,
+      "loss": 0.3367,
+      "step": 15162
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918979622512425e-06,
+      "loss": 0.8301,
+      "step": 15163
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918350804351963e-06,
+      "loss": 0.8862,
+      "step": 15164
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.917722026016729e-06,
+      "loss": 0.8398,
+      "step": 15165
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.917093287518755e-06,
+      "loss": 0.8901,
+      "step": 15166
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.916464588870067e-06,
+      "loss": 0.9136,
+      "step": 15167
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9158359300826948e-06,
+      "loss": 0.8667,
+      "step": 15168
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9152073111686676e-06,
+      "loss": 0.877,
+      "step": 15169
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.91457873214001e-06,
+      "loss": 0.7935,
+      "step": 15170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9139501930087444e-06,
+      "loss": 0.8574,
+      "step": 15171
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.913321693786901e-06,
+      "loss": 0.8711,
+      "step": 15172
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.912693234486501e-06,
+      "loss": 0.9058,
+      "step": 15173
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9120648151195697e-06,
+      "loss": 0.8506,
+      "step": 15174
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9114364356981274e-06,
+      "loss": 0.8384,
+      "step": 15175
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9108080962341977e-06,
+      "loss": 0.8281,
+      "step": 15176
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9101797967398025e-06,
+      "loss": 0.8525,
+      "step": 15177
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.909551537226959e-06,
+      "loss": 0.894,
+      "step": 15178
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.908923317707688e-06,
+      "loss": 0.8652,
+      "step": 15179
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9082951381940117e-06,
+      "loss": 0.8794,
+      "step": 15180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9076669986979433e-06,
+      "loss": 0.8657,
+      "step": 15181
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9070388992315047e-06,
+      "loss": 0.9268,
+      "step": 15182
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9064108398067094e-06,
+      "loss": 0.8389,
+      "step": 15183
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.905782820435572e-06,
+      "loss": 0.3424,
+      "step": 15184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9051548411301116e-06,
+      "loss": 0.8745,
+      "step": 15185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9045269019023382e-06,
+      "loss": 0.3251,
+      "step": 15186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9038990027642666e-06,
+      "loss": 0.9204,
+      "step": 15187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.903271143727913e-06,
+      "loss": 0.8872,
+      "step": 15188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9026433248052852e-06,
+      "loss": 0.811,
+      "step": 15189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9020155460083964e-06,
+      "loss": 0.8569,
+      "step": 15190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9013878073492548e-06,
+      "loss": 0.834,
+      "step": 15191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.900760108839871e-06,
+      "loss": 0.8838,
+      "step": 15192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9001324504922555e-06,
+      "loss": 0.8442,
+      "step": 15193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8995048323184146e-06,
+      "loss": 0.7671,
+      "step": 15194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8988772543303574e-06,
+      "loss": 0.8682,
+      "step": 15195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.898249716540089e-06,
+      "loss": 0.8359,
+      "step": 15196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8976222189596137e-06,
+      "loss": 0.8008,
+      "step": 15197
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.896994761600941e-06,
+      "loss": 0.8799,
+      "step": 15198
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.89636734447607e-06,
+      "loss": 0.8428,
+      "step": 15199
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.895739967597006e-06,
+      "loss": 0.8486,
+      "step": 15200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8951126309757547e-06,
+      "loss": 0.8062,
+      "step": 15201
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.894485334624316e-06,
+      "loss": 0.8291,
+      "step": 15202
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8938580785546907e-06,
+      "loss": 0.8457,
+      "step": 15203
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.893230862778877e-06,
+      "loss": 0.8823,
+      "step": 15204
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.892603687308877e-06,
+      "loss": 0.7852,
+      "step": 15205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8919765521566906e-06,
+      "loss": 0.7993,
+      "step": 15206
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8913494573343132e-06,
+      "loss": 0.8608,
+      "step": 15207
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.890722402853745e-06,
+      "loss": 0.8521,
+      "step": 15208
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8900953887269816e-06,
+      "loss": 0.9214,
+      "step": 15209
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.889468414966016e-06,
+      "loss": 0.8716,
+      "step": 15210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8888414815828474e-06,
+      "loss": 0.8325,
+      "step": 15211
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8882145885894664e-06,
+      "loss": 0.7908,
+      "step": 15212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8875877359978677e-06,
+      "loss": 0.7998,
+      "step": 15213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8869609238200463e-06,
+      "loss": 0.8467,
+      "step": 15214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.886334152067992e-06,
+      "loss": 0.8374,
+      "step": 15215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8857074207536946e-06,
+      "loss": 0.8384,
+      "step": 15216
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8850807298891474e-06,
+      "loss": 0.7974,
+      "step": 15217
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8844540794863376e-06,
+      "loss": 0.8555,
+      "step": 15218
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8838274695572556e-06,
+      "loss": 0.855,
+      "step": 15219
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.883200900113887e-06,
+      "loss": 0.8315,
+      "step": 15220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8825743711682237e-06,
+      "loss": 0.7852,
+      "step": 15221
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8819478827322487e-06,
+      "loss": 0.3409,
+      "step": 15222
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8813214348179466e-06,
+      "loss": 0.8511,
+      "step": 15223
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8806950274373037e-06,
+      "loss": 0.8296,
+      "step": 15224
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.880068660602307e-06,
+      "loss": 0.8291,
+      "step": 15225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.879442334324935e-06,
+      "loss": 0.8164,
+      "step": 15226
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.878816048617175e-06,
+      "loss": 0.8276,
+      "step": 15227
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8781898034910066e-06,
+      "loss": 0.8521,
+      "step": 15228
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.877563598958408e-06,
+      "loss": 0.8818,
+      "step": 15229
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8769374350313655e-06,
+      "loss": 0.8608,
+      "step": 15230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8763113117218533e-06,
+      "loss": 0.8721,
+      "step": 15231
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8756852290418535e-06,
+      "loss": 0.3728,
+      "step": 15232
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.875059187003341e-06,
+      "loss": 0.8955,
+      "step": 15233
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.874433185618297e-06,
+      "loss": 0.9512,
+      "step": 15234
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8738072248986955e-06,
+      "loss": 0.8242,
+      "step": 15235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8731813048565106e-06,
+      "loss": 0.8823,
+      "step": 15236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.872555425503718e-06,
+      "loss": 0.9238,
+      "step": 15237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.871929586852295e-06,
+      "loss": 0.8013,
+      "step": 15238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8713037889142104e-06,
+      "loss": 0.8379,
+      "step": 15239
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8706780317014406e-06,
+      "loss": 0.9087,
+      "step": 15240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8700523152259555e-06,
+      "loss": 0.8022,
+      "step": 15241
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8694266394997238e-06,
+      "loss": 0.8442,
+      "step": 15242
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8688010045347203e-06,
+      "loss": 0.8066,
+      "step": 15243
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8681754103429097e-06,
+      "loss": 0.8833,
+      "step": 15244
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8675498569362633e-06,
+      "loss": 0.292,
+      "step": 15245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8669243443267496e-06,
+      "loss": 0.8682,
+      "step": 15246
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.866298872526333e-06,
+      "loss": 0.8145,
+      "step": 15247
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8656734415469856e-06,
+      "loss": 0.7993,
+      "step": 15248
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865048051400664e-06,
+      "loss": 0.8184,
+      "step": 15249
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8644227020993375e-06,
+      "loss": 0.8032,
+      "step": 15250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8637973936549722e-06,
+      "loss": 0.8955,
+      "step": 15251
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8631721260795265e-06,
+      "loss": 0.8521,
+      "step": 15252
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.862546899384966e-06,
+      "loss": 0.8545,
+      "step": 15253
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.861921713583253e-06,
+      "loss": 0.8843,
+      "step": 15254
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.861296568686347e-06,
+      "loss": 0.8262,
+      "step": 15255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.860671464706207e-06,
+      "loss": 0.8613,
+      "step": 15256
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8600464016547923e-06,
+      "loss": 0.9038,
+      "step": 15257
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.859421379544062e-06,
+      "loss": 0.8657,
+      "step": 15258
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8587963983859746e-06,
+      "loss": 0.8623,
+      "step": 15259
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8581714581924857e-06,
+      "loss": 0.7871,
+      "step": 15260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.857546558975553e-06,
+      "loss": 0.8398,
+      "step": 15261
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8569217007471306e-06,
+      "loss": 0.8413,
+      "step": 15262
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.856296883519172e-06,
+      "loss": 0.8501,
+      "step": 15263
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855672107303633e-06,
+      "loss": 0.8398,
+      "step": 15264
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8550473721124645e-06,
+      "loss": 0.8198,
+      "step": 15265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8544226779576196e-06,
+      "loss": 0.8569,
+      "step": 15266
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8537980248510514e-06,
+      "loss": 0.8364,
+      "step": 15267
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8531734128047094e-06,
+      "loss": 0.8921,
+      "step": 15268
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.852548841830543e-06,
+      "loss": 0.8418,
+      "step": 15269
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.851924311940499e-06,
+      "loss": 0.811,
+      "step": 15270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8512998231465282e-06,
+      "loss": 0.8315,
+      "step": 15271
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.85067537546058e-06,
+      "loss": 0.8853,
+      "step": 15272
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.850050968894596e-06,
+      "loss": 0.8335,
+      "step": 15273
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8494266034605266e-06,
+      "loss": 0.7988,
+      "step": 15274
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8488022791703164e-06,
+      "loss": 0.8232,
+      "step": 15275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8481779960359056e-06,
+      "loss": 0.8486,
+      "step": 15276
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8475537540692426e-06,
+      "loss": 0.8296,
+      "step": 15277
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.846929553282266e-06,
+      "loss": 0.834,
+      "step": 15278
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8463053936869206e-06,
+      "loss": 0.855,
+      "step": 15279
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8456812752951483e-06,
+      "loss": 0.7896,
+      "step": 15280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.845057198118888e-06,
+      "loss": 0.8462,
+      "step": 15281
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.844433162170077e-06,
+      "loss": 0.9116,
+      "step": 15282
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8438091674606594e-06,
+      "loss": 0.8877,
+      "step": 15283
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8431852140025674e-06,
+      "loss": 0.8369,
+      "step": 15284
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8425613018077423e-06,
+      "loss": 0.8462,
+      "step": 15285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8419374308881186e-06,
+      "loss": 0.8232,
+      "step": 15286
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8413136012556336e-06,
+      "loss": 0.3242,
+      "step": 15287
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8406898129222216e-06,
+      "loss": 0.3202,
+      "step": 15288
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8400660658998134e-06,
+      "loss": 0.8101,
+      "step": 15289
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8394423602003455e-06,
+      "loss": 0.8413,
+      "step": 15290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.838818695835752e-06,
+      "loss": 0.2925,
+      "step": 15291
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8381950728179603e-06,
+      "loss": 0.79,
+      "step": 15292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8375714911589057e-06,
+      "loss": 0.8315,
+      "step": 15293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8369479508705156e-06,
+      "loss": 0.8672,
+      "step": 15294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.836324451964718e-06,
+      "loss": 0.8281,
+      "step": 15295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.835700994453445e-06,
+      "loss": 0.8745,
+      "step": 15296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8350775783486205e-06,
+      "loss": 0.8359,
+      "step": 15297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8344542036621753e-06,
+      "loss": 0.8486,
+      "step": 15298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.833830870406032e-06,
+      "loss": 0.8154,
+      "step": 15299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8332075785921192e-06,
+      "loss": 0.9131,
+      "step": 15300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.83258432823236e-06,
+      "loss": 0.8413,
+      "step": 15301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8319611193386766e-06,
+      "loss": 0.8184,
+      "step": 15302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8313379519229923e-06,
+      "loss": 0.853,
+      "step": 15303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.830714825997233e-06,
+      "loss": 0.8408,
+      "step": 15304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8300917415733153e-06,
+      "loss": 0.8447,
+      "step": 15305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.829468698663164e-06,
+      "loss": 0.8545,
+      "step": 15306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.828845697278697e-06,
+      "loss": 0.7993,
+      "step": 15307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8282227374318315e-06,
+      "loss": 0.8657,
+      "step": 15308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.827599819134489e-06,
+      "loss": 0.8765,
+      "step": 15309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8269769423985826e-06,
+      "loss": 0.8096,
+      "step": 15310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8263541072360323e-06,
+      "loss": 0.8428,
+      "step": 15311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8257313136587548e-06,
+      "loss": 0.8027,
+      "step": 15312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8251085616786634e-06,
+      "loss": 0.8843,
+      "step": 15313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8244858513076723e-06,
+      "loss": 0.8921,
+      "step": 15314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8238631825576935e-06,
+      "loss": 0.8613,
+      "step": 15315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.823240555440641e-06,
+      "loss": 0.8853,
+      "step": 15316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8226179699684286e-06,
+      "loss": 0.8135,
+      "step": 15317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8219954261529644e-06,
+      "loss": 0.8804,
+      "step": 15318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.821372924006161e-06,
+      "loss": 0.8604,
+      "step": 15319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.820750463539927e-06,
+      "loss": 0.8906,
+      "step": 15320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8201280447661684e-06,
+      "loss": 0.8447,
+      "step": 15321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8195056676967974e-06,
+      "loss": 0.9287,
+      "step": 15322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.818883332343717e-06,
+      "loss": 0.8125,
+      "step": 15323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.818261038718836e-06,
+      "loss": 0.8257,
+      "step": 15324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8176387868340604e-06,
+      "loss": 0.793,
+      "step": 15325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.817016576701295e-06,
+      "loss": 0.8701,
+      "step": 15326
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.816394408332439e-06,
+      "loss": 0.9043,
+      "step": 15327
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8157722817394017e-06,
+      "loss": 0.8691,
+      "step": 15328
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.815150196934081e-06,
+      "loss": 0.7935,
+      "step": 15329
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8145281539283814e-06,
+      "loss": 0.8428,
+      "step": 15330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8139061527342e-06,
+      "loss": 0.9434,
+      "step": 15331
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8132841933634406e-06,
+      "loss": 0.8291,
+      "step": 15332
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8126622758280008e-06,
+      "loss": 0.8315,
+      "step": 15333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8120404001397765e-06,
+      "loss": 0.8438,
+      "step": 15334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.811418566310669e-06,
+      "loss": 0.8442,
+      "step": 15335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.810796774352571e-06,
+      "loss": 0.8345,
+      "step": 15336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8101750242773807e-06,
+      "loss": 0.9058,
+      "step": 15337
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8095533160969944e-06,
+      "loss": 0.9102,
+      "step": 15338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8089316498233053e-06,
+      "loss": 0.8662,
+      "step": 15339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8083100254682048e-06,
+      "loss": 0.8281,
+      "step": 15340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.807688443043589e-06,
+      "loss": 0.8901,
+      "step": 15341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8070669025613455e-06,
+      "loss": 0.877,
+      "step": 15342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80644540403337e-06,
+      "loss": 0.7871,
+      "step": 15343
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8058239474715488e-06,
+      "loss": 0.8647,
+      "step": 15344
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8052025328877745e-06,
+      "loss": 0.9067,
+      "step": 15345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804581160293935e-06,
+      "loss": 0.8525,
+      "step": 15346
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803959829701915e-06,
+      "loss": 0.8521,
+      "step": 15347
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803338541123604e-06,
+      "loss": 0.7468,
+      "step": 15348
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80271729457089e-06,
+      "loss": 0.9136,
+      "step": 15349
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8020960900556554e-06,
+      "loss": 0.8047,
+      "step": 15350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8014749275897878e-06,
+      "loss": 0.813,
+      "step": 15351
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8008538071851686e-06,
+      "loss": 0.8906,
+      "step": 15352
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80023272885368e-06,
+      "loss": 0.8857,
+      "step": 15353
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.799611692607208e-06,
+      "loss": 0.8706,
+      "step": 15354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.798990698457629e-06,
+      "loss": 0.8501,
+      "step": 15355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.798369746416827e-06,
+      "loss": 0.8354,
+      "step": 15356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7977488364966832e-06,
+      "loss": 0.8237,
+      "step": 15357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7971279687090713e-06,
+      "loss": 0.8247,
+      "step": 15358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7965071430658776e-06,
+      "loss": 0.8628,
+      "step": 15359
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7958863595789703e-06,
+      "loss": 0.8525,
+      "step": 15360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.79526561826023e-06,
+      "loss": 0.7822,
+      "step": 15361
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.794644919121535e-06,
+      "loss": 0.8955,
+      "step": 15362
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.794024262174756e-06,
+      "loss": 0.8423,
+      "step": 15363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7934036474317687e-06,
+      "loss": 0.9102,
+      "step": 15364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7927830749044495e-06,
+      "loss": 0.3494,
+      "step": 15365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7921625446046674e-06,
+      "loss": 0.8579,
+      "step": 15366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7915420565442953e-06,
+      "loss": 0.8521,
+      "step": 15367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790921610735201e-06,
+      "loss": 0.7422,
+      "step": 15368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790301207189258e-06,
+      "loss": 0.8389,
+      "step": 15369
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.789680845918337e-06,
+      "loss": 0.7914,
+      "step": 15370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.789060526934303e-06,
+      "loss": 0.7974,
+      "step": 15371
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.788440250249026e-06,
+      "loss": 0.8545,
+      "step": 15372
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7878200158743727e-06,
+      "loss": 0.8169,
+      "step": 15373
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7871998238222065e-06,
+      "loss": 0.8276,
+      "step": 15374
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.786579674104396e-06,
+      "loss": 0.938,
+      "step": 15375
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7859595667328027e-06,
+      "loss": 0.8521,
+      "step": 15376
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7853395017192918e-06,
+      "loss": 0.8335,
+      "step": 15377
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.784719479075728e-06,
+      "loss": 0.8394,
+      "step": 15378
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7840994988139714e-06,
+      "loss": 0.8188,
+      "step": 15379
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.783479560945883e-06,
+      "loss": 0.876,
+      "step": 15380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.782859665483322e-06,
+      "loss": 0.8701,
+      "step": 15381
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7822398124381488e-06,
+      "loss": 0.8564,
+      "step": 15382
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.781620001822225e-06,
+      "loss": 0.8779,
+      "step": 15383
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.781000233647404e-06,
+      "loss": 0.8682,
+      "step": 15384
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.780380507925548e-06,
+      "loss": 0.8169,
+      "step": 15385
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7797608246685104e-06,
+      "loss": 0.8809,
+      "step": 15386
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7791411838881443e-06,
+      "loss": 0.8291,
+      "step": 15387
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7785215855963087e-06,
+      "loss": 0.8262,
+      "step": 15388
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7779020298048547e-06,
+      "loss": 0.7915,
+      "step": 15389
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7772825165256362e-06,
+      "loss": 0.8784,
+      "step": 15390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.776663045770508e-06,
+      "loss": 0.8022,
+      "step": 15391
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7760436175513184e-06,
+      "loss": 0.8584,
+      "step": 15392
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7754242318799174e-06,
+      "loss": 0.8862,
+      "step": 15393
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7748048887681586e-06,
+      "loss": 0.9048,
+      "step": 15394
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.774185588227887e-06,
+      "loss": 0.8545,
+      "step": 15395
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7735663302709536e-06,
+      "loss": 0.8506,
+      "step": 15396
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.772947114909203e-06,
+      "loss": 0.9185,
+      "step": 15397
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7723279421544857e-06,
+      "loss": 0.8613,
+      "step": 15398
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.771708812018645e-06,
+      "loss": 0.8291,
+      "step": 15399
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7710897245135248e-06,
+      "loss": 0.8618,
+      "step": 15400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7704706796509715e-06,
+      "loss": 0.8193,
+      "step": 15401
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.769851677442826e-06,
+      "loss": 0.8457,
+      "step": 15402
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.769232717900932e-06,
+      "loss": 0.8335,
+      "step": 15403
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.768613801037133e-06,
+      "loss": 0.8877,
+      "step": 15404
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.767994926863269e-06,
+      "loss": 0.8975,
+      "step": 15405
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.767376095391177e-06,
+      "loss": 0.7915,
+      "step": 15406
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7667573066327004e-06,
+      "loss": 0.8247,
+      "step": 15407
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7661385605996737e-06,
+      "loss": 0.8545,
+      "step": 15408
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7655198573039387e-06,
+      "loss": 0.8081,
+      "step": 15409
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7649011967573267e-06,
+      "loss": 0.8379,
+      "step": 15410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7642825789716803e-06,
+      "loss": 0.8916,
+      "step": 15411
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.763664003958831e-06,
+      "loss": 0.8906,
+      "step": 15412
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.763045471730611e-06,
+      "loss": 0.8262,
+      "step": 15413
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.762426982298856e-06,
+      "loss": 0.8413,
+      "step": 15414
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.761808535675401e-06,
+      "loss": 0.7788,
+      "step": 15415
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7611901318720736e-06,
+      "loss": 0.8809,
+      "step": 15416
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7605717709007096e-06,
+      "loss": 0.8979,
+      "step": 15417
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7599534527731352e-06,
+      "loss": 0.8848,
+      "step": 15418
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.75933517750118e-06,
+      "loss": 0.9077,
+      "step": 15419
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7587169450966756e-06,
+      "loss": 0.8022,
+      "step": 15420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7580987555714454e-06,
+      "loss": 0.7605,
+      "step": 15421
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7574806089373197e-06,
+      "loss": 0.3409,
+      "step": 15422
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7568625052061247e-06,
+      "loss": 0.8203,
+      "step": 15423
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.756244444389686e-06,
+      "loss": 0.8101,
+      "step": 15424
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7556264264998255e-06,
+      "loss": 0.8604,
+      "step": 15425
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7550084515483667e-06,
+      "loss": 0.8452,
+      "step": 15426
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7543905195471344e-06,
+      "loss": 0.8037,
+      "step": 15427
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7537726305079515e-06,
+      "loss": 0.8745,
+      "step": 15428
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7531547844426353e-06,
+      "loss": 0.7974,
+      "step": 15429
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7525369813630113e-06,
+      "loss": 0.8628,
+      "step": 15430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7519192212808965e-06,
+      "loss": 0.8901,
+      "step": 15431
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.751301504208107e-06,
+      "loss": 0.8428,
+      "step": 15432
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.750683830156466e-06,
+      "loss": 0.8389,
+      "step": 15433
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.750066199137785e-06,
+      "loss": 0.8574,
+      "step": 15434
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.749448611163884e-06,
+      "loss": 0.8364,
+      "step": 15435
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.748831066246579e-06,
+      "loss": 0.8447,
+      "step": 15436
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7482135643976827e-06,
+      "loss": 0.8555,
+      "step": 15437
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.74759610562901e-06,
+      "loss": 0.7803,
+      "step": 15438
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.746978689952371e-06,
+      "loss": 0.8457,
+      "step": 15439
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7463613173795794e-06,
+      "loss": 0.8418,
+      "step": 15440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7457439879224496e-06,
+      "loss": 0.8481,
+      "step": 15441
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7451267015927875e-06,
+      "loss": 0.8657,
+      "step": 15442
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7445094584024067e-06,
+      "loss": 0.8828,
+      "step": 15443
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7438922583631143e-06,
+      "loss": 0.8691,
+      "step": 15444
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7432751014867164e-06,
+      "loss": 0.8457,
+      "step": 15445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7426579877850235e-06,
+      "loss": 0.8608,
+      "step": 15446
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7420409172698387e-06,
+      "loss": 0.8887,
+      "step": 15447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7414238899529687e-06,
+      "loss": 0.9146,
+      "step": 15448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7408069058462215e-06,
+      "loss": 0.8772,
+      "step": 15449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.740189964961398e-06,
+      "loss": 0.8779,
+      "step": 15450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7395730673102993e-06,
+      "loss": 0.9214,
+      "step": 15451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.738956212904732e-06,
+      "loss": 0.8286,
+      "step": 15452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7383394017564934e-06,
+      "loss": 0.8237,
+      "step": 15453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7377226338773875e-06,
+      "loss": 0.8735,
+      "step": 15454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.737105909279212e-06,
+      "loss": 0.8706,
+      "step": 15455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7364892279737664e-06,
+      "loss": 0.8457,
+      "step": 15456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.73587258997285e-06,
+      "loss": 0.8296,
+      "step": 15457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7352559952882564e-06,
+      "loss": 0.8682,
+      "step": 15458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.734639443931784e-06,
+      "loss": 0.8291,
+      "step": 15459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7340229359152317e-06,
+      "loss": 0.8677,
+      "step": 15460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7334064712503883e-06,
+      "loss": 0.8281,
+      "step": 15461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7327900499490535e-06,
+      "loss": 0.7993,
+      "step": 15462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7321736720230173e-06,
+      "loss": 0.8701,
+      "step": 15463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.731557337484071e-06,
+      "loss": 0.8022,
+      "step": 15464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.730941046344009e-06,
+      "loss": 0.8779,
+      "step": 15465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.730324798614618e-06,
+      "loss": 0.8105,
+      "step": 15466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7297085943076927e-06,
+      "loss": 0.8257,
+      "step": 15467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7290924334350177e-06,
+      "loss": 0.8193,
+      "step": 15468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7284763160083825e-06,
+      "loss": 0.7705,
+      "step": 15469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7278602420395795e-06,
+      "loss": 0.8096,
+      "step": 15470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7272442115403857e-06,
+      "loss": 0.835,
+      "step": 15471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7266282245225916e-06,
+      "loss": 0.416,
+      "step": 15472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.726012280997985e-06,
+      "loss": 0.8188,
+      "step": 15473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7253963809783436e-06,
+      "loss": 0.8057,
+      "step": 15474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7247805244754568e-06,
+      "loss": 0.7993,
+      "step": 15475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.724164711501101e-06,
+      "loss": 0.8506,
+      "step": 15476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7235489420670623e-06,
+      "loss": 0.7666,
+      "step": 15477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7229332161851208e-06,
+      "loss": 0.8203,
+      "step": 15478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7223175338670527e-06,
+      "loss": 0.8853,
+      "step": 15479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7217018951246387e-06,
+      "loss": 0.854,
+      "step": 15480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7210862999696596e-06,
+      "loss": 0.8887,
+      "step": 15481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7204707484138893e-06,
+      "loss": 0.8433,
+      "step": 15482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7198552404691083e-06,
+      "loss": 0.8184,
+      "step": 15483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7192397761470877e-06,
+      "loss": 0.8145,
+      "step": 15484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7186243554596035e-06,
+      "loss": 0.7783,
+      "step": 15485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.718008978418433e-06,
+      "loss": 0.8315,
+      "step": 15486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7173936450353435e-06,
+      "loss": 0.8657,
+      "step": 15487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7167783553221105e-06,
+      "loss": 0.8569,
+      "step": 15488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7161631092905084e-06,
+      "loss": 0.7871,
+      "step": 15489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.715547906952304e-06,
+      "loss": 0.9448,
+      "step": 15490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.714932748319268e-06,
+      "loss": 0.854,
+      "step": 15491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7143176334031685e-06,
+      "loss": 0.8467,
+      "step": 15492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7137025622157733e-06,
+      "loss": 0.7954,
+      "step": 15493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7130875347688533e-06,
+      "loss": 0.8027,
+      "step": 15494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.71247255107417e-06,
+      "loss": 0.876,
+      "step": 15495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7118576111434945e-06,
+      "loss": 0.8096,
+      "step": 15496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.711242714988588e-06,
+      "loss": 0.8247,
+      "step": 15497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7106278626212123e-06,
+      "loss": 0.877,
+      "step": 15498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.710013054053136e-06,
+      "loss": 0.8662,
+      "step": 15499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7093982892961167e-06,
+      "loss": 0.8994,
+      "step": 15500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.708783568361917e-06,
+      "loss": 0.8286,
+      "step": 15501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7081688912623004e-06,
+      "loss": 0.8867,
+      "step": 15502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7075542580090247e-06,
+      "loss": 0.8716,
+      "step": 15503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7069396686138484e-06,
+      "loss": 0.9185,
+      "step": 15504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7063251230885272e-06,
+      "loss": 0.8394,
+      "step": 15505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7057106214448216e-06,
+      "loss": 0.855,
+      "step": 15506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7050961636944886e-06,
+      "loss": 0.8276,
+      "step": 15507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7044817498492804e-06,
+      "loss": 0.8447,
+      "step": 15508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.703867379920955e-06,
+      "loss": 0.9238,
+      "step": 15509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.703253053921266e-06,
+      "loss": 0.8301,
+      "step": 15510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.702638771861963e-06,
+      "loss": 0.8643,
+      "step": 15511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7020245337548017e-06,
+      "loss": 0.8301,
+      "step": 15512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.701410339611531e-06,
+      "loss": 0.8579,
+      "step": 15513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.700796189443903e-06,
+      "loss": 0.8271,
+      "step": 15514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7001820832636677e-06,
+      "loss": 0.8418,
+      "step": 15515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6995680210825737e-06,
+      "loss": 0.9019,
+      "step": 15516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6989540029123673e-06,
+      "loss": 0.8101,
+      "step": 15517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6983400287647975e-06,
+      "loss": 0.834,
+      "step": 15518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.697726098651609e-06,
+      "loss": 0.8584,
+      "step": 15519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6971122125845495e-06,
+      "loss": 0.8218,
+      "step": 15520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6964983705753613e-06,
+      "loss": 0.8428,
+      "step": 15521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6958845726357895e-06,
+      "loss": 0.8623,
+      "step": 15522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6952708187775776e-06,
+      "loss": 0.8306,
+      "step": 15523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6946571090124653e-06,
+      "loss": 0.8452,
+      "step": 15524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6940434433521943e-06,
+      "loss": 0.8433,
+      "step": 15525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6934298218085086e-06,
+      "loss": 0.3718,
+      "step": 15526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6928162443931426e-06,
+      "loss": 0.8965,
+      "step": 15527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.69220271111784e-06,
+      "loss": 0.8096,
+      "step": 15528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.691589221994336e-06,
+      "loss": 0.7979,
+      "step": 15529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6909757770343657e-06,
+      "loss": 0.8574,
+      "step": 15530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6903623762496693e-06,
+      "loss": 0.8643,
+      "step": 15531
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.689749019651978e-06,
+      "loss": 0.8779,
+      "step": 15532
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.689135707253031e-06,
+      "loss": 0.8027,
+      "step": 15533
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.688522439064557e-06,
+      "loss": 0.8706,
+      "step": 15534
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6879092150982933e-06,
+      "loss": 0.8696,
+      "step": 15535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6872960353659693e-06,
+      "loss": 0.876,
+      "step": 15536
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6866828998793154e-06,
+      "loss": 0.8623,
+      "step": 15537
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6860698086500624e-06,
+      "loss": 0.8447,
+      "step": 15538
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6854567616899423e-06,
+      "loss": 0.4133,
+      "step": 15539
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6848437590106804e-06,
+      "loss": 0.8037,
+      "step": 15540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6842308006240058e-06,
+      "loss": 0.8291,
+      "step": 15541
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.683617886541647e-06,
+      "loss": 0.8149,
+      "step": 15542
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6830050167753252e-06,
+      "loss": 0.9219,
+      "step": 15543
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6823921913367706e-06,
+      "loss": 0.9023,
+      "step": 15544
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6817794102377037e-06,
+      "loss": 0.8271,
+      "step": 15545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.68116667348985e-06,
+      "loss": 0.8384,
+      "step": 15546
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6805539811049335e-06,
+      "loss": 0.8232,
+      "step": 15547
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6799413330946744e-06,
+      "loss": 0.8369,
+      "step": 15548
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.679328729470794e-06,
+      "loss": 0.8662,
+      "step": 15549
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6787161702450092e-06,
+      "loss": 0.8354,
+      "step": 15550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6781036554290417e-06,
+      "loss": 0.8296,
+      "step": 15551
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.677491185034613e-06,
+      "loss": 0.8716,
+      "step": 15552
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.676878759073436e-06,
+      "loss": 0.8398,
+      "step": 15553
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.67626637755723e-06,
+      "loss": 0.8071,
+      "step": 15554
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6756540404977104e-06,
+      "loss": 0.3782,
+      "step": 15555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6750417479065905e-06,
+      "loss": 0.8691,
+      "step": 15556
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6744294997955864e-06,
+      "loss": 0.8486,
+      "step": 15557
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.673817296176409e-06,
+      "loss": 0.854,
+      "step": 15558
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6732051370607725e-06,
+      "loss": 0.8535,
+      "step": 15559
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6725930224603903e-06,
+      "loss": 0.7603,
+      "step": 15560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.671980952386971e-06,
+      "loss": 0.8247,
+      "step": 15561
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6713689268522232e-06,
+      "loss": 0.8389,
+      "step": 15562
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6707569458678584e-06,
+      "loss": 0.7925,
+      "step": 15563
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6701450094455827e-06,
+      "loss": 0.8589,
+      "step": 15564
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.669533117597105e-06,
+      "loss": 0.8276,
+      "step": 15565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6689212703341306e-06,
+      "loss": 0.8779,
+      "step": 15566
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6683094676683662e-06,
+      "loss": 0.8447,
+      "step": 15567
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.667697709611517e-06,
+      "loss": 0.8516,
+      "step": 15568
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6670859961752836e-06,
+      "loss": 0.873,
+      "step": 15569
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6664743273713735e-06,
+      "loss": 0.812,
+      "step": 15570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6658627032114835e-06,
+      "loss": 0.8325,
+      "step": 15571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.66525112370732e-06,
+      "loss": 0.7371,
+      "step": 15572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6646395888705825e-06,
+      "loss": 0.8936,
+      "step": 15573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6640280987129695e-06,
+      "loss": 0.8452,
+      "step": 15574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6634166532461786e-06,
+      "loss": 0.8262,
+      "step": 15575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.662805252481911e-06,
+      "loss": 0.8818,
+      "step": 15576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6621938964318593e-06,
+      "loss": 0.8159,
+      "step": 15577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6615825851077247e-06,
+      "loss": 0.8086,
+      "step": 15578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6609713185211984e-06,
+      "loss": 0.9492,
+      "step": 15579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6603600966839766e-06,
+      "loss": 0.8101,
+      "step": 15580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.659748919607757e-06,
+      "loss": 0.8545,
+      "step": 15581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.659137787304224e-06,
+      "loss": 0.856,
+      "step": 15582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6585266997850738e-06,
+      "loss": 0.8345,
+      "step": 15583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.657915657062e-06,
+      "loss": 0.8208,
+      "step": 15584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.657304659146688e-06,
+      "loss": 0.9336,
+      "step": 15585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6566937060508315e-06,
+      "loss": 0.8354,
+      "step": 15586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6560827977861152e-06,
+      "loss": 0.8325,
+      "step": 15587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6554719343642304e-06,
+      "loss": 0.917,
+      "step": 15588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6548611157968618e-06,
+      "loss": 0.835,
+      "step": 15589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6542503420956946e-06,
+      "loss": 0.8296,
+      "step": 15590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6536396132724133e-06,
+      "loss": 0.8594,
+      "step": 15591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.653028929338707e-06,
+      "loss": 0.8887,
+      "step": 15592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6524182903062522e-06,
+      "loss": 0.8066,
+      "step": 15593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.651807696186738e-06,
+      "loss": 0.8184,
+      "step": 15594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.651197146991842e-06,
+      "loss": 0.8477,
+      "step": 15595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6505866427332444e-06,
+      "loss": 0.8096,
+      "step": 15596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6499761834226285e-06,
+      "loss": 0.8242,
+      "step": 15597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.649365769071669e-06,
+      "loss": 0.8589,
+      "step": 15598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.648755399692049e-06,
+      "loss": 0.9238,
+      "step": 15599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6481450752954407e-06,
+      "loss": 0.8418,
+      "step": 15600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.647534795893526e-06,
+      "loss": 0.853,
+      "step": 15601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.646924561497977e-06,
+      "loss": 0.8628,
+      "step": 15602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6463143721204676e-06,
+      "loss": 0.8438,
+      "step": 15603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.645704227772673e-06,
+      "loss": 0.8096,
+      "step": 15604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.645094128466268e-06,
+      "loss": 0.8081,
+      "step": 15605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.644484074212922e-06,
+      "loss": 0.9106,
+      "step": 15606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.643874065024309e-06,
+      "loss": 0.8145,
+      "step": 15607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6432641009120973e-06,
+      "loss": 0.8599,
+      "step": 15608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6426541818879563e-06,
+      "loss": 0.7769,
+      "step": 15609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6420443079635573e-06,
+      "loss": 0.8149,
+      "step": 15610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.641434479150564e-06,
+      "loss": 0.9082,
+      "step": 15611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6408246954606454e-06,
+      "loss": 0.3228,
+      "step": 15612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6402149569054704e-06,
+      "loss": 0.8691,
+      "step": 15613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6396052634967016e-06,
+      "loss": 0.8804,
+      "step": 15614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6389956152460046e-06,
+      "loss": 0.8555,
+      "step": 15615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.638386012165039e-06,
+      "loss": 0.8501,
+      "step": 15616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6377764542654706e-06,
+      "loss": 0.8452,
+      "step": 15617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6371669415589627e-06,
+      "loss": 0.834,
+      "step": 15618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.636557474057173e-06,
+      "loss": 0.8032,
+      "step": 15619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.635948051771765e-06,
+      "loss": 0.8267,
+      "step": 15620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.635338674714397e-06,
+      "loss": 0.8076,
+      "step": 15621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6347293428967236e-06,
+      "loss": 0.8081,
+      "step": 15622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6341200563304066e-06,
+      "loss": 0.8179,
+      "step": 15623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6335108150270994e-06,
+      "loss": 0.8247,
+      "step": 15624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6329016189984595e-06,
+      "loss": 0.8804,
+      "step": 15625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6322924682561436e-06,
+      "loss": 0.8403,
+      "step": 15626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6316833628118046e-06,
+      "loss": 0.7949,
+      "step": 15627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6310743026770917e-06,
+      "loss": 0.8203,
+      "step": 15628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6304652878636628e-06,
+      "loss": 0.8584,
+      "step": 15629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.629856318383165e-06,
+      "loss": 0.9126,
+      "step": 15630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.629247394247253e-06,
+      "loss": 0.9014,
+      "step": 15631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.628638515467572e-06,
+      "loss": 0.8096,
+      "step": 15632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.628029682055775e-06,
+      "loss": 0.8521,
+      "step": 15633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6274208940235075e-06,
+      "loss": 0.8438,
+      "step": 15634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.626812151382415e-06,
+      "loss": 0.8164,
+      "step": 15635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6262034541441485e-06,
+      "loss": 0.8325,
+      "step": 15636
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.625594802320348e-06,
+      "loss": 0.7993,
+      "step": 15637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.624986195922661e-06,
+      "loss": 0.8887,
+      "step": 15638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6243776349627315e-06,
+      "loss": 0.8643,
+      "step": 15639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6237691194522026e-06,
+      "loss": 0.8022,
+      "step": 15640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6231606494027125e-06,
+      "loss": 0.8687,
+      "step": 15641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6225522248259063e-06,
+      "loss": 0.8789,
+      "step": 15642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.621943845733421e-06,
+      "loss": 0.8735,
+      "step": 15643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.621335512136899e-06,
+      "loss": 0.8882,
+      "step": 15644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.620727224047975e-06,
+      "loss": 0.8257,
+      "step": 15645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.62011898147829e-06,
+      "loss": 0.7412,
+      "step": 15646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6195107844394795e-06,
+      "loss": 0.7974,
+      "step": 15647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6189026329431765e-06,
+      "loss": 0.8389,
+      "step": 15648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.618294527001019e-06,
+      "loss": 0.9116,
+      "step": 15649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.617686466624642e-06,
+      "loss": 0.8735,
+      "step": 15650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.617078451825675e-06,
+      "loss": 0.8335,
+      "step": 15651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6164704826157537e-06,
+      "loss": 0.8315,
+      "step": 15652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.615862559006509e-06,
+      "loss": 0.8789,
+      "step": 15653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.615254681009568e-06,
+      "loss": 0.8706,
+      "step": 15654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.614646848636565e-06,
+      "loss": 0.8457,
+      "step": 15655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.614039061899124e-06,
+      "loss": 0.8672,
+      "step": 15656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.613431320808876e-06,
+      "loss": 0.8237,
+      "step": 15657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6128236253774496e-06,
+      "loss": 0.8765,
+      "step": 15658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.612215975616469e-06,
+      "loss": 0.8188,
+      "step": 15659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6116083715375594e-06,
+      "loss": 0.873,
+      "step": 15660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6110008131523435e-06,
+      "loss": 0.8232,
+      "step": 15661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.610393300472447e-06,
+      "loss": 0.854,
+      "step": 15662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6097858335094934e-06,
+      "loss": 0.8169,
+      "step": 15663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.609178412275102e-06,
+      "loss": 0.8364,
+      "step": 15664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.608571036780896e-06,
+      "loss": 0.8433,
+      "step": 15665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.607963707038495e-06,
+      "loss": 0.8384,
+      "step": 15666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.607356423059516e-06,
+      "loss": 0.8882,
+      "step": 15667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6067491848555804e-06,
+      "loss": 0.896,
+      "step": 15668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6061419924383025e-06,
+      "loss": 0.8867,
+      "step": 15669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6055348458193006e-06,
+      "loss": 0.7886,
+      "step": 15670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6049277450101928e-06,
+      "loss": 0.7734,
+      "step": 15671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6043206900225905e-06,
+      "loss": 0.811,
+      "step": 15672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.603713680868109e-06,
+      "loss": 0.8413,
+      "step": 15673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6031067175583593e-06,
+      "loss": 0.8896,
+      "step": 15674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.602499800104954e-06,
+      "loss": 0.8389,
+      "step": 15675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6018929285195084e-06,
+      "loss": 0.8428,
+      "step": 15676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.601286102813628e-06,
+      "loss": 0.8452,
+      "step": 15677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.600679322998926e-06,
+      "loss": 0.8174,
+      "step": 15678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6000725890870092e-06,
+      "loss": 0.8379,
+      "step": 15679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.599465901089484e-06,
+      "loss": 0.8198,
+      "step": 15680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.598859259017961e-06,
+      "loss": 0.8101,
+      "step": 15681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5982526628840417e-06,
+      "loss": 0.7852,
+      "step": 15682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5976461126993336e-06,
+      "loss": 0.8081,
+      "step": 15683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5970396084754434e-06,
+      "loss": 0.8555,
+      "step": 15684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5964331502239716e-06,
+      "loss": 0.834,
+      "step": 15685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5958267379565193e-06,
+      "loss": 0.8462,
+      "step": 15686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5952203716846923e-06,
+      "loss": 0.8755,
+      "step": 15687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.594614051420087e-06,
+      "loss": 0.8892,
+      "step": 15688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5940077771743077e-06,
+      "loss": 0.8496,
+      "step": 15689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.593401548958948e-06,
+      "loss": 0.8687,
+      "step": 15690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.59279536678561e-06,
+      "loss": 0.8232,
+      "step": 15691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5921892306658937e-06,
+      "loss": 0.8398,
+      "step": 15692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.591583140611387e-06,
+      "loss": 0.8232,
+      "step": 15693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5909770966336905e-06,
+      "loss": 0.4187,
+      "step": 15694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5903710987444003e-06,
+      "loss": 0.8281,
+      "step": 15695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.589765146955106e-06,
+      "loss": 0.8535,
+      "step": 15696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5891592412774045e-06,
+      "loss": 0.8564,
+      "step": 15697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5885533817228836e-06,
+      "loss": 0.8613,
+      "step": 15698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5879475683031387e-06,
+      "loss": 0.8374,
+      "step": 15699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.587341801029758e-06,
+      "loss": 0.8062,
+      "step": 15700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5867360799143283e-06,
+      "loss": 0.8496,
+      "step": 15701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5861304049684414e-06,
+      "loss": 0.3704,
+      "step": 15702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5855247762036828e-06,
+      "loss": 0.8784,
+      "step": 15703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5849191936316397e-06,
+      "loss": 0.8115,
+      "step": 15704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5843136572638993e-06,
+      "loss": 0.8931,
+      "step": 15705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5837081671120464e-06,
+      "loss": 0.8091,
+      "step": 15706
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5831027231876615e-06,
+      "loss": 0.8218,
+      "step": 15707
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.582497325502332e-06,
+      "loss": 0.8389,
+      "step": 15708
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5818919740676357e-06,
+      "loss": 0.3302,
+      "step": 15709
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.581286668895159e-06,
+      "loss": 0.4362,
+      "step": 15710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.580681409996477e-06,
+      "loss": 0.7759,
+      "step": 15711
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.580076197383175e-06,
+      "loss": 0.8364,
+      "step": 15712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.579471031066828e-06,
+      "loss": 0.8633,
+      "step": 15713
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.578865911059013e-06,
+      "loss": 0.8735,
+      "step": 15714
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.578260837371308e-06,
+      "loss": 0.8105,
+      "step": 15715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5776558100152914e-06,
+      "loss": 0.8149,
+      "step": 15716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5770508290025344e-06,
+      "loss": 0.4342,
+      "step": 15717
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5764458943446148e-06,
+      "loss": 0.7969,
+      "step": 15718
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5758410060531046e-06,
+      "loss": 0.8638,
+      "step": 15719
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.575236164139574e-06,
+      "loss": 0.7832,
+      "step": 15720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5746313686155983e-06,
+      "loss": 0.3737,
+      "step": 15721
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5740266194927444e-06,
+      "loss": 0.9111,
+      "step": 15722
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.573421916782585e-06,
+      "loss": 0.8711,
+      "step": 15723
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5728172604966888e-06,
+      "loss": 0.8306,
+      "step": 15724
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5722126506466237e-06,
+      "loss": 0.4016,
+      "step": 15725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5716080872439564e-06,
+      "loss": 0.7786,
+      "step": 15726
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5710035703002512e-06,
+      "loss": 0.875,
+      "step": 15727
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.570399099827075e-06,
+      "loss": 0.918,
+      "step": 15728
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.569794675835995e-06,
+      "loss": 0.8013,
+      "step": 15729
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5691902983385697e-06,
+      "loss": 0.8301,
+      "step": 15730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.568585967346367e-06,
+      "loss": 0.8037,
+      "step": 15731
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.567981682870946e-06,
+      "loss": 0.8882,
+      "step": 15732
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5673774449238666e-06,
+      "loss": 0.8506,
+      "step": 15733
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5667732535166914e-06,
+      "loss": 0.8633,
+      "step": 15734
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5661691086609752e-06,
+      "loss": 0.8204,
+      "step": 15735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5655650103682806e-06,
+      "loss": 0.8931,
+      "step": 15736
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5649609586501643e-06,
+      "loss": 0.8579,
+      "step": 15737
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5643569535181834e-06,
+      "loss": 0.8042,
+      "step": 15738
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5637529949838913e-06,
+      "loss": 0.8213,
+      "step": 15739
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.563149083058841e-06,
+      "loss": 0.8916,
+      "step": 15740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5625452177545895e-06,
+      "loss": 0.8135,
+      "step": 15741
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5619413990826907e-06,
+      "loss": 0.8735,
+      "step": 15742
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.561337627054693e-06,
+      "loss": 0.8306,
+      "step": 15743
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5607339016821507e-06,
+      "loss": 0.8457,
+      "step": 15744
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5601302229766133e-06,
+      "loss": 0.8159,
+      "step": 15745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5595265909496276e-06,
+      "loss": 0.7974,
+      "step": 15746
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.558923005612746e-06,
+      "loss": 0.8394,
+      "step": 15747
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5583194669775114e-06,
+      "loss": 0.8169,
+      "step": 15748
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.557715975055473e-06,
+      "loss": 0.8574,
+      "step": 15749
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.557112529858179e-06,
+      "loss": 0.8691,
+      "step": 15750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5565091313971714e-06,
+      "loss": 0.7778,
+      "step": 15751
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5559057796839925e-06,
+      "loss": 0.8643,
+      "step": 15752
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.55530247473019e-06,
+      "loss": 0.8569,
+      "step": 15753
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.554699216547301e-06,
+      "loss": 0.9038,
+      "step": 15754
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.554096005146872e-06,
+      "loss": 0.8335,
+      "step": 15755
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.553492840540438e-06,
+      "loss": 0.8735,
+      "step": 15756
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.552889722739543e-06,
+      "loss": 0.8301,
+      "step": 15757
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.552286651755724e-06,
+      "loss": 0.7759,
+      "step": 15758
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5516836276005176e-06,
+      "loss": 0.8062,
+      "step": 15759
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.55108065028546e-06,
+      "loss": 0.8716,
+      "step": 15760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.550477719822091e-06,
+      "loss": 0.8188,
+      "step": 15761
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.549874836221942e-06,
+      "loss": 0.8433,
+      "step": 15762
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5492719994965494e-06,
+      "loss": 0.8652,
+      "step": 15763
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5486692096574452e-06,
+      "loss": 0.7832,
+      "step": 15764
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5480664667161603e-06,
+      "loss": 0.8379,
+      "step": 15765
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5474637706842297e-06,
+      "loss": 0.8516,
+      "step": 15766
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5468611215731796e-06,
+      "loss": 0.8516,
+      "step": 15767
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5462585193945438e-06,
+      "loss": 0.8721,
+      "step": 15768
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5456559641598478e-06,
+      "loss": 0.8809,
+      "step": 15769
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.545053455880622e-06,
+      "loss": 0.9043,
+      "step": 15770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5444509945683924e-06,
+      "loss": 0.8164,
+      "step": 15771
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5438485802346825e-06,
+      "loss": 0.8354,
+      "step": 15772
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.54324621289102e-06,
+      "loss": 0.8071,
+      "step": 15773
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5426438925489307e-06,
+      "loss": 0.8735,
+      "step": 15774
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5420416192199345e-06,
+      "loss": 0.8613,
+      "step": 15775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5414393929155567e-06,
+      "loss": 0.8354,
+      "step": 15776
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.540837213647318e-06,
+      "loss": 0.7368,
+      "step": 15777
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5402350814267364e-06,
+      "loss": 0.8442,
+      "step": 15778
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5396329962653352e-06,
+      "loss": 0.8516,
+      "step": 15779
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5390309581746307e-06,
+      "loss": 0.7593,
+      "step": 15780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5384289671661416e-06,
+      "loss": 0.8467,
+      "step": 15781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5378270232513875e-06,
+      "loss": 0.8296,
+      "step": 15782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.537225126441882e-06,
+      "loss": 0.8511,
+      "step": 15783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.536623276749141e-06,
+      "loss": 0.8198,
+      "step": 15784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5360214741846766e-06,
+      "loss": 0.9102,
+      "step": 15785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5354197187600038e-06,
+      "loss": 0.8726,
+      "step": 15786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5348180104866375e-06,
+      "loss": 0.874,
+      "step": 15787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5342163493760857e-06,
+      "loss": 0.8135,
+      "step": 15788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.533614735439862e-06,
+      "loss": 0.8813,
+      "step": 15789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5330131686894756e-06,
+      "loss": 0.8267,
+      "step": 15790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5324116491364328e-06,
+      "loss": 0.8638,
+      "step": 15791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5318101767922453e-06,
+      "loss": 0.8584,
+      "step": 15792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.531208751668417e-06,
+      "loss": 0.8252,
+      "step": 15793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5306073737764547e-06,
+      "loss": 0.9014,
+      "step": 15794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.530006043127867e-06,
+      "loss": 0.8174,
+      "step": 15795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5294047597341564e-06,
+      "loss": 0.9229,
+      "step": 15796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5288035236068233e-06,
+      "loss": 0.833,
+      "step": 15797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5282023347573753e-06,
+      "loss": 0.855,
+      "step": 15798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5276011931973097e-06,
+      "loss": 0.8164,
+      "step": 15799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.527000098938131e-06,
+      "loss": 0.8774,
+      "step": 15800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5263990519913357e-06,
+      "loss": 0.8071,
+      "step": 15801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5257980523684243e-06,
+      "loss": 0.3341,
+      "step": 15802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.525197100080899e-06,
+      "loss": 0.8521,
+      "step": 15803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5245961951402485e-06,
+      "loss": 0.8745,
+      "step": 15804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5239953375579764e-06,
+      "loss": 0.8389,
+      "step": 15805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.523394527345573e-06,
+      "loss": 0.854,
+      "step": 15806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5227937645145344e-06,
+      "loss": 0.8525,
+      "step": 15807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.522193049076357e-06,
+      "loss": 0.8828,
+      "step": 15808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.521592381042529e-06,
+      "loss": 0.8955,
+      "step": 15809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.520991760424547e-06,
+      "loss": 0.8306,
+      "step": 15810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5203911872338984e-06,
+      "loss": 0.7993,
+      "step": 15811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5197906614820718e-06,
+      "loss": 0.8843,
+      "step": 15812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51919018318056e-06,
+      "loss": 0.8462,
+      "step": 15813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5185897523408478e-06,
+      "loss": 0.8086,
+      "step": 15814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5179893689744227e-06,
+      "loss": 0.7598,
+      "step": 15815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5173890330927753e-06,
+      "loss": 0.793,
+      "step": 15816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.516788744707387e-06,
+      "loss": 0.894,
+      "step": 15817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5161885038297417e-06,
+      "loss": 0.9282,
+      "step": 15818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5155883104713262e-06,
+      "loss": 0.8281,
+      "step": 15819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.514988164643619e-06,
+      "loss": 0.8398,
+      "step": 15820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5143880663581066e-06,
+      "loss": 0.8418,
+      "step": 15821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.513788015626265e-06,
+      "loss": 0.8745,
+      "step": 15822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5131880124595785e-06,
+      "loss": 0.8174,
+      "step": 15823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.512588056869524e-06,
+      "loss": 0.8691,
+      "step": 15824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5119881488675773e-06,
+      "loss": 0.8223,
+      "step": 15825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5113882884652185e-06,
+      "loss": 0.8193,
+      "step": 15826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5107884756739252e-06,
+      "loss": 0.7998,
+      "step": 15827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.510188710505169e-06,
+      "loss": 0.7661,
+      "step": 15828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5095889929704274e-06,
+      "loss": 0.8501,
+      "step": 15829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5089893230811737e-06,
+      "loss": 0.8384,
+      "step": 15830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.508389700848877e-06,
+      "loss": 0.8701,
+      "step": 15831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5077901262850134e-06,
+      "loss": 0.8564,
+      "step": 15832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5071905994010504e-06,
+      "loss": 0.7861,
+      "step": 15833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5065911202084614e-06,
+      "loss": 0.894,
+      "step": 15834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.505991688718712e-06,
+      "loss": 0.8413,
+      "step": 15835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5053923049432726e-06,
+      "loss": 0.8901,
+      "step": 15836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.50479296889361e-06,
+      "loss": 0.8462,
+      "step": 15837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5041936805811875e-06,
+      "loss": 0.8369,
+      "step": 15838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5035944400174732e-06,
+      "loss": 0.8555,
+      "step": 15839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.502995247213933e-06,
+      "loss": 0.8989,
+      "step": 15840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5023961021820265e-06,
+      "loss": 0.7939,
+      "step": 15841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5017970049332203e-06,
+      "loss": 0.7803,
+      "step": 15842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.501197955478973e-06,
+      "loss": 0.8257,
+      "step": 15843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5005989538307464e-06,
+      "loss": 0.8999,
+      "step": 15844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5000000000000015e-06,
+      "loss": 0.8267,
+      "step": 15845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.499401093998194e-06,
+      "loss": 0.8369,
+      "step": 15846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4988022358367837e-06,
+      "loss": 0.9644,
+      "step": 15847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4982034255272303e-06,
+      "loss": 0.3453,
+      "step": 15848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.497604663080988e-06,
+      "loss": 0.8594,
+      "step": 15849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4970059485095114e-06,
+      "loss": 0.8774,
+      "step": 15850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4964072818242537e-06,
+      "loss": 0.8765,
+      "step": 15851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4958086630366694e-06,
+      "loss": 0.8149,
+      "step": 15852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4952100921582135e-06,
+      "loss": 0.8867,
+      "step": 15853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.494611569200334e-06,
+      "loss": 0.8696,
+      "step": 15854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.494013094174485e-06,
+      "loss": 0.8438,
+      "step": 15855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4934146670921145e-06,
+      "loss": 0.8418,
+      "step": 15856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.492816287964669e-06,
+      "loss": 0.8818,
+      "step": 15857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.492217956803601e-06,
+      "loss": 0.9087,
+      "step": 15858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.491619673620353e-06,
+      "loss": 0.812,
+      "step": 15859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4910214384263735e-06,
+      "loss": 0.812,
+      "step": 15860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.49042325123311e-06,
+      "loss": 0.8486,
+      "step": 15861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4898251120520035e-06,
+      "loss": 0.9009,
+      "step": 15862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.489227020894497e-06,
+      "loss": 0.9033,
+      "step": 15863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.488628977772035e-06,
+      "loss": 0.9009,
+      "step": 15864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.488030982696057e-06,
+      "loss": 0.3317,
+      "step": 15865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4874330356780064e-06,
+      "loss": 0.8066,
+      "step": 15866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.486835136729319e-06,
+      "loss": 0.8345,
+      "step": 15867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.486237285861438e-06,
+      "loss": 0.8164,
+      "step": 15868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4856394830857987e-06,
+      "loss": 0.8408,
+      "step": 15869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4850417284138363e-06,
+      "loss": 0.7988,
+      "step": 15870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4844440218569905e-06,
+      "loss": 0.8608,
+      "step": 15871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.483846363426693e-06,
+      "loss": 0.8643,
+      "step": 15872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.483248753134379e-06,
+      "loss": 0.7891,
+      "step": 15873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.482651190991485e-06,
+      "loss": 0.8613,
+      "step": 15874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.48205367700944e-06,
+      "loss": 0.9019,
+      "step": 15875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4814562111996744e-06,
+      "loss": 0.873,
+      "step": 15876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.480858793573622e-06,
+      "loss": 0.8887,
+      "step": 15877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.480261424142708e-06,
+      "loss": 0.8403,
+      "step": 15878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.479664102918366e-06,
+      "loss": 0.8774,
+      "step": 15879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.479066829912019e-06,
+      "loss": 0.8252,
+      "step": 15880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4784696051350986e-06,
+      "loss": 0.8467,
+      "step": 15881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4778724285990276e-06,
+      "loss": 0.8027,
+      "step": 15882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.47727530031523e-06,
+      "loss": 0.2887,
+      "step": 15883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4766782202951305e-06,
+      "loss": 0.8379,
+      "step": 15884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.476081188550155e-06,
+      "loss": 0.8477,
+      "step": 15885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4754842050917223e-06,
+      "loss": 0.9043,
+      "step": 15886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4748872699312565e-06,
+      "loss": 0.8926,
+      "step": 15887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.474290383080176e-06,
+      "loss": 0.7915,
+      "step": 15888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.473693544549899e-06,
+      "loss": 0.8335,
+      "step": 15889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.473096754351847e-06,
+      "loss": 0.9072,
+      "step": 15890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.472500012497434e-06,
+      "loss": 0.8008,
+      "step": 15891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4719033189980792e-06,
+      "loss": 0.8584,
+      "step": 15892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.471306673865199e-06,
+      "loss": 0.7725,
+      "step": 15893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.470710077110207e-06,
+      "loss": 0.8369,
+      "step": 15894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.470113528744517e-06,
+      "loss": 0.8047,
+      "step": 15895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.46951702877954e-06,
+      "loss": 0.8218,
+      "step": 15896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4689205772266895e-06,
+      "loss": 0.8291,
+      "step": 15897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.468324174097379e-06,
+      "loss": 0.3727,
+      "step": 15898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.467727819403015e-06,
+      "loss": 0.7983,
+      "step": 15899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.467131513155009e-06,
+      "loss": 0.8149,
+      "step": 15900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4665352553647693e-06,
+      "loss": 0.7656,
+      "step": 15901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4659390460437e-06,
+      "loss": 0.8345,
+      "step": 15902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.465342885203212e-06,
+      "loss": 0.8467,
+      "step": 15903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4647467728547068e-06,
+      "loss": 0.8696,
+      "step": 15904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4641507090095912e-06,
+      "loss": 0.7959,
+      "step": 15905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.46355469367927e-06,
+      "loss": 0.8379,
+      "step": 15906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4629587268751447e-06,
+      "loss": 0.8755,
+      "step": 15907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.462362808608617e-06,
+      "loss": 0.8584,
+      "step": 15908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4617669388910854e-06,
+      "loss": 0.834,
+      "step": 15909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.461171117733952e-06,
+      "loss": 0.8413,
+      "step": 15910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.460575345148618e-06,
+      "loss": 0.896,
+      "step": 15911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4599796211464772e-06,
+      "loss": 0.7993,
+      "step": 15912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4593839457389285e-06,
+      "loss": 0.8936,
+      "step": 15913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.458788318937373e-06,
+      "loss": 0.8208,
+      "step": 15914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.458192740753197e-06,
+      "loss": 0.8608,
+      "step": 15915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4575972111978014e-06,
+      "loss": 0.833,
+      "step": 15916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4570017302825753e-06,
+      "loss": 0.8105,
+      "step": 15917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4564062980189135e-06,
+      "loss": 0.8545,
+      "step": 15918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4558109144182097e-06,
+      "loss": 0.8418,
+      "step": 15919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4552155794918497e-06,
+      "loss": 0.8262,
+      "step": 15920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4546202932512283e-06,
+      "loss": 0.8477,
+      "step": 15921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4540250557077313e-06,
+      "loss": 0.8164,
+      "step": 15922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4534298668727453e-06,
+      "loss": 0.8364,
+      "step": 15923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4528347267576604e-06,
+      "loss": 0.8706,
+      "step": 15924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.452239635373859e-06,
+      "loss": 0.8687,
+      "step": 15925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4516445927327287e-06,
+      "loss": 0.876,
+      "step": 15926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4510495988456544e-06,
+      "loss": 0.7666,
+      "step": 15927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4504546537240186e-06,
+      "loss": 0.814,
+      "step": 15928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4498597573792006e-06,
+      "loss": 0.8647,
+      "step": 15929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4492649098225855e-06,
+      "loss": 0.8691,
+      "step": 15930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4486701110655507e-06,
+      "loss": 0.3229,
+      "step": 15931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4480753611194793e-06,
+      "loss": 0.8223,
+      "step": 15932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.447480659995745e-06,
+      "loss": 0.8716,
+      "step": 15933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.44688600770573e-06,
+      "loss": 0.8286,
+      "step": 15934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.446291404260809e-06,
+      "loss": 0.8115,
+      "step": 15935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4456968496723553e-06,
+      "loss": 0.8276,
+      "step": 15936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4451023439517485e-06,
+      "loss": 0.8594,
+      "step": 15937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4445078871103578e-06,
+      "loss": 0.8218,
+      "step": 15938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.443913479159558e-06,
+      "loss": 0.8203,
+      "step": 15939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.443319120110723e-06,
+      "loss": 0.9028,
+      "step": 15940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.442724809975222e-06,
+      "loss": 0.7993,
+      "step": 15941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4421305487644237e-06,
+      "loss": 0.8569,
+      "step": 15942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4415363364897006e-06,
+      "loss": 0.8467,
+      "step": 15943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4409421731624166e-06,
+      "loss": 0.8257,
+      "step": 15944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4403480587939437e-06,
+      "loss": 0.8047,
+      "step": 15945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.439753993395644e-06,
+      "loss": 0.8018,
+      "step": 15946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4391599769788867e-06,
+      "loss": 0.7803,
+      "step": 15947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4385660095550345e-06,
+      "loss": 0.8599,
+      "step": 15948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.437972091135449e-06,
+      "loss": 0.8291,
+      "step": 15949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.437378221731494e-06,
+      "loss": 0.8467,
+      "step": 15950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4367844013545343e-06,
+      "loss": 0.3226,
+      "step": 15951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.436190630015926e-06,
+      "loss": 0.8325,
+      "step": 15952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4355969077270337e-06,
+      "loss": 0.8579,
+      "step": 15953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4350032344992125e-06,
+      "loss": 0.7646,
+      "step": 15954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.43440961034382e-06,
+      "loss": 0.8091,
+      "step": 15955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.433816035272216e-06,
+      "loss": 0.7998,
+      "step": 15956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4332225092957535e-06,
+      "loss": 0.8428,
+      "step": 15957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4326290324257896e-06,
+      "loss": 0.8608,
+      "step": 15958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4320356046736793e-06,
+      "loss": 0.895,
+      "step": 15959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4314422260507743e-06,
+      "loss": 0.8848,
+      "step": 15960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4308488965684273e-06,
+      "loss": 0.8486,
+      "step": 15961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4302556162379877e-06,
+      "loss": 0.8188,
+      "step": 15962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4296623850708073e-06,
+      "loss": 0.8682,
+      "step": 15963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4290692030782374e-06,
+      "loss": 0.8677,
+      "step": 15964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4284760702716235e-06,
+      "loss": 0.8545,
+      "step": 15965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4278829866623166e-06,
+      "loss": 0.8535,
+      "step": 15966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.427289952261661e-06,
+      "loss": 0.8701,
+      "step": 15967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.426696967081001e-06,
+      "loss": 0.8306,
+      "step": 15968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.426104031131685e-06,
+      "loss": 0.9023,
+      "step": 15969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4255111444250527e-06,
+      "loss": 0.7744,
+      "step": 15970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4249183069724493e-06,
+      "loss": 0.8047,
+      "step": 15971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4243255187852184e-06,
+      "loss": 0.8921,
+      "step": 15972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.423732779874699e-06,
+      "loss": 0.7939,
+      "step": 15973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4231400902522313e-06,
+      "loss": 0.8374,
+      "step": 15974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4225474499291525e-06,
+      "loss": 0.814,
+      "step": 15975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4219548589168024e-06,
+      "loss": 0.4156,
+      "step": 15976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.42136231722652e-06,
+      "loss": 0.8438,
+      "step": 15977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.420769824869638e-06,
+      "loss": 0.3093,
+      "step": 15978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4201773818574956e-06,
+      "loss": 0.856,
+      "step": 15979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4195849882014256e-06,
+      "loss": 0.8628,
+      "step": 15980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4189926439127586e-06,
+      "loss": 0.8491,
+      "step": 15981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.418400349002832e-06,
+      "loss": 0.8477,
+      "step": 15982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.417808103482972e-06,
+      "loss": 0.8906,
+      "step": 15983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.417215907364513e-06,
+      "loss": 0.8813,
+      "step": 15984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.416623760658785e-06,
+      "loss": 0.7448,
+      "step": 15985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.416031663377115e-06,
+      "loss": 0.8818,
+      "step": 15986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.41543961553083e-06,
+      "loss": 0.7603,
+      "step": 15987
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4148476171312596e-06,
+      "loss": 0.9102,
+      "step": 15988
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4142556681897257e-06,
+      "loss": 0.8643,
+      "step": 15989
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4136637687175573e-06,
+      "loss": 0.7905,
+      "step": 15990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.413071918726075e-06,
+      "loss": 0.8672,
+      "step": 15991
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4124801182266057e-06,
+      "loss": 0.8159,
+      "step": 15992
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4118883672304688e-06,
+      "loss": 0.8589,
+      "step": 15993
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4112966657489836e-06,
+      "loss": 0.8418,
+      "step": 15994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.410705013793473e-06,
+      "loss": 0.8535,
+      "step": 15995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4101134113752573e-06,
+      "loss": 0.873,
+      "step": 15996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4095218585056514e-06,
+      "loss": 0.8462,
+      "step": 15997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.408930355195976e-06,
+      "loss": 0.8237,
+      "step": 15998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.408338901457546e-06,
+      "loss": 0.8237,
+      "step": 15999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.407747497301675e-06,
+      "loss": 0.8291,
+      "step": 16000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.407156142739681e-06,
+      "loss": 0.8345,
+      "step": 16001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4065648377828742e-06,
+      "loss": 0.7993,
+      "step": 16002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4059735824425705e-06,
+      "loss": 0.8921,
+      "step": 16003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4053823767300783e-06,
+      "loss": 0.8623,
+      "step": 16004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4047912206567116e-06,
+      "loss": 0.8589,
+      "step": 16005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4042001142337777e-06,
+      "loss": 0.9155,
+      "step": 16006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.403609057472585e-06,
+      "loss": 0.8569,
+      "step": 16007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4030180503844426e-06,
+      "loss": 0.8804,
+      "step": 16008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.402427092980658e-06,
+      "loss": 0.8496,
+      "step": 16009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4018361852725357e-06,
+      "loss": 0.8613,
+      "step": 16010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4012453272713828e-06,
+      "loss": 0.9023,
+      "step": 16011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4006545189885016e-06,
+      "loss": 0.8687,
+      "step": 16012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.400063760435194e-06,
+      "loss": 0.9014,
+      "step": 16013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3994730516227656e-06,
+      "loss": 0.8374,
+      "step": 16014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3988823925625134e-06,
+      "loss": 0.8525,
+      "step": 16015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39829178326574e-06,
+      "loss": 0.8325,
+      "step": 16016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.397701223743747e-06,
+      "loss": 0.8066,
+      "step": 16017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.397110714007829e-06,
+      "loss": 0.8184,
+      "step": 16018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3965202540692863e-06,
+      "loss": 0.8628,
+      "step": 16019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.395929843939411e-06,
+      "loss": 0.8306,
+      "step": 16020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3953394836295013e-06,
+      "loss": 0.8354,
+      "step": 16021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3947491731508533e-06,
+      "loss": 0.8271,
+      "step": 16022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3941589125147575e-06,
+      "loss": 0.8794,
+      "step": 16023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39356870173251e-06,
+      "loss": 0.873,
+      "step": 16024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3929785408154e-06,
+      "loss": 0.8618,
+      "step": 16025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3923884297747175e-06,
+      "loss": 0.812,
+      "step": 16026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3917983686217545e-06,
+      "loss": 0.9238,
+      "step": 16027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3912083573677976e-06,
+      "loss": 0.7964,
+      "step": 16028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.390618396024135e-06,
+      "loss": 0.8315,
+      "step": 16029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3900284846020566e-06,
+      "loss": 0.8171,
+      "step": 16030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3894386231128442e-06,
+      "loss": 0.8623,
+      "step": 16031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3888488115677865e-06,
+      "loss": 0.8379,
+      "step": 16032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.388259049978166e-06,
+      "loss": 0.8862,
+      "step": 16033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3876693383552634e-06,
+      "loss": 0.8022,
+      "step": 16034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.387079676710365e-06,
+      "loss": 0.8989,
+      "step": 16035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3864900650547473e-06,
+      "loss": 0.7651,
+      "step": 16036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3859005033996934e-06,
+      "loss": 0.8374,
+      "step": 16037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3853109917564843e-06,
+      "loss": 0.8589,
+      "step": 16038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3847215301363958e-06,
+      "loss": 0.9189,
+      "step": 16039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3841321185507057e-06,
+      "loss": 0.3986,
+      "step": 16040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3835427570106884e-06,
+      "loss": 0.3698,
+      "step": 16041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.382953445527621e-06,
+      "loss": 0.8315,
+      "step": 16042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3823641841127803e-06,
+      "loss": 0.8755,
+      "step": 16043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3817749727774357e-06,
+      "loss": 0.8193,
+      "step": 16044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3811858115328634e-06,
+      "loss": 0.8418,
+      "step": 16045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3805967003903336e-06,
+      "loss": 0.8013,
+      "step": 16046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3800076393611142e-06,
+      "loss": 0.8711,
+      "step": 16047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3794186284564795e-06,
+      "loss": 0.8262,
+      "step": 16048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3788296676876943e-06,
+      "loss": 0.8594,
+      "step": 16049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.378240757066028e-06,
+      "loss": 0.8662,
+      "step": 16050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.377651896602749e-06,
+      "loss": 0.8472,
+      "step": 16051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3770630863091215e-06,
+      "loss": 0.8726,
+      "step": 16052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3764743261964086e-06,
+      "loss": 0.8599,
+      "step": 16053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3758856162758777e-06,
+      "loss": 0.8193,
+      "step": 16054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.375296956558788e-06,
+      "loss": 0.7871,
+      "step": 16055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.374708347056406e-06,
+      "loss": 0.7949,
+      "step": 16056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3741197877799883e-06,
+      "loss": 0.8101,
+      "step": 16057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.373531278740798e-06,
+      "loss": 0.7689,
+      "step": 16058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.372942819950094e-06,
+      "loss": 0.8682,
+      "step": 16059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3723544114191316e-06,
+      "loss": 0.8896,
+      "step": 16060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3717660531591697e-06,
+      "loss": 0.8359,
+      "step": 16061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3711777451814665e-06,
+      "loss": 0.855,
+      "step": 16062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3705894874972747e-06,
+      "loss": 0.8135,
+      "step": 16063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.37000128011785e-06,
+      "loss": 0.8457,
+      "step": 16064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.369413123054446e-06,
+      "loss": 0.8135,
+      "step": 16065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.368825016318312e-06,
+      "loss": 0.8667,
+      "step": 16066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.368236959920704e-06,
+      "loss": 0.8452,
+      "step": 16067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3676489538728675e-06,
+      "loss": 0.8687,
+      "step": 16068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.367060998186057e-06,
+      "loss": 0.7778,
+      "step": 16069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3664730928715164e-06,
+      "loss": 0.8359,
+      "step": 16070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3658852379404973e-06,
+      "loss": 0.7998,
+      "step": 16071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.365297433404245e-06,
+      "loss": 0.8325,
+      "step": 16072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.364709679274002e-06,
+      "loss": 0.3914,
+      "step": 16073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3641219755610163e-06,
+      "loss": 0.8037,
+      "step": 16074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3635343222765322e-06,
+      "loss": 0.8525,
+      "step": 16075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.362946719431789e-06,
+      "loss": 0.8145,
+      "step": 16076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3623591670380324e-06,
+      "loss": 0.8765,
+      "step": 16077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.361771665106502e-06,
+      "loss": 0.8574,
+      "step": 16078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.361184213648435e-06,
+      "loss": 0.8257,
+      "step": 16079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.360596812675073e-06,
+      "loss": 0.8633,
+      "step": 16080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3600094621976527e-06,
+      "loss": 0.8369,
+      "step": 16081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.359422162227411e-06,
+      "loss": 0.9438,
+      "step": 16082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.358834912775586e-06,
+      "loss": 0.8496,
+      "step": 16083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.358247713853411e-06,
+      "loss": 0.853,
+      "step": 16084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.357660565472121e-06,
+      "loss": 0.8682,
+      "step": 16085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3570734676429453e-06,
+      "loss": 0.8555,
+      "step": 16086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3564864203771195e-06,
+      "loss": 0.8545,
+      "step": 16087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3558994236858755e-06,
+      "loss": 0.8936,
+      "step": 16088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3553124775804403e-06,
+      "loss": 0.835,
+      "step": 16089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3547255820720473e-06,
+      "loss": 0.8296,
+      "step": 16090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.354138737171922e-06,
+      "loss": 0.3922,
+      "step": 16091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.35355194289129e-06,
+      "loss": 0.8477,
+      "step": 16092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.352965199241382e-06,
+      "loss": 0.8564,
+      "step": 16093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.352378506233419e-06,
+      "loss": 0.8135,
+      "step": 16094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3517918638786267e-06,
+      "loss": 0.9106,
+      "step": 16095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3512052721882316e-06,
+      "loss": 0.856,
+      "step": 16096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3506187311734536e-06,
+      "loss": 0.8159,
+      "step": 16097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3500322408455117e-06,
+      "loss": 0.9487,
+      "step": 16098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3494458012156312e-06,
+      "loss": 0.7637,
+      "step": 16099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.348859412295027e-06,
+      "loss": 0.8403,
+      "step": 16100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3482730740949223e-06,
+      "loss": 0.8057,
+      "step": 16101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.34768678662653e-06,
+      "loss": 0.8862,
+      "step": 16102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3471005499010712e-06,
+      "loss": 0.834,
+      "step": 16103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3465143639297593e-06,
+      "loss": 0.8662,
+      "step": 16104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.345928228723807e-06,
+      "loss": 0.855,
+      "step": 16105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.345342144294432e-06,
+      "loss": 0.8477,
+      "step": 16106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3447561106528432e-06,
+      "loss": 0.8511,
+      "step": 16107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.344170127810254e-06,
+      "loss": 0.8364,
+      "step": 16108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.343584195777877e-06,
+      "loss": 0.8511,
+      "step": 16109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3429983145669207e-06,
+      "loss": 0.8511,
+      "step": 16110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3424124841885917e-06,
+      "loss": 0.8501,
+      "step": 16111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.341826704654101e-06,
+      "loss": 0.8384,
+      "step": 16112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.341240975974653e-06,
+      "loss": 0.8076,
+      "step": 16113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.340655298161456e-06,
+      "loss": 0.8037,
+      "step": 16114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3400696712257127e-06,
+      "loss": 0.7998,
+      "step": 16115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.339484095178629e-06,
+      "loss": 0.8896,
+      "step": 16116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.338898570031408e-06,
+      "loss": 0.8203,
+      "step": 16117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.338313095795248e-06,
+      "loss": 0.8218,
+      "step": 16118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3377276724813537e-06,
+      "loss": 0.8579,
+      "step": 16119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3371423001009254e-06,
+      "loss": 0.8032,
+      "step": 16120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.336556978665159e-06,
+      "loss": 0.896,
+      "step": 16121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.335971708185257e-06,
+      "loss": 0.8013,
+      "step": 16122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3353864886724144e-06,
+      "loss": 0.8501,
+      "step": 16123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.334801320137825e-06,
+      "loss": 0.8296,
+      "step": 16124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3342162025926886e-06,
+      "loss": 0.853,
+      "step": 16125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.333631136048194e-06,
+      "loss": 0.8838,
+      "step": 16126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.333046120515538e-06,
+      "loss": 0.8516,
+      "step": 16127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3324611560059146e-06,
+      "loss": 0.9058,
+      "step": 16128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3318762425305125e-06,
+      "loss": 0.8472,
+      "step": 16129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.331291380100522e-06,
+      "loss": 0.8525,
+      "step": 16130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3307065687271313e-06,
+      "loss": 0.8633,
+      "step": 16131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.33012180842153e-06,
+      "loss": 0.772,
+      "step": 16132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.329537099194908e-06,
+      "loss": 0.8867,
+      "step": 16133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3289524410584468e-06,
+      "loss": 0.8052,
+      "step": 16134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3283678340233366e-06,
+      "loss": 0.8506,
+      "step": 16135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3277832781007597e-06,
+      "loss": 0.8462,
+      "step": 16136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.327198773301898e-06,
+      "loss": 0.8296,
+      "step": 16137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3266143196379376e-06,
+      "loss": 0.8389,
+      "step": 16138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3260299171200557e-06,
+      "loss": 0.8945,
+      "step": 16139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3254455657594356e-06,
+      "loss": 0.8726,
+      "step": 16140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3248612655672575e-06,
+      "loss": 0.8569,
+      "step": 16141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3242770165546973e-06,
+      "loss": 0.8218,
+      "step": 16142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3236928187329383e-06,
+      "loss": 0.8765,
+      "step": 16143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.323108672113149e-06,
+      "loss": 0.8604,
+      "step": 16144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.322524576706509e-06,
+      "loss": 0.8623,
+      "step": 16145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.321940532524195e-06,
+      "loss": 0.8296,
+      "step": 16146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.321356539577377e-06,
+      "loss": 0.8911,
+      "step": 16147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3207725978772296e-06,
+      "loss": 0.833,
+      "step": 16148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3201887074349256e-06,
+      "loss": 0.897,
+      "step": 16149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3196048682616346e-06,
+      "loss": 0.877,
+      "step": 16150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3190210803685266e-06,
+      "loss": 0.8701,
+      "step": 16151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.318437343766768e-06,
+      "loss": 0.9043,
+      "step": 16152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.317853658467528e-06,
+      "loss": 0.4385,
+      "step": 16153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.317270024481977e-06,
+      "loss": 0.8301,
+      "step": 16154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.316686441821275e-06,
+      "loss": 0.8218,
+      "step": 16155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3161029104965925e-06,
+      "loss": 0.8672,
+      "step": 16156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3155194305190902e-06,
+      "loss": 0.8496,
+      "step": 16157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3149360018999294e-06,
+      "loss": 0.8394,
+      "step": 16158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3143526246502764e-06,
+      "loss": 0.8203,
+      "step": 16159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3137692987812875e-06,
+      "loss": 0.8638,
+      "step": 16160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.313186024304125e-06,
+      "loss": 0.8066,
+      "step": 16161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3126028012299495e-06,
+      "loss": 0.8281,
+      "step": 16162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3120196295699172e-06,
+      "loss": 0.8496,
+      "step": 16163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.311436509335183e-06,
+      "loss": 0.8511,
+      "step": 16164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3108534405369078e-06,
+      "loss": 0.8315,
+      "step": 16165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.310270423186241e-06,
+      "loss": 0.8335,
+      "step": 16166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3096874572943416e-06,
+      "loss": 0.8804,
+      "step": 16167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.309104542872359e-06,
+      "loss": 0.8594,
+      "step": 16168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3085216799314487e-06,
+      "loss": 0.9136,
+      "step": 16169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3079388684827596e-06,
+      "loss": 0.8193,
+      "step": 16170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.307356108537441e-06,
+      "loss": 0.8159,
+      "step": 16171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3067734001066437e-06,
+      "loss": 0.9092,
+      "step": 16172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3061907432015142e-06,
+      "loss": 0.8501,
+      "step": 16173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3056081378332007e-06,
+      "loss": 0.8838,
+      "step": 16174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.305025584012851e-06,
+      "loss": 0.834,
+      "step": 16175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.304443081751609e-06,
+      "loss": 0.8486,
+      "step": 16176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.303860631060616e-06,
+      "loss": 0.855,
+      "step": 16177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.30327823195102e-06,
+      "loss": 0.8511,
+      "step": 16178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.302695884433959e-06,
+      "loss": 0.7847,
+      "step": 16179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.302113588520578e-06,
+      "loss": 0.8335,
+      "step": 16180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.301531344222014e-06,
+      "loss": 0.9619,
+      "step": 16181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.300949151549409e-06,
+      "loss": 0.8257,
+      "step": 16182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3003670105139e-06,
+      "loss": 0.8447,
+      "step": 16183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.299784921126622e-06,
+      "loss": 0.8101,
+      "step": 16184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2992028833987136e-06,
+      "loss": 0.8423,
+      "step": 16185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.298620897341312e-06,
+      "loss": 0.832,
+      "step": 16186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2980389629655476e-06,
+      "loss": 0.2793,
+      "step": 16187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2974570802825572e-06,
+      "loss": 0.8379,
+      "step": 16188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.296875249303472e-06,
+      "loss": 0.8271,
+      "step": 16189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2962934700394207e-06,
+      "loss": 0.8662,
+      "step": 16190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.295711742501538e-06,
+      "loss": 0.8467,
+      "step": 16191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2951300667009493e-06,
+      "loss": 0.8262,
+      "step": 16192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.294548442648784e-06,
+      "loss": 0.8062,
+      "step": 16193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.293966870356173e-06,
+      "loss": 0.8364,
+      "step": 16194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2933853498342394e-06,
+      "loss": 0.9043,
+      "step": 16195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2928038810941095e-06,
+      "loss": 0.8335,
+      "step": 16196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2922224641469047e-06,
+      "loss": 0.8535,
+      "step": 16197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.291641099003752e-06,
+      "loss": 0.8574,
+      "step": 16198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2910597856757744e-06,
+      "loss": 0.8101,
+      "step": 16199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.29047852417409e-06,
+      "loss": 0.8379,
+      "step": 16200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.289897314509823e-06,
+      "loss": 0.8442,
+      "step": 16201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2893161566940914e-06,
+      "loss": 0.8804,
+      "step": 16202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2887350507380106e-06,
+      "loss": 0.8013,
+      "step": 16203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.288153996652703e-06,
+      "loss": 0.8325,
+      "step": 16204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2875729944492807e-06,
+      "loss": 0.8228,
+      "step": 16205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2869920441388615e-06,
+      "loss": 0.8696,
+      "step": 16206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2864111457325614e-06,
+      "loss": 0.7759,
+      "step": 16207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2858302992414926e-06,
+      "loss": 0.8149,
+      "step": 16208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2852495046767668e-06,
+      "loss": 0.854,
+      "step": 16209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.284668762049494e-06,
+      "loss": 0.8169,
+      "step": 16210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2840880713707865e-06,
+      "loss": 0.834,
+      "step": 16211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.283507432651756e-06,
+      "loss": 0.8457,
+      "step": 16212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.282926845903507e-06,
+      "loss": 0.8467,
+      "step": 16213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2823463111371513e-06,
+      "loss": 0.8003,
+      "step": 16214
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2817658283637926e-06,
+      "loss": 0.9229,
+      "step": 16215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.281185397594535e-06,
+      "loss": 0.8486,
+      "step": 16216
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2806050188404875e-06,
+      "loss": 0.8481,
+      "step": 16217
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.280024692112749e-06,
+      "loss": 0.8862,
+      "step": 16218
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2794444174224245e-06,
+      "loss": 0.8354,
+      "step": 16219
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2788641947806173e-06,
+      "loss": 0.8115,
+      "step": 16220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2782840241984254e-06,
+      "loss": 0.8584,
+      "step": 16221
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2777039056869483e-06,
+      "loss": 0.8457,
+      "step": 16222
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2771238392572858e-06,
+      "loss": 0.8237,
+      "step": 16223
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.276543824920534e-06,
+      "loss": 0.8013,
+      "step": 16224
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2759638626877923e-06,
+      "loss": 0.8115,
+      "step": 16225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.275383952570152e-06,
+      "loss": 0.8027,
+      "step": 16226
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2748040945787124e-06,
+      "loss": 0.8901,
+      "step": 16227
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2742242887245653e-06,
+      "loss": 0.8735,
+      "step": 16228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2736445350188003e-06,
+      "loss": 0.9194,
+      "step": 16229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.273064833472512e-06,
+      "loss": 0.8574,
+      "step": 16230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.272485184096792e-06,
+      "loss": 0.8242,
+      "step": 16231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2719055869027266e-06,
+      "loss": 0.792,
+      "step": 16232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2713260419014083e-06,
+      "loss": 0.8862,
+      "step": 16233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.270746549103923e-06,
+      "loss": 0.8745,
+      "step": 16234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2701671085213543e-06,
+      "loss": 0.8115,
+      "step": 16235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.269587720164793e-06,
+      "loss": 0.8687,
+      "step": 16236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2690083840453194e-06,
+      "loss": 0.8022,
+      "step": 16237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2684291001740207e-06,
+      "loss": 0.875,
+      "step": 16238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2678498685619756e-06,
+      "loss": 0.8174,
+      "step": 16239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2672706892202702e-06,
+      "loss": 0.8552,
+      "step": 16240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.266691562159982e-06,
+      "loss": 0.8184,
+      "step": 16241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.26611248739219e-06,
+      "loss": 0.8555,
+      "step": 16242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.265533464927974e-06,
+      "loss": 0.8599,
+      "step": 16243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.264954494778414e-06,
+      "loss": 0.8076,
+      "step": 16244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.264375576954582e-06,
+      "loss": 0.873,
+      "step": 16245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2637967114675587e-06,
+      "loss": 0.8335,
+      "step": 16246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.263217898328415e-06,
+      "loss": 0.8296,
+      "step": 16247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.262639137548225e-06,
+      "loss": 0.8325,
+      "step": 16248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.262060429138063e-06,
+      "loss": 0.8315,
+      "step": 16249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.261481773108997e-06,
+      "loss": 0.9121,
+      "step": 16250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2609031694721008e-06,
+      "loss": 0.7959,
+      "step": 16251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2603246182384448e-06,
+      "loss": 0.4332,
+      "step": 16252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.259746119419094e-06,
+      "loss": 0.8618,
+      "step": 16253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.259167673025122e-06,
+      "loss": 0.8345,
+      "step": 16254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2585892790675878e-06,
+      "loss": 0.8003,
+      "step": 16255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.258010937557561e-06,
+      "loss": 0.8418,
+      "step": 16256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2574326485061073e-06,
+      "loss": 0.8745,
+      "step": 16257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2568544119242864e-06,
+      "loss": 0.8643,
+      "step": 16258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2562762278231637e-06,
+      "loss": 0.8691,
+      "step": 16259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.255698096213802e-06,
+      "loss": 0.874,
+      "step": 16260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2551200171072605e-06,
+      "loss": 0.8252,
+      "step": 16261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.254541990514598e-06,
+      "loss": 0.8359,
+      "step": 16262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2539640164468724e-06,
+      "loss": 0.856,
+      "step": 16263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.253386094915142e-06,
+      "loss": 0.8877,
+      "step": 16264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2528082259304656e-06,
+      "loss": 0.8569,
+      "step": 16265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.252230409503895e-06,
+      "loss": 0.8281,
+      "step": 16266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2516526456464886e-06,
+      "loss": 0.8086,
+      "step": 16267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2510749343692977e-06,
+      "loss": 0.8716,
+      "step": 16268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2504972756833733e-06,
+      "loss": 0.7856,
+      "step": 16269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2499196695997704e-06,
+      "loss": 0.8394,
+      "step": 16270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2493421161295364e-06,
+      "loss": 0.8269,
+      "step": 16271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.248764615283722e-06,
+      "loss": 0.8828,
+      "step": 16272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.248187167073378e-06,
+      "loss": 0.8867,
+      "step": 16273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.247609771509549e-06,
+      "loss": 0.8809,
+      "step": 16274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.247032428603283e-06,
+      "loss": 0.7866,
+      "step": 16275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.246455138365623e-06,
+      "loss": 0.8643,
+      "step": 16276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.245877900807615e-06,
+      "loss": 0.8091,
+      "step": 16277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2453007159403045e-06,
+      "loss": 0.8159,
+      "step": 16278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2447235837747306e-06,
+      "loss": 0.834,
+      "step": 16279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.244146504321938e-06,
+      "loss": 0.8271,
+      "step": 16280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.243569477592965e-06,
+      "loss": 0.8013,
+      "step": 16281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.24299250359885e-06,
+      "loss": 0.8555,
+      "step": 16282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.242415582350635e-06,
+      "loss": 0.8784,
+      "step": 16283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2418387138593535e-06,
+      "loss": 0.8701,
+      "step": 16284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2412618981360433e-06,
+      "loss": 0.8564,
+      "step": 16285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2406851351917424e-06,
+      "loss": 0.8135,
+      "step": 16286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.240108425037483e-06,
+      "loss": 0.8525,
+      "step": 16287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2395317676842964e-06,
+      "loss": 0.812,
+      "step": 16288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2389551631432186e-06,
+      "loss": 0.7646,
+      "step": 16289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.238378611425278e-06,
+      "loss": 0.8159,
+      "step": 16290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.237802112541508e-06,
+      "loss": 0.8184,
+      "step": 16291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2372256665029352e-06,
+      "loss": 0.873,
+      "step": 16292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2366492733205903e-06,
+      "loss": 0.8257,
+      "step": 16293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2360729330054986e-06,
+      "loss": 0.9136,
+      "step": 16294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.235496645568686e-06,
+      "loss": 0.8159,
+      "step": 16295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.234920411021179e-06,
+      "loss": 0.8647,
+      "step": 16296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.234344229374003e-06,
+      "loss": 0.8174,
+      "step": 16297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2337681006381788e-06,
+      "loss": 0.833,
+      "step": 16298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2331920248247313e-06,
+      "loss": 0.9009,
+      "step": 16299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.232616001944681e-06,
+      "loss": 0.7239,
+      "step": 16300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2320400320090445e-06,
+      "loss": 0.9028,
+      "step": 16301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2314641150288462e-06,
+      "loss": 0.8481,
+      "step": 16302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.230888251015101e-06,
+      "loss": 0.8638,
+      "step": 16303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.230312439978828e-06,
+      "loss": 0.8086,
+      "step": 16304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2297366819310412e-06,
+      "loss": 0.8042,
+      "step": 16305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2291609768827583e-06,
+      "loss": 0.895,
+      "step": 16306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2285853248449927e-06,
+      "loss": 0.834,
+      "step": 16307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2280097258287554e-06,
+      "loss": 0.8755,
+      "step": 16308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2274341798450594e-06,
+      "loss": 0.8379,
+      "step": 16309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2268586869049193e-06,
+      "loss": 0.853,
+      "step": 16310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2262832470193404e-06,
+      "loss": 0.8501,
+      "step": 16311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2257078601993347e-06,
+      "loss": 0.8521,
+      "step": 16312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2251325264559105e-06,
+      "loss": 0.8232,
+      "step": 16313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2245572458000714e-06,
+      "loss": 0.876,
+      "step": 16314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.223982018242828e-06,
+      "loss": 0.854,
+      "step": 16315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2234068437951807e-06,
+      "loss": 0.8745,
+      "step": 16316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2228317224681363e-06,
+      "loss": 0.8042,
+      "step": 16317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.222256654272699e-06,
+      "loss": 0.8203,
+      "step": 16318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.221681639219869e-06,
+      "loss": 0.8223,
+      "step": 16319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.221106677320647e-06,
+      "loss": 0.8354,
+      "step": 16320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.220531768586032e-06,
+      "loss": 0.8628,
+      "step": 16321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.219956913027023e-06,
+      "loss": 0.8779,
+      "step": 16322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.219382110654622e-06,
+      "loss": 0.8506,
+      "step": 16323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.218807361479821e-06,
+      "loss": 0.896,
+      "step": 16324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.218232665513619e-06,
+      "loss": 0.8618,
+      "step": 16325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2176580227670107e-06,
+      "loss": 0.8574,
+      "step": 16326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2170834332509867e-06,
+      "loss": 0.8711,
+      "step": 16327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.216508896976543e-06,
+      "loss": 0.8237,
+      "step": 16328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2159344139546697e-06,
+      "loss": 0.8994,
+      "step": 16329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2153599841963567e-06,
+      "loss": 0.8823,
+      "step": 16330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2147856077125986e-06,
+      "loss": 0.8662,
+      "step": 16331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.21421128451438e-06,
+      "loss": 0.7788,
+      "step": 16332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2136370146126874e-06,
+      "loss": 0.8086,
+      "step": 16333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2130627980185115e-06,
+      "loss": 0.7993,
+      "step": 16334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2124886347428344e-06,
+      "loss": 0.8135,
+      "step": 16335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.211914524796644e-06,
+      "loss": 0.8271,
+      "step": 16336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.21134046819092e-06,
+      "loss": 0.7363,
+      "step": 16337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2107664649366494e-06,
+      "loss": 0.8027,
+      "step": 16338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2101925150448115e-06,
+      "loss": 0.8369,
+      "step": 16339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.209618618526385e-06,
+      "loss": 0.8516,
+      "step": 16340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.209044775392354e-06,
+      "loss": 0.8833,
+      "step": 16341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.208470985653692e-06,
+      "loss": 0.8398,
+      "step": 16342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2078972493213784e-06,
+      "loss": 0.8442,
+      "step": 16343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.207323566406393e-06,
+      "loss": 0.8384,
+      "step": 16344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2067499369197082e-06,
+      "loss": 0.8667,
+      "step": 16345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.206176360872297e-06,
+      "loss": 0.8701,
+      "step": 16346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.205602838275136e-06,
+      "loss": 0.8472,
+      "step": 16347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2050293691391944e-06,
+      "loss": 0.8667,
+      "step": 16348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2044559534754472e-06,
+      "loss": 0.8394,
+      "step": 16349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2038825912948613e-06,
+      "loss": 0.8257,
+      "step": 16350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.203309282608409e-06,
+      "loss": 0.8066,
+      "step": 16351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.202736027427057e-06,
+      "loss": 0.8623,
+      "step": 16352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2021628257617715e-06,
+      "loss": 0.8506,
+      "step": 16353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2015896776235195e-06,
+      "loss": 0.8311,
+      "step": 16354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2010165830232687e-06,
+      "loss": 0.8984,
+      "step": 16355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.200443541971979e-06,
+      "loss": 0.833,
+      "step": 16356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.199870554480619e-06,
+      "loss": 0.8633,
+      "step": 16357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.199297620560147e-06,
+      "loss": 0.7959,
+      "step": 16358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.198724740221523e-06,
+      "loss": 0.9019,
+      "step": 16359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1981519134757106e-06,
+      "loss": 0.8623,
+      "step": 16360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1975791403336654e-06,
+      "loss": 0.8667,
+      "step": 16361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.197006420806348e-06,
+      "loss": 0.7153,
+      "step": 16362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.196433754904716e-06,
+      "loss": 0.8652,
+      "step": 16363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1958611426397224e-06,
+      "loss": 0.8975,
+      "step": 16364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1952885840223274e-06,
+      "loss": 0.8491,
+      "step": 16365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1947160790634783e-06,
+      "loss": 0.9004,
+      "step": 16366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1941436277741308e-06,
+      "loss": 0.8215,
+      "step": 16367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.193571230165239e-06,
+      "loss": 0.8472,
+      "step": 16368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1929988862477504e-06,
+      "loss": 0.7847,
+      "step": 16369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.192426596032616e-06,
+      "loss": 0.8999,
+      "step": 16370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.191854359530787e-06,
+      "loss": 0.8169,
+      "step": 16371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1912821767532095e-06,
+      "loss": 0.7827,
+      "step": 16372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1907100477108302e-06,
+      "loss": 0.8057,
+      "step": 16373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1901379724145922e-06,
+      "loss": 0.834,
+      "step": 16374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1895659508754425e-06,
+      "loss": 0.8438,
+      "step": 16375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1889939831043265e-06,
+      "loss": 0.7949,
+      "step": 16376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.188422069112184e-06,
+      "loss": 0.8018,
+      "step": 16377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.187850208909959e-06,
+      "loss": 0.833,
+      "step": 16378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.187278402508591e-06,
+      "loss": 0.8091,
+      "step": 16379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.186706649919018e-06,
+      "loss": 0.8115,
+      "step": 16380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1861349511521817e-06,
+      "loss": 0.7793,
+      "step": 16381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1855633062190164e-06,
+      "loss": 0.8438,
+      "step": 16382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1849917151304595e-06,
+      "loss": 0.9141,
+      "step": 16383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.184420177897449e-06,
+      "loss": 0.8066,
+      "step": 16384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1838486945309173e-06,
+      "loss": 0.7734,
+      "step": 16385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1832772650417976e-06,
+      "loss": 0.8154,
+      "step": 16386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1827058894410208e-06,
+      "loss": 0.854,
+      "step": 16387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.18213456773952e-06,
+      "loss": 0.835,
+      "step": 16388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1815632999482267e-06,
+      "loss": 0.8379,
+      "step": 16389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1809920860780675e-06,
+      "loss": 0.8086,
+      "step": 16390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1804209261399733e-06,
+      "loss": 0.8418,
+      "step": 16391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.17984982014487e-06,
+      "loss": 0.8467,
+      "step": 16392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.179278768103682e-06,
+      "loss": 0.895,
+      "step": 16393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.178707770027337e-06,
+      "loss": 0.9058,
+      "step": 16394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.178136825926758e-06,
+      "loss": 0.8413,
+      "step": 16395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.177565935812867e-06,
+      "loss": 0.8472,
+      "step": 16396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1769950996965895e-06,
+      "loss": 0.8149,
+      "step": 16397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.176424317588844e-06,
+      "loss": 0.8364,
+      "step": 16398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.175853589500549e-06,
+      "loss": 0.8208,
+      "step": 16399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.175282915442627e-06,
+      "loss": 0.833,
+      "step": 16400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174712295425992e-06,
+      "loss": 0.8164,
+      "step": 16401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1741417294615653e-06,
+      "loss": 0.8408,
+      "step": 16402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1735712175602576e-06,
+      "loss": 0.8579,
+      "step": 16403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.173000759732989e-06,
+      "loss": 0.8516,
+      "step": 16404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.17243035599067e-06,
+      "loss": 0.814,
+      "step": 16405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1718600063442125e-06,
+      "loss": 0.8818,
+      "step": 16406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171289710804532e-06,
+      "loss": 0.8013,
+      "step": 16407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.170719469382534e-06,
+      "loss": 0.8599,
+      "step": 16408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1701492820891316e-06,
+      "loss": 0.8672,
+      "step": 16409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1695791489352346e-06,
+      "loss": 0.7617,
+      "step": 16410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1690090699317486e-06,
+      "loss": 0.8169,
+      "step": 16411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1684390450895783e-06,
+      "loss": 0.8315,
+      "step": 16412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1678690744196323e-06,
+      "loss": 0.8892,
+      "step": 16413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.167299157932813e-06,
+      "loss": 0.811,
+      "step": 16414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1667292956400253e-06,
+      "loss": 0.8164,
+      "step": 16415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1661594875521696e-06,
+      "loss": 0.9072,
+      "step": 16416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1655897336801505e-06,
+      "loss": 0.8862,
+      "step": 16417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.165020034034866e-06,
+      "loss": 0.3201,
+      "step": 16418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1644503886272137e-06,
+      "loss": 0.813,
+      "step": 16419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1638807974680944e-06,
+      "loss": 0.8564,
+      "step": 16420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.163311260568406e-06,
+      "loss": 0.8599,
+      "step": 16421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1627417779390414e-06,
+      "loss": 0.8013,
+      "step": 16422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1621723495908993e-06,
+      "loss": 0.8945,
+      "step": 16423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.161602975534872e-06,
+      "loss": 0.8472,
+      "step": 16424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1610336557818507e-06,
+      "loss": 0.8335,
+      "step": 16425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1604643903427312e-06,
+      "loss": 0.8613,
+      "step": 16426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1598951792284006e-06,
+      "loss": 0.4055,
+      "step": 16427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1593260224497507e-06,
+      "loss": 0.8369,
+      "step": 16428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.158756920017672e-06,
+      "loss": 0.8389,
+      "step": 16429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.158187871943051e-06,
+      "loss": 0.8872,
+      "step": 16430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1576188782367736e-06,
+      "loss": 0.8525,
+      "step": 16431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1570499389097248e-06,
+      "loss": 0.4871,
+      "step": 16432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1564810539727905e-06,
+      "loss": 0.8433,
+      "step": 16433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1559122234368555e-06,
+      "loss": 0.8589,
+      "step": 16434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1553434473128005e-06,
+      "loss": 0.8384,
+      "step": 16435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1547747256115097e-06,
+      "loss": 0.7539,
+      "step": 16436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1542060583438617e-06,
+      "loss": 0.8535,
+      "step": 16437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1536374455207345e-06,
+      "loss": 0.8486,
+      "step": 16438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1530688871530102e-06,
+      "loss": 0.8252,
+      "step": 16439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1525003832515624e-06,
+      "loss": 0.8027,
+      "step": 16440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1519319338272693e-06,
+      "loss": 0.8301,
+      "step": 16441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1513635388910083e-06,
+      "loss": 0.9272,
+      "step": 16442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1507951984536513e-06,
+      "loss": 0.8481,
+      "step": 16443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1502269125260722e-06,
+      "loss": 0.9028,
+      "step": 16444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.149658681119141e-06,
+      "loss": 0.8638,
+      "step": 16445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.149090504243731e-06,
+      "loss": 0.7881,
+      "step": 16446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1485223819107138e-06,
+      "loss": 0.9023,
+      "step": 16447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.147954314130955e-06,
+      "loss": 0.856,
+      "step": 16448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.147386300915326e-06,
+      "loss": 0.8804,
+      "step": 16449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1468183422746925e-06,
+      "loss": 0.8203,
+      "step": 16450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1462504382199182e-06,
+      "loss": 0.8618,
+      "step": 16451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1456825887618723e-06,
+      "loss": 0.8193,
+      "step": 16452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.145114793911414e-06,
+      "loss": 0.8706,
+      "step": 16453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.144547053679409e-06,
+      "loss": 0.8208,
+      "step": 16454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1439793680767194e-06,
+      "loss": 0.8359,
+      "step": 16455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1434117371142057e-06,
+      "loss": 0.8579,
+      "step": 16456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1428441608027244e-06,
+      "loss": 0.8276,
+      "step": 16457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.142276639153139e-06,
+      "loss": 0.8662,
+      "step": 16458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1417091721763035e-06,
+      "loss": 0.8887,
+      "step": 16459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1411417598830764e-06,
+      "loss": 0.8633,
+      "step": 16460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1405744022843113e-06,
+      "loss": 0.8242,
+      "step": 16461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.140007099390865e-06,
+      "loss": 0.8159,
+      "step": 16462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.13943985121359e-06,
+      "loss": 0.8076,
+      "step": 16463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.138872657763336e-06,
+      "loss": 0.8325,
+      "step": 16464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1383055190509576e-06,
+      "loss": 0.8442,
+      "step": 16465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1377384350873055e-06,
+      "loss": 0.7817,
+      "step": 16466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1371714058832254e-06,
+      "loss": 0.8301,
+      "step": 16467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.136604431449569e-06,
+      "loss": 0.7773,
+      "step": 16468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.136037511797183e-06,
+      "loss": 0.8555,
+      "step": 16469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.13547064693691e-06,
+      "loss": 0.8003,
+      "step": 16470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1349038368795997e-06,
+      "loss": 0.7837,
+      "step": 16471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.134337081636092e-06,
+      "loss": 0.9028,
+      "step": 16472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.133770381217233e-06,
+      "loss": 0.8867,
+      "step": 16473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1332037356338613e-06,
+      "loss": 0.8047,
+      "step": 16474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1326371448968205e-06,
+      "loss": 0.834,
+      "step": 16475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1320706090169524e-06,
+      "loss": 0.7544,
+      "step": 16476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.13150412800509e-06,
+      "loss": 0.8774,
+      "step": 16477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.130937701872073e-06,
+      "loss": 0.877,
+      "step": 16478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.130371330628741e-06,
+      "loss": 0.8623,
+      "step": 16479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1298050142859254e-06,
+      "loss": 0.7949,
+      "step": 16480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1292387528544647e-06,
+      "loss": 0.7393,
+      "step": 16481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1286725463451886e-06,
+      "loss": 0.833,
+      "step": 16482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.128106394768933e-06,
+      "loss": 0.8306,
+      "step": 16483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.127540298136528e-06,
+      "loss": 0.8643,
+      "step": 16484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1269742564588012e-06,
+      "loss": 0.9097,
+      "step": 16485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1264082697465845e-06,
+      "loss": 0.8618,
+      "step": 16486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1258423380107074e-06,
+      "loss": 0.8408,
+      "step": 16487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1252764612619943e-06,
+      "loss": 0.8481,
+      "step": 16488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1247106395112737e-06,
+      "loss": 0.8535,
+      "step": 16489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1241448727693696e-06,
+      "loss": 0.8618,
+      "step": 16490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.123579161047104e-06,
+      "loss": 0.8311,
+      "step": 16491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1230135043553033e-06,
+      "loss": 0.8354,
+      "step": 16492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1224479027047866e-06,
+      "loss": 0.8447,
+      "step": 16493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1218823561063762e-06,
+      "loss": 0.8442,
+      "step": 16494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1213168645708927e-06,
+      "loss": 0.8579,
+      "step": 16495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.120751428109154e-06,
+      "loss": 0.8633,
+      "step": 16496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.120186046731978e-06,
+      "loss": 0.875,
+      "step": 16497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.119620720450179e-06,
+      "loss": 0.4478,
+      "step": 16498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.119055449274574e-06,
+      "loss": 0.8638,
+      "step": 16499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.118490233215981e-06,
+      "loss": 0.8589,
+      "step": 16500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1179250722852074e-06,
+      "loss": 0.8569,
+      "step": 16501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.117359966493071e-06,
+      "loss": 0.8354,
+      "step": 16502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1167949158503814e-06,
+      "loss": 0.7915,
+      "step": 16503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1162299203679463e-06,
+      "loss": 0.7852,
+      "step": 16504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1156649800565788e-06,
+      "loss": 0.8486,
+      "step": 16505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.115100094927084e-06,
+      "loss": 0.8579,
+      "step": 16506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.11453526499027e-06,
+      "loss": 0.8667,
+      "step": 16507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1139704902569452e-06,
+      "loss": 0.8511,
+      "step": 16508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1134057707379126e-06,
+      "loss": 0.7729,
+      "step": 16509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1128411064439763e-06,
+      "loss": 0.7969,
+      "step": 16510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1122764973859377e-06,
+      "loss": 0.8228,
+      "step": 16511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1117119435746003e-06,
+      "loss": 0.8638,
+      "step": 16512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1111474450207663e-06,
+      "loss": 0.8877,
+      "step": 16513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.110583001735232e-06,
+      "loss": 0.8667,
+      "step": 16514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1100186137288005e-06,
+      "loss": 0.8516,
+      "step": 16515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.109454281012266e-06,
+      "loss": 0.7974,
+      "step": 16516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.108890003596425e-06,
+      "loss": 0.835,
+      "step": 16517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1083257814920753e-06,
+      "loss": 0.8545,
+      "step": 16518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1077616147100087e-06,
+      "loss": 0.8145,
+      "step": 16519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.10719750326102e-06,
+      "loss": 0.7983,
+      "step": 16520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.106633447155903e-06,
+      "loss": 0.8833,
+      "step": 16521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1060694464054484e-06,
+      "loss": 0.8403,
+      "step": 16522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1055055010204427e-06,
+      "loss": 0.8452,
+      "step": 16523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1049416110116804e-06,
+      "loss": 0.833,
+      "step": 16524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1043777763899447e-06,
+      "loss": 0.874,
+      "step": 16525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1038139971660277e-06,
+      "loss": 0.7822,
+      "step": 16526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1032502733507104e-06,
+      "loss": 0.8057,
+      "step": 16527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1026866049547822e-06,
+      "loss": 0.8022,
+      "step": 16528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1021229919890247e-06,
+      "loss": 0.8052,
+      "step": 16529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.101559434464219e-06,
+      "loss": 0.8379,
+      "step": 16530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1009959323911483e-06,
+      "loss": 0.8164,
+      "step": 16531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.100432485780596e-06,
+      "loss": 0.7754,
+      "step": 16532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0998690946433375e-06,
+      "loss": 0.877,
+      "step": 16533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.099305758990155e-06,
+      "loss": 0.8594,
+      "step": 16534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0987424788318245e-06,
+      "loss": 0.8379,
+      "step": 16535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0981792541791204e-06,
+      "loss": 0.8721,
+      "step": 16536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.097616085042821e-06,
+      "loss": 0.9067,
+      "step": 16537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0970529714336986e-06,
+      "loss": 0.8452,
+      "step": 16538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0964899133625283e-06,
+      "loss": 0.8154,
+      "step": 16539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0959269108400798e-06,
+      "loss": 0.8325,
+      "step": 16540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0953639638771277e-06,
+      "loss": 0.8496,
+      "step": 16541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0948010724844405e-06,
+      "loss": 0.8364,
+      "step": 16542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0942382366727843e-06,
+      "loss": 0.8618,
+      "step": 16543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.09367545645293e-06,
+      "loss": 0.8887,
+      "step": 16544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0931127318356455e-06,
+      "loss": 0.8574,
+      "step": 16545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0925500628316936e-06,
+      "loss": 0.8477,
+      "step": 16546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0919874494518423e-06,
+      "loss": 0.8857,
+      "step": 16547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.091424891706853e-06,
+      "loss": 0.8696,
+      "step": 16548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0908623896074874e-06,
+      "loss": 0.8994,
+      "step": 16549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0902999431645103e-06,
+      "loss": 0.7876,
+      "step": 16550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.089737552388679e-06,
+      "loss": 0.8618,
+      "step": 16551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0891752172907543e-06,
+      "loss": 0.9131,
+      "step": 16552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.088612937881496e-06,
+      "loss": 0.856,
+      "step": 16553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.08805071417166e-06,
+      "loss": 0.8843,
+      "step": 16554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.087488546172003e-06,
+      "loss": 0.8569,
+      "step": 16555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.086926433893277e-06,
+      "loss": 0.8438,
+      "step": 16556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0863643773462394e-06,
+      "loss": 0.8198,
+      "step": 16557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0858023765416443e-06,
+      "loss": 0.8647,
+      "step": 16558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0852404314902404e-06,
+      "loss": 0.8848,
+      "step": 16559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0846785422027814e-06,
+      "loss": 0.8257,
+      "step": 16560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.084116708690016e-06,
+      "loss": 0.8467,
+      "step": 16561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.083554930962691e-06,
+      "loss": 0.812,
+      "step": 16562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.082993209031558e-06,
+      "loss": 0.8638,
+      "step": 16563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0824315429073595e-06,
+      "loss": 0.894,
+      "step": 16564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.081869932600843e-06,
+      "loss": 0.875,
+      "step": 16565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.081308378122755e-06,
+      "loss": 0.8931,
+      "step": 16566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.080746879483837e-06,
+      "loss": 0.8496,
+      "step": 16567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0801854366948297e-06,
+      "loss": 0.834,
+      "step": 16568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.079624049766478e-06,
+      "loss": 0.8613,
+      "step": 16569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.079062718709518e-06,
+      "loss": 0.8521,
+      "step": 16570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.078501443534693e-06,
+      "loss": 0.8452,
+      "step": 16571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0779402242527367e-06,
+      "loss": 0.8398,
+      "step": 16572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0773790608743918e-06,
+      "loss": 0.8037,
+      "step": 16573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0768179534103896e-06,
+      "loss": 0.8291,
+      "step": 16574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0762569018714645e-06,
+      "loss": 0.8706,
+      "step": 16575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0756959062683545e-06,
+      "loss": 0.8687,
+      "step": 16576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.075134966611788e-06,
+      "loss": 0.8477,
+      "step": 16577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0745740829124987e-06,
+      "loss": 0.8911,
+      "step": 16578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0740132551812188e-06,
+      "loss": 0.9062,
+      "step": 16579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0734524834286764e-06,
+      "loss": 0.7793,
+      "step": 16580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0728917676655968e-06,
+      "loss": 0.8232,
+      "step": 16581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.072331107902713e-06,
+      "loss": 0.7817,
+      "step": 16582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0717705041507464e-06,
+      "loss": 0.8447,
+      "step": 16583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0712099564204264e-06,
+      "loss": 0.8447,
+      "step": 16584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0706494647224734e-06,
+      "loss": 0.792,
+      "step": 16585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0700890290676133e-06,
+      "loss": 0.8394,
+      "step": 16586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0695286494665677e-06,
+      "loss": 0.8638,
+      "step": 16587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.068968325930055e-06,
+      "loss": 0.825,
+      "step": 16588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.068408058468797e-06,
+      "loss": 0.8018,
+      "step": 16589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.067847847093514e-06,
+      "loss": 0.8569,
+      "step": 16590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.067287691814921e-06,
+      "loss": 0.8618,
+      "step": 16591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0667275926437368e-06,
+      "loss": 0.8101,
+      "step": 16592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0661675495906748e-06,
+      "loss": 0.3475,
+      "step": 16593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.065607562666452e-06,
+      "loss": 0.8208,
+      "step": 16594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0650476318817814e-06,
+      "loss": 0.7974,
+      "step": 16595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.064487757247372e-06,
+      "loss": 0.3645,
+      "step": 16596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0639279387739376e-06,
+      "loss": 0.8521,
+      "step": 16597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.06336817647219e-06,
+      "loss": 0.896,
+      "step": 16598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.062808470352836e-06,
+      "loss": 0.749,
+      "step": 16599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.062248820426585e-06,
+      "loss": 0.8691,
+      "step": 16600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0616892267041446e-06,
+      "loss": 0.8687,
+      "step": 16601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.061129689196217e-06,
+      "loss": 0.8618,
+      "step": 16602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.060570207913511e-06,
+      "loss": 0.8784,
+      "step": 16603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0600107828667274e-06,
+      "loss": 0.7603,
+      "step": 16604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0594514140665705e-06,
+      "loss": 0.8296,
+      "step": 16605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0588921015237433e-06,
+      "loss": 0.3801,
+      "step": 16606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.058332845248945e-06,
+      "loss": 0.8433,
+      "step": 16607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0577736452528745e-06,
+      "loss": 0.7832,
+      "step": 16608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.057214501546229e-06,
+      "loss": 0.834,
+      "step": 16609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.056655414139708e-06,
+      "loss": 0.8516,
+      "step": 16610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0560963830440082e-06,
+      "loss": 0.7832,
+      "step": 16611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.055537408269822e-06,
+      "loss": 0.8159,
+      "step": 16612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0549784898278464e-06,
+      "loss": 0.8066,
+      "step": 16613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0544196277287735e-06,
+      "loss": 0.8306,
+      "step": 16614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053860821983293e-06,
+      "loss": 0.832,
+      "step": 16615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053302072602099e-06,
+      "loss": 0.8623,
+      "step": 16616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0527433795958785e-06,
+      "loss": 0.8691,
+      "step": 16617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.052184742975321e-06,
+      "loss": 0.8232,
+      "step": 16618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0516261627511164e-06,
+      "loss": 0.8149,
+      "step": 16619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0510676389339494e-06,
+      "loss": 0.7651,
+      "step": 16620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0505091715345053e-06,
+      "loss": 0.7766,
+      "step": 16621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.049950760563467e-06,
+      "loss": 0.9038,
+      "step": 16622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0493924060315194e-06,
+      "loss": 0.8364,
+      "step": 16623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0488341079493464e-06,
+      "loss": 0.8325,
+      "step": 16624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0482758663276254e-06,
+      "loss": 0.8838,
+      "step": 16625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0477176811770404e-06,
+      "loss": 0.9097,
+      "step": 16626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0471595525082684e-06,
+      "loss": 0.8633,
+      "step": 16627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0466014803319856e-06,
+      "loss": 0.8218,
+      "step": 16628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.046043464658872e-06,
+      "loss": 0.8755,
+      "step": 16629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0454855054996e-06,
+      "loss": 0.8643,
+      "step": 16630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0449276028648457e-06,
+      "loss": 0.8604,
+      "step": 16631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0443697567652852e-06,
+      "loss": 0.8433,
+      "step": 16632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0438119672115885e-06,
+      "loss": 0.8589,
+      "step": 16633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0432542342144257e-06,
+      "loss": 0.8193,
+      "step": 16634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0426965577844703e-06,
+      "loss": 0.3768,
+      "step": 16635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.042138937932388e-06,
+      "loss": 0.835,
+      "step": 16636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0415813746688507e-06,
+      "loss": 0.8257,
+      "step": 16637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.041023868004522e-06,
+      "loss": 0.8311,
+      "step": 16638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.040466417950071e-06,
+      "loss": 0.8145,
+      "step": 16639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0399090245161614e-06,
+      "loss": 0.8672,
+      "step": 16640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0393516877134547e-06,
+      "loss": 0.8438,
+      "step": 16641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0387944075526166e-06,
+      "loss": 0.8481,
+      "step": 16642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.038237184044307e-06,
+      "loss": 0.8354,
+      "step": 16643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0376800171991872e-06,
+      "loss": 0.8452,
+      "step": 16644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.037122907027918e-06,
+      "loss": 0.793,
+      "step": 16645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.036565853541157e-06,
+      "loss": 0.8545,
+      "step": 16646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.036008856749559e-06,
+      "loss": 0.8237,
+      "step": 16647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0354519166637836e-06,
+      "loss": 0.8423,
+      "step": 16648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.034895033294483e-06,
+      "loss": 0.8174,
+      "step": 16649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0343382066523154e-06,
+      "loss": 0.8906,
+      "step": 16650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0337814367479287e-06,
+      "loss": 0.8667,
+      "step": 16651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0332247235919793e-06,
+      "loss": 0.8364,
+      "step": 16652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.032668067195116e-06,
+      "loss": 0.8047,
+      "step": 16653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0321114675679866e-06,
+      "loss": 0.8027,
+      "step": 16654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.031554924721241e-06,
+      "loss": 0.772,
+      "step": 16655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.03099843866553e-06,
+      "loss": 0.7954,
+      "step": 16656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0304420094114957e-06,
+      "loss": 0.8594,
+      "step": 16657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.029885636969787e-06,
+      "loss": 0.8198,
+      "step": 16658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.029329321351046e-06,
+      "loss": 0.8735,
+      "step": 16659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.028773062565915e-06,
+      "loss": 0.7949,
+      "step": 16660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0282168606250384e-06,
+      "loss": 0.8228,
+      "step": 16661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0276607155390547e-06,
+      "loss": 0.8452,
+      "step": 16662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0271046273186055e-06,
+      "loss": 0.855,
+      "step": 16663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0265485959743316e-06,
+      "loss": 0.8501,
+      "step": 16664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0259926215168684e-06,
+      "loss": 0.8501,
+      "step": 16665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0254367039568524e-06,
+      "loss": 0.8398,
+      "step": 16666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0248808433049183e-06,
+      "loss": 0.8359,
+      "step": 16667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0243250395717016e-06,
+      "loss": 0.8457,
+      "step": 16668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0237692927678375e-06,
+      "loss": 0.8745,
+      "step": 16669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0232136029039556e-06,
+      "loss": 0.8335,
+      "step": 16670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0226579699906893e-06,
+      "loss": 0.7959,
+      "step": 16671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0221023940386684e-06,
+      "loss": 0.8784,
+      "step": 16672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0215468750585197e-06,
+      "loss": 0.8486,
+      "step": 16673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0209914130608744e-06,
+      "loss": 0.8711,
+      "step": 16674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0204360080563556e-06,
+      "loss": 0.8525,
+      "step": 16675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.019880660055591e-06,
+      "loss": 0.8506,
+      "step": 16676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0193253690692076e-06,
+      "loss": 0.7822,
+      "step": 16677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0187701351078267e-06,
+      "loss": 0.9092,
+      "step": 16678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.018214958182071e-06,
+      "loss": 0.8823,
+      "step": 16679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.01765983830256e-06,
+      "loss": 0.7988,
+      "step": 16680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.017104775479916e-06,
+      "loss": 0.8765,
+      "step": 16681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.01654976972476e-06,
+      "loss": 0.8521,
+      "step": 16682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0159948210477057e-06,
+      "loss": 0.8486,
+      "step": 16683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0154399294593757e-06,
+      "loss": 0.8662,
+      "step": 16684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.014885094970382e-06,
+      "loss": 0.853,
+      "step": 16685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0143303175913392e-06,
+      "loss": 0.7998,
+      "step": 16686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0137755973328636e-06,
+      "loss": 0.8862,
+      "step": 16687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0132209342055647e-06,
+      "loss": 0.8184,
+      "step": 16688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.012666328220056e-06,
+      "loss": 0.8472,
+      "step": 16689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.01211177938695e-06,
+      "loss": 0.8901,
+      "step": 16690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.011557287716854e-06,
+      "loss": 0.8623,
+      "step": 16691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0110028532203742e-06,
+      "loss": 0.8003,
+      "step": 16692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.010448475908122e-06,
+      "loss": 0.8491,
+      "step": 16693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0098941557906986e-06,
+      "loss": 0.8555,
+      "step": 16694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0093398928787144e-06,
+      "loss": 0.7969,
+      "step": 16695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0087856871827694e-06,
+      "loss": 0.8911,
+      "step": 16696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.008231538713469e-06,
+      "loss": 0.9131,
+      "step": 16697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0076774474814136e-06,
+      "loss": 0.8579,
+      "step": 16698
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0071234134972023e-06,
+      "loss": 0.7861,
+      "step": 16699
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.006569436771436e-06,
+      "loss": 0.8359,
+      "step": 16700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0060155173147157e-06,
+      "loss": 0.8262,
+      "step": 16701
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.005461655137635e-06,
+      "loss": 0.8599,
+      "step": 16702
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0049078502507924e-06,
+      "loss": 0.8965,
+      "step": 16703
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.004354102664781e-06,
+      "loss": 0.8218,
+      "step": 16704
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0038004123901984e-06,
+      "loss": 0.4116,
+      "step": 16705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.003246779437635e-06,
+      "loss": 0.8213,
+      "step": 16706
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.002693203817681e-06,
+      "loss": 0.8555,
+      "step": 16707
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0021396855409315e-06,
+      "loss": 0.8628,
+      "step": 16708
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.001586224617971e-06,
+      "loss": 0.8589,
+      "step": 16709
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.001032821059392e-06,
+      "loss": 0.8013,
+      "step": 16710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.000479474875783e-06,
+      "loss": 0.8418,
+      "step": 16711
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9999261860777267e-06,
+      "loss": 0.832,
+      "step": 16712
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9993729546758097e-06,
+      "loss": 0.8667,
+      "step": 16713
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9988197806806177e-06,
+      "loss": 0.8467,
+      "step": 16714
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.998266664102731e-06,
+      "loss": 0.8213,
+      "step": 16715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9977136049527348e-06,
+      "loss": 0.877,
+      "step": 16716
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9971606032412062e-06,
+      "loss": 0.8271,
+      "step": 16717
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9966076589787293e-06,
+      "loss": 0.8638,
+      "step": 16718
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9960547721758806e-06,
+      "loss": 0.9111,
+      "step": 16719
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.995501942843236e-06,
+      "loss": 0.8232,
+      "step": 16720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9949491709913737e-06,
+      "loss": 0.853,
+      "step": 16721
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.994396456630871e-06,
+      "loss": 0.8589,
+      "step": 16722
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.993843799772298e-06,
+      "loss": 0.8789,
+      "step": 16723
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.993291200426233e-06,
+      "loss": 0.8105,
+      "step": 16724
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9927386586032444e-06,
+      "loss": 0.8628,
+      "step": 16725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.992186174313903e-06,
+      "loss": 0.8643,
+      "step": 16726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.991633747568781e-06,
+      "loss": 0.8145,
+      "step": 16727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.991081378378445e-06,
+      "loss": 0.8301,
+      "step": 16728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9905290667534625e-06,
+      "loss": 0.8989,
+      "step": 16729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9899768127044046e-06,
+      "loss": 0.8516,
+      "step": 16730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9894246162418325e-06,
+      "loss": 0.8433,
+      "step": 16731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9888724773763115e-06,
+      "loss": 0.7227,
+      "step": 16732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9883203961184033e-06,
+      "loss": 0.8188,
+      "step": 16733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9877683724786718e-06,
+      "loss": 0.8716,
+      "step": 16734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.98721640646768e-06,
+      "loss": 0.8906,
+      "step": 16735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9866644980959842e-06,
+      "loss": 0.8428,
+      "step": 16736
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9861126473741465e-06,
+      "loss": 0.8105,
+      "step": 16737
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.985560854312723e-06,
+      "loss": 0.8887,
+      "step": 16738
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9850091189222693e-06,
+      "loss": 0.8726,
+      "step": 16739
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9844574412133434e-06,
+      "loss": 0.7988,
+      "step": 16740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.983905821196497e-06,
+      "loss": 0.8149,
+      "step": 16741
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9833542588822846e-06,
+      "loss": 0.7983,
+      "step": 16742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.982802754281261e-06,
+      "loss": 0.9126,
+      "step": 16743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9822513074039756e-06,
+      "loss": 0.8823,
+      "step": 16744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9816999182609776e-06,
+      "loss": 0.8198,
+      "step": 16745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.981148586862815e-06,
+      "loss": 0.8545,
+      "step": 16746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9805973132200367e-06,
+      "loss": 0.8818,
+      "step": 16747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.980046097343192e-06,
+      "loss": 0.8623,
+      "step": 16748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.979494939242822e-06,
+      "loss": 0.823,
+      "step": 16749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9789438389294753e-06,
+      "loss": 0.9414,
+      "step": 16750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.978392796413694e-06,
+      "loss": 0.8647,
+      "step": 16751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9778418117060177e-06,
+      "loss": 0.8545,
+      "step": 16752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9772908848169917e-06,
+      "loss": 0.8188,
+      "step": 16753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976740015757153e-06,
+      "loss": 0.7964,
+      "step": 16754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976189204537042e-06,
+      "loss": 0.8613,
+      "step": 16755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.975638451167197e-06,
+      "loss": 0.8374,
+      "step": 16756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9750877556581553e-06,
+      "loss": 0.8608,
+      "step": 16757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.97453711802045e-06,
+      "loss": 0.7666,
+      "step": 16758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.973986538264618e-06,
+      "loss": 0.7959,
+      "step": 16759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9734360164011906e-06,
+      "loss": 0.8574,
+      "step": 16760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.972885552440703e-06,
+      "loss": 0.8906,
+      "step": 16761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9723351463936835e-06,
+      "loss": 0.9263,
+      "step": 16762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.971784798270665e-06,
+      "loss": 0.8677,
+      "step": 16763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9712345080821754e-06,
+      "loss": 0.7886,
+      "step": 16764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.970684275838741e-06,
+      "loss": 0.8398,
+      "step": 16765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9701341015508894e-06,
+      "loss": 0.9272,
+      "step": 16766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.969583985229149e-06,
+      "loss": 0.3368,
+      "step": 16767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.969033926884041e-06,
+      "loss": 0.9028,
+      "step": 16768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9684839265260915e-06,
+      "loss": 0.855,
+      "step": 16769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9679339841658214e-06,
+      "loss": 0.8057,
+      "step": 16770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9673840998137505e-06,
+      "loss": 0.8677,
+      "step": 16771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9668342734804024e-06,
+      "loss": 0.8237,
+      "step": 16772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9662845051762923e-06,
+      "loss": 0.8242,
+      "step": 16773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9657347949119422e-06,
+      "loss": 0.8218,
+      "step": 16774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9651851426978653e-06,
+      "loss": 0.8394,
+      "step": 16775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.96463554854458e-06,
+      "loss": 0.8076,
+      "step": 16776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9640860124626003e-06,
+      "loss": 0.7949,
+      "step": 16777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9635365344624373e-06,
+      "loss": 0.8965,
+      "step": 16778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.962987114554605e-06,
+      "loss": 0.8833,
+      "step": 16779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9624377527496176e-06,
+      "loss": 0.8088,
+      "step": 16780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.96188844905798e-06,
+      "loss": 0.8135,
+      "step": 16781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9613392034902057e-06,
+      "loss": 0.8281,
+      "step": 16782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.960790016056801e-06,
+      "loss": 0.8027,
+      "step": 16783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9602408867682705e-06,
+      "loss": 0.8608,
+      "step": 16784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.959691815635124e-06,
+      "loss": 0.8267,
+      "step": 16785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.959142802667862e-06,
+      "loss": 0.8169,
+      "step": 16786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.95859384787699e-06,
+      "loss": 0.8618,
+      "step": 16787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.958044951273012e-06,
+      "loss": 0.8291,
+      "step": 16788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9574961128664276e-06,
+      "loss": 0.8335,
+      "step": 16789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9569473326677373e-06,
+      "loss": 0.9185,
+      "step": 16790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.956398610687438e-06,
+      "loss": 0.9072,
+      "step": 16791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9558499469360287e-06,
+      "loss": 0.8613,
+      "step": 16792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9553013414240093e-06,
+      "loss": 0.8706,
+      "step": 16793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.954752794161871e-06,
+      "loss": 0.8374,
+      "step": 16794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9542043051601122e-06,
+      "loss": 0.8252,
+      "step": 16795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.953655874429225e-06,
+      "loss": 0.7876,
+      "step": 16796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9531075019796995e-06,
+      "loss": 0.8677,
+      "step": 16797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9525591878220307e-06,
+      "loss": 0.8032,
+      "step": 16798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.952010931966705e-06,
+      "loss": 0.8506,
+      "step": 16799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9514627344242127e-06,
+      "loss": 0.8618,
+      "step": 16800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950914595205044e-06,
+      "loss": 0.8364,
+      "step": 16801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950366514319684e-06,
+      "loss": 0.8276,
+      "step": 16802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9498184917786166e-06,
+      "loss": 0.8149,
+      "step": 16803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.94927052759233e-06,
+      "loss": 0.8076,
+      "step": 16804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9487226217713034e-06,
+      "loss": 0.8521,
+      "step": 16805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.948174774326023e-06,
+      "loss": 0.7856,
+      "step": 16806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9476269852669667e-06,
+      "loss": 0.8301,
+      "step": 16807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9470792546046175e-06,
+      "loss": 0.8794,
+      "step": 16808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9465315823494535e-06,
+      "loss": 0.7739,
+      "step": 16809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9459839685119496e-06,
+      "loss": 0.8843,
+      "step": 16810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.945436413102587e-06,
+      "loss": 0.8389,
+      "step": 16811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.944888916131838e-06,
+      "loss": 0.7974,
+      "step": 16812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9443414776101777e-06,
+      "loss": 0.856,
+      "step": 16813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.943794097548082e-06,
+      "loss": 0.873,
+      "step": 16814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9432467759560192e-06,
+      "loss": 0.8296,
+      "step": 16815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9426995128444643e-06,
+      "loss": 0.7637,
+      "step": 16816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9421523082238852e-06,
+      "loss": 0.8428,
+      "step": 16817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9416051621047492e-06,
+      "loss": 0.8169,
+      "step": 16818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.941058074497528e-06,
+      "loss": 0.8169,
+      "step": 16819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9405110454126847e-06,
+      "loss": 0.8613,
+      "step": 16820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.939964074860686e-06,
+      "loss": 0.8091,
+      "step": 16821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9394171628519976e-06,
+      "loss": 0.8477,
+      "step": 16822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9388703093970822e-06,
+      "loss": 0.8223,
+      "step": 16823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9383235145064e-06,
+      "loss": 0.8271,
+      "step": 16824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9377767781904157e-06,
+      "loss": 0.853,
+      "step": 16825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9372301004595846e-06,
+      "loss": 0.8213,
+      "step": 16826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.93668348132437e-06,
+      "loss": 0.854,
+      "step": 16827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936136920795226e-06,
+      "loss": 0.8428,
+      "step": 16828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.935590418882612e-06,
+      "loss": 0.8257,
+      "step": 16829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9350439755969825e-06,
+      "loss": 0.9199,
+      "step": 16830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.93449759094879e-06,
+      "loss": 0.8652,
+      "step": 16831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9339512649484887e-06,
+      "loss": 0.8501,
+      "step": 16832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9334049976065334e-06,
+      "loss": 0.8457,
+      "step": 16833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.932858788933371e-06,
+      "loss": 0.8687,
+      "step": 16834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.932312638939455e-06,
+      "loss": 0.8335,
+      "step": 16835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.931766547635232e-06,
+      "loss": 0.7834,
+      "step": 16836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9312205150311486e-06,
+      "loss": 0.8345,
+      "step": 16837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9306745411376537e-06,
+      "loss": 0.8169,
+      "step": 16838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.93012862596519e-06,
+      "loss": 0.8721,
+      "step": 16839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9295827695242035e-06,
+      "loss": 0.854,
+      "step": 16840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9290369718251385e-06,
+      "loss": 0.8564,
+      "step": 16841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9284912328784356e-06,
+      "loss": 0.8608,
+      "step": 16842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.927945552694535e-06,
+      "loss": 0.8745,
+      "step": 16843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.927399931283876e-06,
+      "loss": 0.8682,
+      "step": 16844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9268543686568975e-06,
+      "loss": 0.7886,
+      "step": 16845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.926308864824039e-06,
+      "loss": 0.8975,
+      "step": 16846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9257634197957347e-06,
+      "loss": 0.8467,
+      "step": 16847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.925218033582422e-06,
+      "loss": 0.8877,
+      "step": 16848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9246727061945333e-06,
+      "loss": 0.8447,
+      "step": 16849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9241274376425e-06,
+      "loss": 0.8789,
+      "step": 16850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9235822279367576e-06,
+      "loss": 0.8711,
+      "step": 16851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9230370770877337e-06,
+      "loss": 0.8164,
+      "step": 16852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9224919851058586e-06,
+      "loss": 0.9287,
+      "step": 16853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.921946952001564e-06,
+      "loss": 0.8369,
+      "step": 16854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.921401977785274e-06,
+      "loss": 0.8413,
+      "step": 16855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9208570624674156e-06,
+      "loss": 0.874,
+      "step": 16856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.920312206058412e-06,
+      "loss": 0.894,
+      "step": 16857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9197674085686884e-06,
+      "loss": 0.8813,
+      "step": 16858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9192226700086703e-06,
+      "loss": 0.8643,
+      "step": 16859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9186779903887754e-06,
+      "loss": 0.8008,
+      "step": 16860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9181333697194276e-06,
+      "loss": 0.8271,
+      "step": 16861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.917588808011045e-06,
+      "loss": 0.7988,
+      "step": 16862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9170443052740445e-06,
+      "loss": 0.8286,
+      "step": 16863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.916499861518846e-06,
+      "loss": 0.8262,
+      "step": 16864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.915955476755862e-06,
+      "loss": 0.834,
+      "step": 16865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9154111509955105e-06,
+      "loss": 0.874,
+      "step": 16866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.914866884248205e-06,
+      "loss": 0.854,
+      "step": 16867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9143226765243584e-06,
+      "loss": 0.8823,
+      "step": 16868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9137785278343795e-06,
+      "loss": 0.8027,
+      "step": 16869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.913234438188682e-06,
+      "loss": 0.8779,
+      "step": 16870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9126904075976728e-06,
+      "loss": 0.8384,
+      "step": 16871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9121464360717625e-06,
+      "loss": 0.4493,
+      "step": 16872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.911602523621355e-06,
+      "loss": 0.8662,
+      "step": 16873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.911058670256859e-06,
+      "loss": 0.8394,
+      "step": 16874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.910514875988679e-06,
+      "loss": 0.8521,
+      "step": 16875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.909971140827216e-06,
+      "loss": 0.8096,
+      "step": 16876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9094274647828757e-06,
+      "loss": 0.8472,
+      "step": 16877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.908883847866057e-06,
+      "loss": 0.8359,
+      "step": 16878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.908340290087161e-06,
+      "loss": 0.8081,
+      "step": 16879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9077967914565885e-06,
+      "loss": 0.7659,
+      "step": 16880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9072533519847364e-06,
+      "loss": 0.8472,
+      "step": 16881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9067099716819998e-06,
+      "loss": 0.7974,
+      "step": 16882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9061666505587773e-06,
+      "loss": 0.8457,
+      "step": 16883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9056233886254599e-06,
+      "loss": 0.874,
+      "step": 16884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9050801858924456e-06,
+      "loss": 0.8643,
+      "step": 16885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9045370423701221e-06,
+      "loss": 0.8232,
+      "step": 16886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.903993958068885e-06,
+      "loss": 0.812,
+      "step": 16887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.903450932999122e-06,
+      "loss": 0.8687,
+      "step": 16888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9029079671712203e-06,
+      "loss": 0.8306,
+      "step": 16889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9023650605955701e-06,
+      "loss": 0.7036,
+      "step": 16890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9018222132825587e-06,
+      "loss": 0.8535,
+      "step": 16891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9012794252425692e-06,
+      "loss": 0.3093,
+      "step": 16892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9007366964859885e-06,
+      "loss": 0.8125,
+      "step": 16893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9001940270231988e-06,
+      "loss": 0.8457,
+      "step": 16894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8996514168645809e-06,
+      "loss": 0.7939,
+      "step": 16895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8991088660205182e-06,
+      "loss": 0.8726,
+      "step": 16896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8985663745013877e-06,
+      "loss": 0.8037,
+      "step": 16897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8980239423175695e-06,
+      "loss": 0.8862,
+      "step": 16898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8974815694794429e-06,
+      "loss": 0.8511,
+      "step": 16899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8969392559973832e-06,
+      "loss": 0.8838,
+      "step": 16900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8963970018817657e-06,
+      "loss": 0.8477,
+      "step": 16901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8958548071429616e-06,
+      "loss": 0.8906,
+      "step": 16902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8953126717913473e-06,
+      "loss": 0.8088,
+      "step": 16903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8947705958372953e-06,
+      "loss": 0.874,
+      "step": 16904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8942285792911735e-06,
+      "loss": 0.8735,
+      "step": 16905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.893686622163355e-06,
+      "loss": 0.8643,
+      "step": 16906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.893144724464206e-06,
+      "loss": 0.8501,
+      "step": 16907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8926028862040924e-06,
+      "loss": 0.8037,
+      "step": 16908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8920611073933842e-06,
+      "loss": 0.8511,
+      "step": 16909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8915193880424425e-06,
+      "loss": 0.8408,
+      "step": 16910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8909777281616337e-06,
+      "loss": 0.854,
+      "step": 16911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8904361277613221e-06,
+      "loss": 0.8882,
+      "step": 16912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.889894586851867e-06,
+      "loss": 0.8076,
+      "step": 16913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8893531054436292e-06,
+      "loss": 0.8823,
+      "step": 16914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8888116835469666e-06,
+      "loss": 0.8159,
+      "step": 16915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8882703211722393e-06,
+      "loss": 0.8555,
+      "step": 16916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8877290183298058e-06,
+      "loss": 0.8345,
+      "step": 16917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8871877750300194e-06,
+      "loss": 0.7642,
+      "step": 16918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8866465912832372e-06,
+      "loss": 0.7964,
+      "step": 16919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8861054670998119e-06,
+      "loss": 0.8481,
+      "step": 16920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8855644024900937e-06,
+      "loss": 0.8862,
+      "step": 16921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8850233974644383e-06,
+      "loss": 0.7979,
+      "step": 16922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8844824520331923e-06,
+      "loss": 0.8174,
+      "step": 16923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8839415662067068e-06,
+      "loss": 0.8428,
+      "step": 16924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8834007399953303e-06,
+      "loss": 0.9043,
+      "step": 16925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8828599734094078e-06,
+      "loss": 0.8472,
+      "step": 16926
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.882319266459287e-06,
+      "loss": 0.3173,
+      "step": 16927
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8817786191553121e-06,
+      "loss": 0.8047,
+      "step": 16928
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8812380315078239e-06,
+      "loss": 0.8164,
+      "step": 16929
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8806975035271685e-06,
+      "loss": 0.835,
+      "step": 16930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8801570352236837e-06,
+      "loss": 0.8618,
+      "step": 16931
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.879616626607711e-06,
+      "loss": 0.8721,
+      "step": 16932
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.879076277689591e-06,
+      "loss": 0.8887,
+      "step": 16933
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.87853598847966e-06,
+      "loss": 0.8433,
+      "step": 16934
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8779957589882525e-06,
+      "loss": 0.8584,
+      "step": 16935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8774555892257074e-06,
+      "loss": 0.8457,
+      "step": 16936
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8769154792023559e-06,
+      "loss": 0.8135,
+      "step": 16937
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.876375428928534e-06,
+      "loss": 0.9229,
+      "step": 16938
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.875835438414571e-06,
+      "loss": 0.8174,
+      "step": 16939
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8752955076708007e-06,
+      "loss": 0.8682,
+      "step": 16940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8747556367075508e-06,
+      "loss": 0.8652,
+      "step": 16941
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8742158255351494e-06,
+      "loss": 0.781,
+      "step": 16942
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8736760741639265e-06,
+      "loss": 0.8101,
+      "step": 16943
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8731363826042049e-06,
+      "loss": 0.4075,
+      "step": 16944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8725967508663111e-06,
+      "loss": 0.8315,
+      "step": 16945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.872057178960572e-06,
+      "loss": 0.8657,
+      "step": 16946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8715176668973079e-06,
+      "loss": 0.8481,
+      "step": 16947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8709782146868388e-06,
+      "loss": 0.8184,
+      "step": 16948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8704388223394892e-06,
+      "loss": 0.8564,
+      "step": 16949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8698994898655747e-06,
+      "loss": 0.8354,
+      "step": 16950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8693602172754173e-06,
+      "loss": 0.8735,
+      "step": 16951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8688210045793303e-06,
+      "loss": 0.811,
+      "step": 16952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8682818517876334e-06,
+      "loss": 0.8604,
+      "step": 16953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8677427589106394e-06,
+      "loss": 0.8301,
+      "step": 16954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8672037259586606e-06,
+      "loss": 0.8164,
+      "step": 16955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8666647529420113e-06,
+      "loss": 0.8257,
+      "step": 16956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8661258398710047e-06,
+      "loss": 0.8677,
+      "step": 16957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8655869867559472e-06,
+      "loss": 0.8521,
+      "step": 16958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8650481936071519e-06,
+      "loss": 0.8555,
+      "step": 16959
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.864509460434924e-06,
+      "loss": 0.8262,
+      "step": 16960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8639707872495699e-06,
+      "loss": 0.8364,
+      "step": 16961
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.863432174061398e-06,
+      "loss": 0.8379,
+      "step": 16962
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8628936208807096e-06,
+      "loss": 0.8838,
+      "step": 16963
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8623551277178097e-06,
+      "loss": 0.8965,
+      "step": 16964
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8618166945830019e-06,
+      "loss": 0.917,
+      "step": 16965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8612783214865864e-06,
+      "loss": 0.8228,
+      "step": 16966
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.860740008438862e-06,
+      "loss": 0.8389,
+      "step": 16967
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8602017554501268e-06,
+      "loss": 0.9004,
+      "step": 16968
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8596635625306796e-06,
+      "loss": 0.8398,
+      "step": 16969
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.859125429690819e-06,
+      "loss": 0.7798,
+      "step": 16970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8585873569408363e-06,
+      "loss": 0.7871,
+      "step": 16971
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8580493442910296e-06,
+      "loss": 0.8359,
+      "step": 16972
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8575113917516902e-06,
+      "loss": 0.8203,
+      "step": 16973
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8569734993331079e-06,
+      "loss": 0.8145,
+      "step": 16974
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.856435667045577e-06,
+      "loss": 0.772,
+      "step": 16975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8558978948993833e-06,
+      "loss": 0.8647,
+      "step": 16976
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.855360182904818e-06,
+      "loss": 0.8687,
+      "step": 16977
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8548225310721696e-06,
+      "loss": 0.8472,
+      "step": 16978
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8542849394117218e-06,
+      "loss": 0.8271,
+      "step": 16979
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8537474079337603e-06,
+      "loss": 0.8516,
+      "step": 16980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8532099366485673e-06,
+      "loss": 0.8545,
+      "step": 16981
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8526725255664268e-06,
+      "loss": 0.8828,
+      "step": 16982
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8521351746976218e-06,
+      "loss": 0.8384,
+      "step": 16983
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8515978840524302e-06,
+      "loss": 0.8135,
+      "step": 16984
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8510606536411335e-06,
+      "loss": 0.8599,
+      "step": 16985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8505234834740094e-06,
+      "loss": 0.8999,
+      "step": 16986
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8499863735613316e-06,
+      "loss": 0.8491,
+      "step": 16987
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8494493239133803e-06,
+      "loss": 0.8184,
+      "step": 16988
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8489123345404263e-06,
+      "loss": 0.8467,
+      "step": 16989
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8483754054527448e-06,
+      "loss": 0.9009,
+      "step": 16990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.84783853666061e-06,
+      "loss": 0.8384,
+      "step": 16991
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8473017281742906e-06,
+      "loss": 0.8794,
+      "step": 16992
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.846764980004056e-06,
+      "loss": 0.8423,
+      "step": 16993
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.846228292160178e-06,
+      "loss": 0.876,
+      "step": 16994
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8456916646529205e-06,
+      "loss": 0.8237,
+      "step": 16995
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.845155097492554e-06,
+      "loss": 0.8545,
+      "step": 16996
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.84461859068934e-06,
+      "loss": 0.792,
+      "step": 16997
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.844082144253546e-06,
+      "loss": 0.8652,
+      "step": 16998
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8435457581954342e-06,
+      "loss": 0.8657,
+      "step": 16999
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8430094325252645e-06,
+      "loss": 0.853,
+      "step": 17000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8424731672532985e-06,
+      "loss": 0.8281,
+      "step": 17001
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8419369623897988e-06,
+      "loss": 0.8994,
+      "step": 17002
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8414008179450194e-06,
+      "loss": 0.8193,
+      "step": 17003
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8408647339292213e-06,
+      "loss": 0.8228,
+      "step": 17004
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.840328710352659e-06,
+      "loss": 0.8267,
+      "step": 17005
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.839792747225586e-06,
+      "loss": 0.8872,
+      "step": 17006
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8392568445582594e-06,
+      "loss": 0.8843,
+      "step": 17007
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8387210023609282e-06,
+      "loss": 0.7153,
+      "step": 17008
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8381852206438477e-06,
+      "loss": 0.8276,
+      "step": 17009
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8376494994172644e-06,
+      "loss": 0.8193,
+      "step": 17010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.837113838691431e-06,
+      "loss": 0.7939,
+      "step": 17011
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8365782384765945e-06,
+      "loss": 0.833,
+      "step": 17012
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8360426987829993e-06,
+      "loss": 0.8179,
+      "step": 17013
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8355072196208928e-06,
+      "loss": 0.832,
+      "step": 17014
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8349718010005212e-06,
+      "loss": 0.8662,
+      "step": 17015
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8344364429321255e-06,
+      "loss": 0.8535,
+      "step": 17016
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8339011454259504e-06,
+      "loss": 0.8711,
+      "step": 17017
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8333659084922357e-06,
+      "loss": 0.8569,
+      "step": 17018
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8328307321412192e-06,
+      "loss": 0.8579,
+      "step": 17019
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8322956163831435e-06,
+      "loss": 0.9165,
+      "step": 17020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8317605612282425e-06,
+      "loss": 0.8359,
+      "step": 17021
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8312255666867545e-06,
+      "loss": 0.8101,
+      "step": 17022
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8306906327689166e-06,
+      "loss": 0.3665,
+      "step": 17023
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.830155759484961e-06,
+      "loss": 0.7983,
+      "step": 17024
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.829620946845121e-06,
+      "loss": 0.8994,
+      "step": 17025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.829086194859626e-06,
+      "loss": 0.8262,
+      "step": 17026
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.828551503538709e-06,
+      "loss": 0.8286,
+      "step": 17027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8280168728926017e-06,
+      "loss": 0.8193,
+      "step": 17028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8274823029315275e-06,
+      "loss": 0.8667,
+      "step": 17029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8269477936657182e-06,
+      "loss": 0.8477,
+      "step": 17030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8264133451053978e-06,
+      "loss": 0.8511,
+      "step": 17031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8258789572607893e-06,
+      "loss": 0.8237,
+      "step": 17032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8253446301421197e-06,
+      "loss": 0.8013,
+      "step": 17033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8248103637596082e-06,
+      "loss": 0.8364,
+      "step": 17034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.824276158123478e-06,
+      "loss": 0.8491,
+      "step": 17035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.823742013243951e-06,
+      "loss": 0.9038,
+      "step": 17036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8232079291312427e-06,
+      "loss": 0.7944,
+      "step": 17037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8226739057955738e-06,
+      "loss": 0.8252,
+      "step": 17038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8221399432471609e-06,
+      "loss": 0.8384,
+      "step": 17039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8216060414962167e-06,
+      "loss": 0.9097,
+      "step": 17040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8210722005529585e-06,
+      "loss": 0.8262,
+      "step": 17041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8205384204275973e-06,
+      "loss": 0.8232,
+      "step": 17042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8200047011303462e-06,
+      "loss": 0.8311,
+      "step": 17043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.819471042671418e-06,
+      "loss": 0.8057,
+      "step": 17044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.818937445061021e-06,
+      "loss": 0.8281,
+      "step": 17045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8184039083093625e-06,
+      "loss": 0.9136,
+      "step": 17046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.81787043242665e-06,
+      "loss": 0.7886,
+      "step": 17047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8173370174230899e-06,
+      "loss": 0.8008,
+      "step": 17048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8168036633088898e-06,
+      "loss": 0.8682,
+      "step": 17049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8162703700942502e-06,
+      "loss": 0.7954,
+      "step": 17050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8157371377893769e-06,
+      "loss": 0.813,
+      "step": 17051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.81520396640447e-06,
+      "loss": 0.8037,
+      "step": 17052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8146708559497279e-06,
+      "loss": 0.8931,
+      "step": 17053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8141378064353538e-06,
+      "loss": 0.7959,
+      "step": 17054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8136048178715422e-06,
+      "loss": 0.8203,
+      "step": 17055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8130718902684908e-06,
+      "loss": 0.8994,
+      "step": 17056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8125390236363983e-06,
+      "loss": 0.7409,
+      "step": 17057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8120062179854576e-06,
+      "loss": 0.8167,
+      "step": 17058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8114734733258592e-06,
+      "loss": 0.8916,
+      "step": 17059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8109407896677994e-06,
+      "loss": 0.834,
+      "step": 17060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8104081670214663e-06,
+      "loss": 0.8013,
+      "step": 17061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8098756053970528e-06,
+      "loss": 0.8711,
+      "step": 17062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8093431048047445e-06,
+      "loss": 0.8755,
+      "step": 17063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8088106652547316e-06,
+      "loss": 0.8887,
+      "step": 17064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8082782867571997e-06,
+      "loss": 0.8711,
+      "step": 17065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8077459693223315e-06,
+      "loss": 0.8501,
+      "step": 17066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.807213712960314e-06,
+      "loss": 0.8057,
+      "step": 17067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8066815176813306e-06,
+      "loss": 0.8174,
+      "step": 17068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.80614938349556e-06,
+      "loss": 0.875,
+      "step": 17069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.805617310413187e-06,
+      "loss": 0.8477,
+      "step": 17070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.805085298444388e-06,
+      "loss": 0.8276,
+      "step": 17071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8045533475993405e-06,
+      "loss": 0.8564,
+      "step": 17072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8040214578882242e-06,
+      "loss": 0.8315,
+      "step": 17073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8034896293212118e-06,
+      "loss": 0.8882,
+      "step": 17074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8029578619084804e-06,
+      "loss": 0.8296,
+      "step": 17075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8024261556602042e-06,
+      "loss": 0.4108,
+      "step": 17076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.801894510586555e-06,
+      "loss": 0.8027,
+      "step": 17077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.801362926697703e-06,
+      "loss": 0.8711,
+      "step": 17078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8008314040038167e-06,
+      "loss": 0.8911,
+      "step": 17079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.800299942515067e-06,
+      "loss": 0.8135,
+      "step": 17080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7997685422416233e-06,
+      "loss": 0.8203,
+      "step": 17081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.799237203193649e-06,
+      "loss": 0.7905,
+      "step": 17082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7987059253813116e-06,
+      "loss": 0.8916,
+      "step": 17083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7981747088147755e-06,
+      "loss": 0.8535,
+      "step": 17084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7976435535042002e-06,
+      "loss": 0.8457,
+      "step": 17085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7971124594597522e-06,
+      "loss": 0.8872,
+      "step": 17086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7965814266915876e-06,
+      "loss": 0.854,
+      "step": 17087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.796050455209869e-06,
+      "loss": 0.8472,
+      "step": 17088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7955195450247552e-06,
+      "loss": 0.3962,
+      "step": 17089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7949886961464025e-06,
+      "loss": 0.8755,
+      "step": 17090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7944579085849662e-06,
+      "loss": 0.9102,
+      "step": 17091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7939271823505999e-06,
+      "loss": 0.8345,
+      "step": 17092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.793396517453459e-06,
+      "loss": 0.8784,
+      "step": 17093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7928659139036968e-06,
+      "loss": 0.8589,
+      "step": 17094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7923353717114622e-06,
+      "loss": 0.8608,
+      "step": 17095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7918048908869084e-06,
+      "loss": 0.8237,
+      "step": 17096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7912744714401824e-06,
+      "loss": 0.769,
+      "step": 17097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7907441133814303e-06,
+      "loss": 0.856,
+      "step": 17098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7902138167208021e-06,
+      "loss": 0.8809,
+      "step": 17099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.78968358146844e-06,
+      "loss": 0.8198,
+      "step": 17100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7891534076344895e-06,
+      "loss": 0.812,
+      "step": 17101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.788623295229096e-06,
+      "loss": 0.8584,
+      "step": 17102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7880932442623993e-06,
+      "loss": 0.8584,
+      "step": 17103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7875632547445387e-06,
+      "loss": 0.8613,
+      "step": 17104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7870333266856566e-06,
+      "loss": 0.8408,
+      "step": 17105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7865034600958886e-06,
+      "loss": 0.8579,
+      "step": 17106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7859736549853745e-06,
+      "loss": 0.8027,
+      "step": 17107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7854439113642475e-06,
+      "loss": 0.8384,
+      "step": 17108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7849142292426458e-06,
+      "loss": 0.8696,
+      "step": 17109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.784384608630701e-06,
+      "loss": 0.8989,
+      "step": 17110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7838550495385443e-06,
+      "loss": 0.8809,
+      "step": 17111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7833255519763105e-06,
+      "loss": 0.3792,
+      "step": 17112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7827961159541257e-06,
+      "loss": 0.3091,
+      "step": 17113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.782266741482121e-06,
+      "loss": 0.8604,
+      "step": 17114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7817374285704253e-06,
+      "loss": 0.8193,
+      "step": 17115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7812081772291645e-06,
+      "loss": 0.7832,
+      "step": 17116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7806789874684616e-06,
+      "loss": 0.8887,
+      "step": 17117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7801498592984445e-06,
+      "loss": 0.8716,
+      "step": 17118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7796207927292326e-06,
+      "loss": 0.8677,
+      "step": 17119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7790917877709512e-06,
+      "loss": 0.8584,
+      "step": 17120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7785628444337183e-06,
+      "loss": 0.8418,
+      "step": 17121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7780339627276554e-06,
+      "loss": 0.7437,
+      "step": 17122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7775051426628803e-06,
+      "loss": 0.8442,
+      "step": 17123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7769763842495086e-06,
+      "loss": 0.8481,
+      "step": 17124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7764476874976572e-06,
+      "loss": 0.8564,
+      "step": 17125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7759190524174425e-06,
+      "loss": 0.7964,
+      "step": 17126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7753904790189757e-06,
+      "loss": 0.8984,
+      "step": 17127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7748619673123717e-06,
+      "loss": 0.8418,
+      "step": 17128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7743335173077409e-06,
+      "loss": 0.9048,
+      "step": 17129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7738051290151908e-06,
+      "loss": 0.3269,
+      "step": 17130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7732768024448343e-06,
+      "loss": 0.3079,
+      "step": 17131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.772748537606776e-06,
+      "loss": 0.8647,
+      "step": 17132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7722203345111232e-06,
+      "loss": 0.812,
+      "step": 17133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7716921931679836e-06,
+      "loss": 0.7847,
+      "step": 17134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.771164113587459e-06,
+      "loss": 0.7725,
+      "step": 17135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7706360957796536e-06,
+      "loss": 0.7896,
+      "step": 17136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7701081397546665e-06,
+      "loss": 0.7983,
+      "step": 17137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7695802455226007e-06,
+      "loss": 0.8647,
+      "step": 17138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7690524130935565e-06,
+      "loss": 0.8052,
+      "step": 17139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7685246424776297e-06,
+      "loss": 0.8481,
+      "step": 17140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.76799693368492e-06,
+      "loss": 0.8384,
+      "step": 17141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7674692867255222e-06,
+      "loss": 0.8672,
+      "step": 17142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.766941701609529e-06,
+      "loss": 0.9136,
+      "step": 17143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7664141783470378e-06,
+      "loss": 0.8491,
+      "step": 17144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.765886716948137e-06,
+      "loss": 0.8896,
+      "step": 17145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7653593174229195e-06,
+      "loss": 0.7803,
+      "step": 17146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7648319797814767e-06,
+      "loss": 0.8804,
+      "step": 17147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7643047040338967e-06,
+      "loss": 0.8911,
+      "step": 17148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7637774901902665e-06,
+      "loss": 0.8462,
+      "step": 17149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7632503382606703e-06,
+      "loss": 0.8535,
+      "step": 17150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7627232482551965e-06,
+      "loss": 0.8711,
+      "step": 17151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7621962201839293e-06,
+      "loss": 0.8638,
+      "step": 17152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.761669254056949e-06,
+      "loss": 0.8237,
+      "step": 17153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7611423498843388e-06,
+      "loss": 0.8394,
+      "step": 17154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.760615507676181e-06,
+      "loss": 0.8135,
+      "step": 17155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7600887274425533e-06,
+      "loss": 0.8716,
+      "step": 17156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.759562009193534e-06,
+      "loss": 0.8135,
+      "step": 17157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7590353529391973e-06,
+      "loss": 0.8599,
+      "step": 17158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.758508758689622e-06,
+      "loss": 0.8384,
+      "step": 17159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7579822264548834e-06,
+      "loss": 0.8418,
+      "step": 17160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7574557562450523e-06,
+      "loss": 0.7881,
+      "step": 17161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7569293480702038e-06,
+      "loss": 0.7947,
+      "step": 17162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7564030019404071e-06,
+      "loss": 0.8579,
+      "step": 17163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7558767178657305e-06,
+      "loss": 0.8252,
+      "step": 17164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7553504958562462e-06,
+      "loss": 0.8447,
+      "step": 17165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7548243359220184e-06,
+      "loss": 0.8428,
+      "step": 17166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7542982380731144e-06,
+      "loss": 0.8076,
+      "step": 17167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7537722023196018e-06,
+      "loss": 0.874,
+      "step": 17168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7532462286715423e-06,
+      "loss": 0.8159,
+      "step": 17169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7527203171389967e-06,
+      "loss": 0.8281,
+      "step": 17170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.75219446773203e-06,
+      "loss": 0.8726,
+      "step": 17171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7516686804607004e-06,
+      "loss": 0.812,
+      "step": 17172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7511429553350684e-06,
+      "loss": 0.854,
+      "step": 17173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7506172923651898e-06,
+      "loss": 0.792,
+      "step": 17174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7500916915611248e-06,
+      "loss": 0.8164,
+      "step": 17175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7495661529329267e-06,
+      "loss": 0.8101,
+      "step": 17176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7490406764906481e-06,
+      "loss": 0.8164,
+      "step": 17177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7485152622443463e-06,
+      "loss": 0.8418,
+      "step": 17178
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7479899102040698e-06,
+      "loss": 0.8706,
+      "step": 17179
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7474646203798705e-06,
+      "loss": 0.8643,
+      "step": 17180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7469393927818e-06,
+      "loss": 0.8833,
+      "step": 17181
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7464142274199053e-06,
+      "loss": 0.8452,
+      "step": 17182
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7458891243042314e-06,
+      "loss": 0.8481,
+      "step": 17183
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7453640834448288e-06,
+      "loss": 0.8208,
+      "step": 17184
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7448391048517378e-06,
+      "loss": 0.8477,
+      "step": 17185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7443141885350057e-06,
+      "loss": 0.7661,
+      "step": 17186
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.743789334504672e-06,
+      "loss": 0.8579,
+      "step": 17187
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7432645427707812e-06,
+      "loss": 0.8452,
+      "step": 17188
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7427398133433714e-06,
+      "loss": 0.8496,
+      "step": 17189
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7422151462324804e-06,
+      "loss": 0.8911,
+      "step": 17190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7416905414481479e-06,
+      "loss": 0.8833,
+      "step": 17191
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7411659990004103e-06,
+      "loss": 0.3875,
+      "step": 17192
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7406415188993014e-06,
+      "loss": 0.8184,
+      "step": 17193
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.740117101154858e-06,
+      "loss": 0.9102,
+      "step": 17194
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.739592745777111e-06,
+      "loss": 0.8252,
+      "step": 17195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7390684527760915e-06,
+      "loss": 0.8569,
+      "step": 17196
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.738544222161832e-06,
+      "loss": 0.8638,
+      "step": 17197
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7380200539443597e-06,
+      "loss": 0.813,
+      "step": 17198
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7374959481337046e-06,
+      "loss": 0.854,
+      "step": 17199
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7369719047398936e-06,
+      "loss": 0.8701,
+      "step": 17200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7364479237729526e-06,
+      "loss": 0.4823,
+      "step": 17201
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7359240052429055e-06,
+      "loss": 0.8296,
+      "step": 17202
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7354001491597739e-06,
+      "loss": 0.8486,
+      "step": 17203
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7348763555335818e-06,
+      "loss": 0.8198,
+      "step": 17204
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7343526243743525e-06,
+      "loss": 0.8315,
+      "step": 17205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7338289556921012e-06,
+      "loss": 0.8267,
+      "step": 17206
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7333053494968505e-06,
+      "loss": 0.8706,
+      "step": 17207
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7327818057986162e-06,
+      "loss": 0.8428,
+      "step": 17208
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7322583246074132e-06,
+      "loss": 0.4182,
+      "step": 17209
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.731734905933259e-06,
+      "loss": 0.8047,
+      "step": 17210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7312115497861642e-06,
+      "loss": 0.8213,
+      "step": 17211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.730688256176144e-06,
+      "loss": 0.8271,
+      "step": 17212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7301650251132107e-06,
+      "loss": 0.8042,
+      "step": 17213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7296418566073736e-06,
+      "loss": 0.8525,
+      "step": 17214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7291187506686407e-06,
+      "loss": 0.8188,
+      "step": 17215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7285957073070186e-06,
+      "loss": 0.8008,
+      "step": 17216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7280727265325158e-06,
+      "loss": 0.8579,
+      "step": 17217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7275498083551396e-06,
+      "loss": 0.7642,
+      "step": 17218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7270269527848905e-06,
+      "loss": 0.8164,
+      "step": 17219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.726504159831775e-06,
+      "loss": 0.3895,
+      "step": 17220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.725981429505793e-06,
+      "loss": 0.8374,
+      "step": 17221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7254587618169445e-06,
+      "loss": 0.834,
+      "step": 17222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7249361567752309e-06,
+      "loss": 0.9258,
+      "step": 17223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7244136143906482e-06,
+      "loss": 0.7661,
+      "step": 17224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7238911346731946e-06,
+      "loss": 0.8535,
+      "step": 17225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7233687176328679e-06,
+      "loss": 0.8706,
+      "step": 17226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7228463632796604e-06,
+      "loss": 0.7642,
+      "step": 17227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.722324071623565e-06,
+      "loss": 0.8311,
+      "step": 17228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.721801842674577e-06,
+      "loss": 0.8638,
+      "step": 17229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.721279676442683e-06,
+      "loss": 0.8193,
+      "step": 17230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7207575729378778e-06,
+      "loss": 0.8428,
+      "step": 17231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7202355321701458e-06,
+      "loss": 0.771,
+      "step": 17232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.719713554149478e-06,
+      "loss": 0.3219,
+      "step": 17233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7191916388858592e-06,
+      "loss": 0.7803,
+      "step": 17234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7186697863892727e-06,
+      "loss": 0.7856,
+      "step": 17235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7181479966697034e-06,
+      "loss": 0.8325,
+      "step": 17236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7176262697371365e-06,
+      "loss": 0.8892,
+      "step": 17237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7171046056015494e-06,
+      "loss": 0.8623,
+      "step": 17238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7165830042729265e-06,
+      "loss": 0.8022,
+      "step": 17239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7160614657612447e-06,
+      "loss": 0.8359,
+      "step": 17240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7155399900764802e-06,
+      "loss": 0.8027,
+      "step": 17241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7150185772286128e-06,
+      "loss": 0.7632,
+      "step": 17242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714497227227615e-06,
+      "loss": 0.8052,
+      "step": 17243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7139759400834639e-06,
+      "loss": 0.7998,
+      "step": 17244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7134547158061299e-06,
+      "loss": 0.8438,
+      "step": 17245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7129335544055874e-06,
+      "loss": 0.8623,
+      "step": 17246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7124124558918066e-06,
+      "loss": 0.7681,
+      "step": 17247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7118914202747539e-06,
+      "loss": 0.9092,
+      "step": 17248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7113704475643993e-06,
+      "loss": 0.8203,
+      "step": 17249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.710849537770713e-06,
+      "loss": 0.8301,
+      "step": 17250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7103286909036559e-06,
+      "loss": 0.8301,
+      "step": 17251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.709807906973196e-06,
+      "loss": 0.8032,
+      "step": 17252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7092871859892962e-06,
+      "loss": 0.8335,
+      "step": 17253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7087665279619165e-06,
+      "loss": 0.8086,
+      "step": 17254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7082459329010203e-06,
+      "loss": 0.8071,
+      "step": 17255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7077254008165656e-06,
+      "loss": 0.8618,
+      "step": 17256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.707204931718512e-06,
+      "loss": 0.8296,
+      "step": 17257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.706684525616818e-06,
+      "loss": 0.7829,
+      "step": 17258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7061641825214387e-06,
+      "loss": 0.8042,
+      "step": 17259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7056439024423293e-06,
+      "loss": 0.8193,
+      "step": 17260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7051236853894414e-06,
+      "loss": 0.8472,
+      "step": 17261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.70460353137273e-06,
+      "loss": 0.8809,
+      "step": 17262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7040834404021473e-06,
+      "loss": 0.8794,
+      "step": 17263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7035634124876404e-06,
+      "loss": 0.8652,
+      "step": 17264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.70304344763916e-06,
+      "loss": 0.7812,
+      "step": 17265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7025235458666555e-06,
+      "loss": 0.8545,
+      "step": 17266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7020037071800715e-06,
+      "loss": 0.8853,
+      "step": 17267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7014839315893539e-06,
+      "loss": 0.8521,
+      "step": 17268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7009642191044445e-06,
+      "loss": 0.8086,
+      "step": 17269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7004445697352884e-06,
+      "loss": 0.8662,
+      "step": 17270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6999249834918292e-06,
+      "loss": 0.8994,
+      "step": 17271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6994054603840033e-06,
+      "loss": 0.8877,
+      "step": 17272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6988860004217534e-06,
+      "loss": 0.875,
+      "step": 17273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6983666036150164e-06,
+      "loss": 0.8711,
+      "step": 17274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6978472699737276e-06,
+      "loss": 0.8584,
+      "step": 17275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6973279995078258e-06,
+      "loss": 0.8877,
+      "step": 17276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.696808792227242e-06,
+      "loss": 0.3452,
+      "step": 17277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6962896481419105e-06,
+      "loss": 0.8213,
+      "step": 17278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6957705672617664e-06,
+      "loss": 0.8164,
+      "step": 17279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.695251549596738e-06,
+      "loss": 0.8018,
+      "step": 17280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6947325951567551e-06,
+      "loss": 0.8677,
+      "step": 17281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6942137039517447e-06,
+      "loss": 0.8408,
+      "step": 17282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.693694875991635e-06,
+      "loss": 0.9141,
+      "step": 17283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6931761112863543e-06,
+      "loss": 0.8208,
+      "step": 17284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6926574098458237e-06,
+      "loss": 0.8735,
+      "step": 17285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6921387716799703e-06,
+      "loss": 0.8262,
+      "step": 17286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6916201967987144e-06,
+      "loss": 0.7534,
+      "step": 17287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6911016852119766e-06,
+      "loss": 0.8706,
+      "step": 17288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6905832369296787e-06,
+      "loss": 0.8467,
+      "step": 17289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6900648519617375e-06,
+      "loss": 0.8218,
+      "step": 17290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6895465303180707e-06,
+      "loss": 0.8208,
+      "step": 17291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6890282720085977e-06,
+      "loss": 0.8936,
+      "step": 17292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6885100770432311e-06,
+      "loss": 0.8184,
+      "step": 17293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6879919454318833e-06,
+      "loss": 0.8394,
+      "step": 17294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6874738771844702e-06,
+      "loss": 0.8584,
+      "step": 17295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6869558723108998e-06,
+      "loss": 0.7734,
+      "step": 17296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6864379308210864e-06,
+      "loss": 0.834,
+      "step": 17297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6859200527249354e-06,
+      "loss": 0.8618,
+      "step": 17298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.685402238032357e-06,
+      "loss": 0.29,
+      "step": 17299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6848844867532577e-06,
+      "loss": 0.8916,
+      "step": 17300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6843667988975404e-06,
+      "loss": 0.8516,
+      "step": 17301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6838491744751107e-06,
+      "loss": 0.833,
+      "step": 17302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6833316134958739e-06,
+      "loss": 0.3531,
+      "step": 17303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6828141159697276e-06,
+      "loss": 0.3345,
+      "step": 17304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6822966819065767e-06,
+      "loss": 0.8735,
+      "step": 17305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6817793113163178e-06,
+      "loss": 0.8433,
+      "step": 17306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6812620042088484e-06,
+      "loss": 0.8379,
+      "step": 17307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.680744760594068e-06,
+      "loss": 0.832,
+      "step": 17308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6802275804818691e-06,
+      "loss": 0.8711,
+      "step": 17309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6797104638821494e-06,
+      "loss": 0.8149,
+      "step": 17310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6791934108047992e-06,
+      "loss": 0.8545,
+      "step": 17311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6786764212597139e-06,
+      "loss": 0.832,
+      "step": 17312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.678159495256782e-06,
+      "loss": 0.8086,
+      "step": 17313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.677642632805892e-06,
+      "loss": 0.856,
+      "step": 17314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.677125833916934e-06,
+      "loss": 0.8525,
+      "step": 17315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6766090985997962e-06,
+      "loss": 0.8486,
+      "step": 17316
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.676092426864362e-06,
+      "loss": 0.8057,
+      "step": 17317
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.675575818720519e-06,
+      "loss": 0.8477,
+      "step": 17318
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6750592741781496e-06,
+      "loss": 0.8242,
+      "step": 17319
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6745427932471336e-06,
+      "loss": 0.8433,
+      "step": 17320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6740263759373566e-06,
+      "loss": 0.8276,
+      "step": 17321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6735100222586937e-06,
+      "loss": 0.3726,
+      "step": 17322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6729937322210265e-06,
+      "loss": 0.8618,
+      "step": 17323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6724775058342335e-06,
+      "loss": 0.8755,
+      "step": 17324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6719613431081889e-06,
+      "loss": 0.8447,
+      "step": 17325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6714452440527685e-06,
+      "loss": 0.8164,
+      "step": 17326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6709292086778439e-06,
+      "loss": 0.8016,
+      "step": 17327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6704132369932896e-06,
+      "loss": 0.3964,
+      "step": 17328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6698973290089788e-06,
+      "loss": 0.8447,
+      "step": 17329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6693814847347778e-06,
+      "loss": 0.8306,
+      "step": 17330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.668865704180559e-06,
+      "loss": 0.8213,
+      "step": 17331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.668349987356188e-06,
+      "loss": 0.8457,
+      "step": 17332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6678343342715308e-06,
+      "loss": 0.854,
+      "step": 17333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6673187449364552e-06,
+      "loss": 0.8525,
+      "step": 17334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6668032193608218e-06,
+      "loss": 0.8423,
+      "step": 17335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6662877575544955e-06,
+      "loss": 0.7993,
+      "step": 17336
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.665772359527339e-06,
+      "loss": 0.834,
+      "step": 17337
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6652570252892114e-06,
+      "loss": 0.8833,
+      "step": 17338
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6647417548499706e-06,
+      "loss": 0.8149,
+      "step": 17339
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6642265482194769e-06,
+      "loss": 0.8545,
+      "step": 17340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6637114054075849e-06,
+      "loss": 0.8086,
+      "step": 17341
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6631963264241518e-06,
+      "loss": 0.8208,
+      "step": 17342
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6626813112790296e-06,
+      "loss": 0.8936,
+      "step": 17343
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6621663599820754e-06,
+      "loss": 0.8584,
+      "step": 17344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6616514725431377e-06,
+      "loss": 0.7713,
+      "step": 17345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6611366489720665e-06,
+      "loss": 0.8799,
+      "step": 17346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.660621889278714e-06,
+      "loss": 0.8506,
+      "step": 17347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6601071934729257e-06,
+      "loss": 0.8267,
+      "step": 17348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6595925615645498e-06,
+      "loss": 0.8857,
+      "step": 17349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6590779935634332e-06,
+      "loss": 0.8755,
+      "step": 17350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6585634894794195e-06,
+      "loss": 0.8823,
+      "step": 17351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6580490493223507e-06,
+      "loss": 0.234,
+      "step": 17352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6575346731020703e-06,
+      "loss": 0.8735,
+      "step": 17353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.657020360828418e-06,
+      "loss": 0.8721,
+      "step": 17354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6565061125112353e-06,
+      "loss": 0.8457,
+      "step": 17355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6559919281603576e-06,
+      "loss": 0.8213,
+      "step": 17356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6554778077856253e-06,
+      "loss": 0.8574,
+      "step": 17357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6549637513968736e-06,
+      "loss": 0.8027,
+      "step": 17358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6544497590039344e-06,
+      "loss": 0.7959,
+      "step": 17359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.653935830616643e-06,
+      "loss": 0.8677,
+      "step": 17360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6534219662448336e-06,
+      "loss": 0.8735,
+      "step": 17361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6529081658983343e-06,
+      "loss": 0.7871,
+      "step": 17362
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6523944295869776e-06,
+      "loss": 0.8521,
+      "step": 17363
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6518807573205903e-06,
+      "loss": 0.8442,
+      "step": 17364
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6513671491089989e-06,
+      "loss": 0.8013,
+      "step": 17365
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6508536049620323e-06,
+      "loss": 0.8823,
+      "step": 17366
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6503401248895118e-06,
+      "loss": 0.8228,
+      "step": 17367
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6498267089012638e-06,
+      "loss": 0.8535,
+      "step": 17368
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6493133570071114e-06,
+      "loss": 0.8647,
+      "step": 17369
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6488000692168742e-06,
+      "loss": 0.8574,
+      "step": 17370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6482868455403727e-06,
+      "loss": 0.8442,
+      "step": 17371
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6477736859874239e-06,
+      "loss": 0.8101,
+      "step": 17372
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6472605905678467e-06,
+      "loss": 0.8193,
+      "step": 17373
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.64674755929146e-06,
+      "loss": 0.8369,
+      "step": 17374
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6462345921680746e-06,
+      "loss": 0.8599,
+      "step": 17375
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6457216892075057e-06,
+      "loss": 0.8779,
+      "step": 17376
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.645208850419569e-06,
+      "loss": 0.8442,
+      "step": 17377
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.644696075814073e-06,
+      "loss": 0.369,
+      "step": 17378
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6441833654008282e-06,
+      "loss": 0.8354,
+      "step": 17379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6436707191896429e-06,
+      "loss": 0.8438,
+      "step": 17380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.643158137190325e-06,
+      "loss": 0.7725,
+      "step": 17381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6426456194126833e-06,
+      "loss": 0.8467,
+      "step": 17382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6421331658665203e-06,
+      "loss": 0.8228,
+      "step": 17383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6416207765616426e-06,
+      "loss": 0.855,
+      "step": 17384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6411084515078512e-06,
+      "loss": 0.8662,
+      "step": 17385
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.640596190714947e-06,
+      "loss": 0.8252,
+      "step": 17386
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6400839941927328e-06,
+      "loss": 0.8398,
+      "step": 17387
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6395718619510048e-06,
+      "loss": 0.3813,
+      "step": 17388
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6390597939995623e-06,
+      "loss": 0.8564,
+      "step": 17389
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6385477903482038e-06,
+      "loss": 0.8037,
+      "step": 17390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6380358510067229e-06,
+      "loss": 0.8423,
+      "step": 17391
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6375239759849143e-06,
+      "loss": 0.7939,
+      "step": 17392
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6370121652925691e-06,
+      "loss": 0.812,
+      "step": 17393
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.63650041893948e-06,
+      "loss": 0.7827,
+      "step": 17394
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6359887369354404e-06,
+      "loss": 0.8115,
+      "step": 17395
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6354771192902352e-06,
+      "loss": 0.8252,
+      "step": 17396
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6349655660136565e-06,
+      "loss": 0.7822,
+      "step": 17397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6344540771154889e-06,
+      "loss": 0.8516,
+      "step": 17398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6339426526055168e-06,
+      "loss": 0.8481,
+      "step": 17399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6334312924935275e-06,
+      "loss": 0.8162,
+      "step": 17400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.632919996789301e-06,
+      "loss": 0.8682,
+      "step": 17401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6324087655026204e-06,
+      "loss": 0.8896,
+      "step": 17402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6318975986432694e-06,
+      "loss": 0.8389,
+      "step": 17403
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.631386496221024e-06,
+      "loss": 0.834,
+      "step": 17404
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.630875458245662e-06,
+      "loss": 0.8467,
+      "step": 17405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6303644847269629e-06,
+      "loss": 0.8589,
+      "step": 17406
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6298535756746998e-06,
+      "loss": 0.894,
+      "step": 17407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6293427310986498e-06,
+      "loss": 0.8364,
+      "step": 17408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6288319510085831e-06,
+      "loss": 0.8379,
+      "step": 17409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6283212354142758e-06,
+      "loss": 0.855,
+      "step": 17410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6278105843254955e-06,
+      "loss": 0.8062,
+      "step": 17411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6272999977520115e-06,
+      "loss": 0.8638,
+      "step": 17412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6267894757035952e-06,
+      "loss": 0.8203,
+      "step": 17413
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6262790181900095e-06,
+      "loss": 0.7402,
+      "step": 17414
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6257686252210226e-06,
+      "loss": 0.7664,
+      "step": 17415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6252582968064006e-06,
+      "loss": 0.8574,
+      "step": 17416
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6247480329559057e-06,
+      "loss": 0.8076,
+      "step": 17417
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.624237833679298e-06,
+      "loss": 0.3879,
+      "step": 17418
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6237276989863416e-06,
+      "loss": 0.8809,
+      "step": 17419
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6232176288867929e-06,
+      "loss": 0.8662,
+      "step": 17420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6227076233904137e-06,
+      "loss": 0.895,
+      "step": 17421
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.622197682506958e-06,
+      "loss": 0.8491,
+      "step": 17422
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.621687806246185e-06,
+      "loss": 0.8779,
+      "step": 17423
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.621177994617848e-06,
+      "loss": 0.3214,
+      "step": 17424
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6206682476316987e-06,
+      "loss": 0.8291,
+      "step": 17425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.620158565297491e-06,
+      "loss": 0.8525,
+      "step": 17426
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6196489476249777e-06,
+      "loss": 0.7959,
+      "step": 17427
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6191393946239048e-06,
+      "loss": 0.8589,
+      "step": 17428
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6186299063040246e-06,
+      "loss": 0.8228,
+      "step": 17429
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6181204826750823e-06,
+      "loss": 0.7817,
+      "step": 17430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6176111237468234e-06,
+      "loss": 0.8906,
+      "step": 17431
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6171018295289953e-06,
+      "loss": 0.8252,
+      "step": 17432
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.616592600031338e-06,
+      "loss": 0.834,
+      "step": 17433
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6160834352635956e-06,
+      "loss": 0.8525,
+      "step": 17434
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6155743352355118e-06,
+      "loss": 0.7905,
+      "step": 17435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6150652999568235e-06,
+      "loss": 0.8262,
+      "step": 17436
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6145563294372702e-06,
+      "loss": 0.7754,
+      "step": 17437
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6140474236865876e-06,
+      "loss": 0.7979,
+      "step": 17438
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6135385827145135e-06,
+      "loss": 0.8989,
+      "step": 17439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6130298065307843e-06,
+      "loss": 0.8096,
+      "step": 17440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6125210951451304e-06,
+      "loss": 0.8462,
+      "step": 17441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6120124485672872e-06,
+      "loss": 0.8057,
+      "step": 17442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6115038668069854e-06,
+      "loss": 0.8374,
+      "step": 17443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6109953498739522e-06,
+      "loss": 0.8662,
+      "step": 17444
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6104868977779198e-06,
+      "loss": 0.8779,
+      "step": 17445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6099785105286132e-06,
+      "loss": 0.8271,
+      "step": 17446
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6094701881357593e-06,
+      "loss": 0.8301,
+      "step": 17447
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6089619306090848e-06,
+      "loss": 0.8301,
+      "step": 17448
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6084537379583127e-06,
+      "loss": 0.751,
+      "step": 17449
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6079456101931652e-06,
+      "loss": 0.8955,
+      "step": 17450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6074375473233612e-06,
+      "loss": 0.854,
+      "step": 17451
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6069295493586235e-06,
+      "loss": 0.8799,
+      "step": 17452
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6064216163086716e-06,
+      "loss": 0.8403,
+      "step": 17453
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6059137481832198e-06,
+      "loss": 0.8008,
+      "step": 17454
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6054059449919885e-06,
+      "loss": 0.832,
+      "step": 17455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6048982067446905e-06,
+      "loss": 0.363,
+      "step": 17456
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6043905334510384e-06,
+      "loss": 0.8022,
+      "step": 17457
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6038829251207477e-06,
+      "loss": 0.8794,
+      "step": 17458
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6033753817635266e-06,
+      "loss": 0.8457,
+      "step": 17459
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.602867903389087e-06,
+      "loss": 0.8799,
+      "step": 17460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6023604900071394e-06,
+      "loss": 0.8101,
+      "step": 17461
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6018531416273897e-06,
+      "loss": 0.8481,
+      "step": 17462
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.601345858259542e-06,
+      "loss": 0.8193,
+      "step": 17463
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6008386399133057e-06,
+      "loss": 0.8364,
+      "step": 17464
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.600331486598381e-06,
+      "loss": 0.7998,
+      "step": 17465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5998243983244739e-06,
+      "loss": 0.8091,
+      "step": 17466
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.599317375101282e-06,
+      "loss": 0.8311,
+      "step": 17467
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5988104169385093e-06,
+      "loss": 0.7837,
+      "step": 17468
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5983035238458532e-06,
+      "loss": 0.8491,
+      "step": 17469
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5977966958330094e-06,
+      "loss": 0.814,
+      "step": 17470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5972899329096757e-06,
+      "loss": 0.8066,
+      "step": 17471
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5967832350855494e-06,
+      "loss": 0.8096,
+      "step": 17472
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5962766023703208e-06,
+      "loss": 0.2992,
+      "step": 17473
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5957700347736864e-06,
+      "loss": 0.832,
+      "step": 17474
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5952635323053355e-06,
+      "loss": 0.8174,
+      "step": 17475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.594757094974957e-06,
+      "loss": 0.8579,
+      "step": 17476
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5942507227922427e-06,
+      "loss": 0.8257,
+      "step": 17477
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.593744415766877e-06,
+      "loss": 0.8984,
+      "step": 17478
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5932381739085501e-06,
+      "loss": 0.7998,
+      "step": 17479
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5927319972269434e-06,
+      "loss": 0.8569,
+      "step": 17480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5922258857317447e-06,
+      "loss": 0.8618,
+      "step": 17481
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.591719839432635e-06,
+      "loss": 0.8887,
+      "step": 17482
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5912138583392932e-06,
+      "loss": 0.8242,
+      "step": 17483
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5907079424614024e-06,
+      "loss": 0.832,
+      "step": 17484
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.590202091808643e-06,
+      "loss": 0.8247,
+      "step": 17485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5896963063906884e-06,
+      "loss": 0.8472,
+      "step": 17486
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5891905862172196e-06,
+      "loss": 0.8931,
+      "step": 17487
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.588684931297908e-06,
+      "loss": 0.8242,
+      "step": 17488
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5881793416424307e-06,
+      "loss": 0.7988,
+      "step": 17489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5876738172604589e-06,
+      "loss": 0.9365,
+      "step": 17490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5871683581616626e-06,
+      "loss": 0.8481,
+      "step": 17491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.586662964355713e-06,
+      "loss": 0.8594,
+      "step": 17492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5861576358522817e-06,
+      "loss": 0.8232,
+      "step": 17493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.585652372661033e-06,
+      "loss": 0.875,
+      "step": 17494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5851471747916365e-06,
+      "loss": 0.8823,
+      "step": 17495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.584642042253755e-06,
+      "loss": 0.8198,
+      "step": 17496
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5841369750570523e-06,
+      "loss": 0.8799,
+      "step": 17497
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5836319732111933e-06,
+      "loss": 0.8267,
+      "step": 17498
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.583127036725836e-06,
+      "loss": 0.8652,
+      "step": 17499
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5826221656106433e-06,
+      "loss": 0.8154,
+      "step": 17500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5821173598752754e-06,
+      "loss": 0.8521,
+      "step": 17501
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5816126195293886e-06,
+      "loss": 0.7285,
+      "step": 17502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5811079445826388e-06,
+      "loss": 0.7646,
+      "step": 17503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5806033350446798e-06,
+      "loss": 0.8872,
+      "step": 17504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5800987909251668e-06,
+      "loss": 0.8735,
+      "step": 17505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5795943122337555e-06,
+      "loss": 0.8496,
+      "step": 17506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.579089898980093e-06,
+      "loss": 0.8623,
+      "step": 17507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.578585551173833e-06,
+      "loss": 0.8369,
+      "step": 17508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5780812688246228e-06,
+      "loss": 0.8452,
+      "step": 17509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5775770519421086e-06,
+      "loss": 0.8447,
+      "step": 17510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5770729005359403e-06,
+      "loss": 0.8311,
+      "step": 17511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5765688146157593e-06,
+      "loss": 0.8213,
+      "step": 17512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5760647941912116e-06,
+      "loss": 0.7996,
+      "step": 17513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5755608392719418e-06,
+      "loss": 0.8071,
+      "step": 17514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5750569498675893e-06,
+      "loss": 0.8799,
+      "step": 17515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5745531259877944e-06,
+      "loss": 0.3265,
+      "step": 17516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5740493676421943e-06,
+      "loss": 0.8813,
+      "step": 17517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5735456748404281e-06,
+      "loss": 0.8379,
+      "step": 17518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5730420475921349e-06,
+      "loss": 0.8311,
+      "step": 17519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5725384859069454e-06,
+      "loss": 0.8799,
+      "step": 17520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.572034989794497e-06,
+      "loss": 0.8589,
+      "step": 17521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5715315592644214e-06,
+      "loss": 0.8828,
+      "step": 17522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5710281943263483e-06,
+      "loss": 0.8174,
+      "step": 17523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5705248949899104e-06,
+      "loss": 0.8721,
+      "step": 17524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5700216612647346e-06,
+      "loss": 0.8652,
+      "step": 17525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5695184931604484e-06,
+      "loss": 0.7871,
+      "step": 17526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5690153906866818e-06,
+      "loss": 0.8105,
+      "step": 17527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5685123538530567e-06,
+      "loss": 0.3417,
+      "step": 17528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.568009382669196e-06,
+      "loss": 0.8857,
+      "step": 17529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5675064771447257e-06,
+      "loss": 0.8428,
+      "step": 17530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5670036372892638e-06,
+      "loss": 0.8589,
+      "step": 17531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5665008631124335e-06,
+      "loss": 0.8462,
+      "step": 17532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5659981546238506e-06,
+      "loss": 0.8965,
+      "step": 17533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.565495511833136e-06,
+      "loss": 0.3352,
+      "step": 17534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5649929347499048e-06,
+      "loss": 0.7881,
+      "step": 17535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.56449042338377e-06,
+      "loss": 0.8486,
+      "step": 17536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5639879777443462e-06,
+      "loss": 0.8774,
+      "step": 17537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5634855978412495e-06,
+      "loss": 0.8662,
+      "step": 17538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5629832836840864e-06,
+      "loss": 0.8013,
+      "step": 17539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5624810352824709e-06,
+      "loss": 0.8042,
+      "step": 17540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.56197885264601e-06,
+      "loss": 0.8345,
+      "step": 17541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5614767357843097e-06,
+      "loss": 0.8291,
+      "step": 17542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5609746847069795e-06,
+      "loss": 0.915,
+      "step": 17543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5604726994236214e-06,
+      "loss": 0.8472,
+      "step": 17544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.559970779943842e-06,
+      "loss": 0.853,
+      "step": 17545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5594689262772407e-06,
+      "loss": 0.8735,
+      "step": 17546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5589671384334221e-06,
+      "loss": 0.8066,
+      "step": 17547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5584654164219847e-06,
+      "loss": 0.8525,
+      "step": 17548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5579637602525255e-06,
+      "loss": 0.8516,
+      "step": 17549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5574621699346433e-06,
+      "loss": 0.8608,
+      "step": 17550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5569606454779362e-06,
+      "loss": 0.8643,
+      "step": 17551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.556459186891996e-06,
+      "loss": 0.8599,
+      "step": 17552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5559577941864195e-06,
+      "loss": 0.8174,
+      "step": 17553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5554564673707973e-06,
+      "loss": 0.8804,
+      "step": 17554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5549552064547195e-06,
+      "loss": 0.3357,
+      "step": 17555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5544540114477785e-06,
+      "loss": 0.7729,
+      "step": 17556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5539528823595607e-06,
+      "loss": 0.8037,
+      "step": 17557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5534518191996544e-06,
+      "loss": 0.8784,
+      "step": 17558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5529508219776473e-06,
+      "loss": 0.835,
+      "step": 17559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5524498907031232e-06,
+      "loss": 0.3894,
+      "step": 17560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.551949025385665e-06,
+      "loss": 0.8271,
+      "step": 17561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.551448226034854e-06,
+      "loss": 0.8359,
+      "step": 17562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5509474926602725e-06,
+      "loss": 0.9019,
+      "step": 17563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5504468252715022e-06,
+      "loss": 0.8564,
+      "step": 17564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5499462238781183e-06,
+      "loss": 0.8433,
+      "step": 17565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5494456884897014e-06,
+      "loss": 0.8472,
+      "step": 17566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5489452191158262e-06,
+      "loss": 0.8633,
+      "step": 17567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5484448157660648e-06,
+      "loss": 0.8447,
+      "step": 17568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5479444784499954e-06,
+      "loss": 0.8501,
+      "step": 17569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5474442071771855e-06,
+      "loss": 0.7939,
+      "step": 17570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.546944001957209e-06,
+      "loss": 0.8438,
+      "step": 17571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.546443862799637e-06,
+      "loss": 0.8462,
+      "step": 17572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5459437897140357e-06,
+      "loss": 0.8223,
+      "step": 17573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5454437827099706e-06,
+      "loss": 0.8032,
+      "step": 17574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5449438417970113e-06,
+      "loss": 0.813,
+      "step": 17575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.54444396698472e-06,
+      "loss": 0.8418,
+      "step": 17576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.543944158282662e-06,
+      "loss": 0.8604,
+      "step": 17577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5434444157003963e-06,
+      "loss": 0.7642,
+      "step": 17578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5429447392474878e-06,
+      "loss": 0.8501,
+      "step": 17579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5424451289334941e-06,
+      "loss": 0.8267,
+      "step": 17580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5419455847679721e-06,
+      "loss": 0.853,
+      "step": 17581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5414461067604814e-06,
+      "loss": 0.7749,
+      "step": 17582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5409466949205754e-06,
+      "loss": 0.8892,
+      "step": 17583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5404473492578097e-06,
+      "loss": 0.7729,
+      "step": 17584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.53994806978174e-06,
+      "loss": 0.8726,
+      "step": 17585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5394488565019155e-06,
+      "loss": 0.8569,
+      "step": 17586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5389497094278861e-06,
+      "loss": 0.8374,
+      "step": 17587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.538450628569204e-06,
+      "loss": 0.8599,
+      "step": 17588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.537951613935415e-06,
+      "loss": 0.8608,
+      "step": 17589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5374526655360683e-06,
+      "loss": 0.9185,
+      "step": 17590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5369537833807068e-06,
+      "loss": 0.8486,
+      "step": 17591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5364549674788786e-06,
+      "loss": 0.8711,
+      "step": 17592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5359562178401244e-06,
+      "loss": 0.8906,
+      "step": 17593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5354575344739847e-06,
+      "loss": 0.3125,
+      "step": 17594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.534958917390002e-06,
+      "loss": 0.7729,
+      "step": 17595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.534460366597717e-06,
+      "loss": 0.8457,
+      "step": 17596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5339618821066648e-06,
+      "loss": 0.8198,
+      "step": 17597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5334634639263846e-06,
+      "loss": 0.8403,
+      "step": 17598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5329651120664102e-06,
+      "loss": 0.876,
+      "step": 17599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5324668265362774e-06,
+      "loss": 0.8115,
+      "step": 17600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5319686073455188e-06,
+      "loss": 0.8667,
+      "step": 17601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5314704545036634e-06,
+      "loss": 0.8442,
+      "step": 17602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.530972368020245e-06,
+      "loss": 0.8691,
+      "step": 17603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5304743479047923e-06,
+      "loss": 0.8267,
+      "step": 17604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5299763941668317e-06,
+      "loss": 0.718,
+      "step": 17605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5294785068158919e-06,
+      "loss": 0.8579,
+      "step": 17606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5289806858614975e-06,
+      "loss": 0.8677,
+      "step": 17607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5284829313131704e-06,
+      "loss": 0.854,
+      "step": 17608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.527985243180437e-06,
+      "loss": 0.8838,
+      "step": 17609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5274876214728158e-06,
+      "loss": 0.8408,
+      "step": 17610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5269900661998277e-06,
+      "loss": 0.8623,
+      "step": 17611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5264925773709943e-06,
+      "loss": 0.8579,
+      "step": 17612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.525995154995832e-06,
+      "loss": 0.8486,
+      "step": 17613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.525497799083856e-06,
+      "loss": 0.8218,
+      "step": 17614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5250005096445807e-06,
+      "loss": 0.8174,
+      "step": 17615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.524503286687522e-06,
+      "loss": 0.8799,
+      "step": 17616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5240061302221932e-06,
+      "loss": 0.8623,
+      "step": 17617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5235090402581033e-06,
+      "loss": 0.874,
+      "step": 17618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5230120168047652e-06,
+      "loss": 0.8652,
+      "step": 17619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5225150598716864e-06,
+      "loss": 0.8418,
+      "step": 17620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5220181694683727e-06,
+      "loss": 0.8516,
+      "step": 17621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5215213456043336e-06,
+      "loss": 0.8208,
+      "step": 17622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5210245882890711e-06,
+      "loss": 0.3411,
+      "step": 17623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5205278975320908e-06,
+      "loss": 0.873,
+      "step": 17624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5200312733428958e-06,
+      "loss": 0.8506,
+      "step": 17625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5195347157309865e-06,
+      "loss": 0.8584,
+      "step": 17626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5190382247058621e-06,
+      "loss": 0.8359,
+      "step": 17627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5185418002770208e-06,
+      "loss": 0.8691,
+      "step": 17628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.518045442453961e-06,
+      "loss": 0.7881,
+      "step": 17629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.51754915124618e-06,
+      "loss": 0.8618,
+      "step": 17630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.51705292666317e-06,
+      "loss": 0.8428,
+      "step": 17631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5165567687144273e-06,
+      "loss": 0.8599,
+      "step": 17632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5160606774094434e-06,
+      "loss": 0.8276,
+      "step": 17633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5155646527577068e-06,
+      "loss": 0.7896,
+      "step": 17634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5150686947687104e-06,
+      "loss": 0.8315,
+      "step": 17635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5145728034519401e-06,
+      "loss": 0.8135,
+      "step": 17636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5140769788168847e-06,
+      "loss": 0.8374,
+      "step": 17637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.513581220873031e-06,
+      "loss": 0.8784,
+      "step": 17638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.513085529629863e-06,
+      "loss": 0.791,
+      "step": 17639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5125899050968619e-06,
+      "loss": 0.8257,
+      "step": 17640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5120943472835125e-06,
+      "loss": 0.8735,
+      "step": 17641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5115988561992928e-06,
+      "loss": 0.873,
+      "step": 17642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.511103431853686e-06,
+      "loss": 0.8599,
+      "step": 17643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5106080742561667e-06,
+      "loss": 0.8257,
+      "step": 17644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5101127834162149e-06,
+      "loss": 0.8555,
+      "step": 17645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5096175593433054e-06,
+      "loss": 0.3362,
+      "step": 17646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5091224020469102e-06,
+      "loss": 0.834,
+      "step": 17647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5086273115365063e-06,
+      "loss": 0.8247,
+      "step": 17648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5081322878215615e-06,
+      "loss": 0.7954,
+      "step": 17649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5076373309115488e-06,
+      "loss": 0.8916,
+      "step": 17650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5071424408159392e-06,
+      "loss": 0.7881,
+      "step": 17651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5066476175441985e-06,
+      "loss": 0.8457,
+      "step": 17652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5061528611057917e-06,
+      "loss": 0.8218,
+      "step": 17653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5056581715101887e-06,
+      "loss": 0.8643,
+      "step": 17654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5051635487668491e-06,
+      "loss": 0.8735,
+      "step": 17655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5046689928852398e-06,
+      "loss": 0.8242,
+      "step": 17656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504174503874819e-06,
+      "loss": 0.8384,
+      "step": 17657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5036800817450503e-06,
+      "loss": 0.8271,
+      "step": 17658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5031857265053913e-06,
+      "loss": 0.8569,
+      "step": 17659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5026914381652973e-06,
+      "loss": 0.833,
+      "step": 17660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021972167342281e-06,
+      "loss": 0.79,
+      "step": 17661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501703062221639e-06,
+      "loss": 0.8062,
+      "step": 17662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012089746369811e-06,
+      "loss": 0.8188,
+      "step": 17663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5007149539897114e-06,
+      "loss": 0.9214,
+      "step": 17664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500221000289278e-06,
+      "loss": 0.8462,
+      "step": 17665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4997271135451307e-06,
+      "loss": 0.8555,
+      "step": 17666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992332937667215e-06,
+      "loss": 0.8179,
+      "step": 17667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987395409634936e-06,
+      "loss": 0.8125,
+      "step": 17668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498245855144896e-06,
+      "loss": 0.8716,
+      "step": 17669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4977522363203744e-06,
+      "loss": 0.7871,
+      "step": 17670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4972586844993713e-06,
+      "loss": 0.8643,
+      "step": 17671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4967651996913295e-06,
+      "loss": 0.8164,
+      "step": 17672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4962717819056883e-06,
+      "loss": 0.8169,
+      "step": 17673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.495778431151889e-06,
+      "loss": 0.8237,
+      "step": 17674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.495285147439372e-06,
+      "loss": 0.9087,
+      "step": 17675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4947919307775716e-06,
+      "loss": 0.8442,
+      "step": 17676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4942987811759264e-06,
+      "loss": 0.853,
+      "step": 17677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4938056986438693e-06,
+      "loss": 0.8755,
+      "step": 17678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4933126831908335e-06,
+      "loss": 0.8867,
+      "step": 17679
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492819734826253e-06,
+      "loss": 0.8667,
+      "step": 17680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923268535595559e-06,
+      "loss": 0.8325,
+      "step": 17681
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918340394001735e-06,
+      "loss": 0.9287,
+      "step": 17682
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491341292357536e-06,
+      "loss": 0.8452,
+      "step": 17683
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4908486124410686e-06,
+      "loss": 0.9141,
+      "step": 17684
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4903559996601969e-06,
+      "loss": 0.8032,
+      "step": 17685
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4898634540243434e-06,
+      "loss": 0.8643,
+      "step": 17686
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4893709755429337e-06,
+      "loss": 0.8315,
+      "step": 17687
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.488878564225391e-06,
+      "loss": 0.8584,
+      "step": 17688
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4883862200811322e-06,
+      "loss": 0.8291,
+      "step": 17689
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.487893943119581e-06,
+      "loss": 0.8501,
+      "step": 17690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4874017333501522e-06,
+      "loss": 0.8418,
+      "step": 17691
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4869095907822622e-06,
+      "loss": 0.7788,
+      "step": 17692
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.486417515425329e-06,
+      "loss": 0.9028,
+      "step": 17693
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4859255072887647e-06,
+      "loss": 0.7988,
+      "step": 17694
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4854335663819818e-06,
+      "loss": 0.8091,
+      "step": 17695
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4849416927143956e-06,
+      "loss": 0.8457,
+      "step": 17696
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4844498862954132e-06,
+      "loss": 0.833,
+      "step": 17697
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4839581471344428e-06,
+      "loss": 0.7744,
+      "step": 17698
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4834664752408956e-06,
+      "loss": 0.812,
+      "step": 17699
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4829748706241743e-06,
+      "loss": 0.8379,
+      "step": 17700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4824833332936873e-06,
+      "loss": 0.9155,
+      "step": 17701
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4819918632588355e-06,
+      "loss": 0.832,
+      "step": 17702
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4815004605290245e-06,
+      "loss": 0.8613,
+      "step": 17703
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4810091251136543e-06,
+      "loss": 0.7964,
+      "step": 17704
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4805178570221234e-06,
+      "loss": 0.812,
+      "step": 17705
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.480026656263832e-06,
+      "loss": 0.8501,
+      "step": 17706
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4795355228481789e-06,
+      "loss": 0.8159,
+      "step": 17707
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.479044456784557e-06,
+      "loss": 0.8101,
+      "step": 17708
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4785534580823647e-06,
+      "loss": 0.8101,
+      "step": 17709
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4780625267509924e-06,
+      "loss": 0.8413,
+      "step": 17710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.477571662799835e-06,
+      "loss": 0.8081,
+      "step": 17711
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.477080866238283e-06,
+      "loss": 0.8184,
+      "step": 17712
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4765901370757236e-06,
+      "loss": 0.8242,
+      "step": 17713
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4760994753215486e-06,
+      "loss": 0.8359,
+      "step": 17714
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4756088809851416e-06,
+      "loss": 0.897,
+      "step": 17715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4751183540758911e-06,
+      "loss": 0.8389,
+      "step": 17716
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4746278946031823e-06,
+      "loss": 0.8457,
+      "step": 17717
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4741375025763971e-06,
+      "loss": 0.8345,
+      "step": 17718
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4736471780049162e-06,
+      "loss": 0.7656,
+      "step": 17719
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4731569208981234e-06,
+      "loss": 0.8037,
+      "step": 17720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.472666731265394e-06,
+      "loss": 0.354,
+      "step": 17721
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4721766091161105e-06,
+      "loss": 0.8125,
+      "step": 17722
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.471686554459646e-06,
+      "loss": 0.8599,
+      "step": 17723
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.471196567305379e-06,
+      "loss": 0.8184,
+      "step": 17724
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.470706647662683e-06,
+      "loss": 0.8262,
+      "step": 17725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.470216795540928e-06,
+      "loss": 0.8223,
+      "step": 17726
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4697270109494882e-06,
+      "loss": 0.8745,
+      "step": 17727
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4692372938977356e-06,
+      "loss": 0.8682,
+      "step": 17728
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4687476443950355e-06,
+      "loss": 0.8472,
+      "step": 17729
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4682580624507593e-06,
+      "loss": 0.8545,
+      "step": 17730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4677685480742714e-06,
+      "loss": 0.8682,
+      "step": 17731
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4672791012749355e-06,
+      "loss": 0.8628,
+      "step": 17732
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4667897220621191e-06,
+      "loss": 0.3796,
+      "step": 17733
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4663004104451817e-06,
+      "loss": 0.8594,
+      "step": 17734
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4658111664334857e-06,
+      "loss": 0.7661,
+      "step": 17735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4653219900363925e-06,
+      "loss": 0.7866,
+      "step": 17736
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.46483288126326e-06,
+      "loss": 0.855,
+      "step": 17737
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.464343840123445e-06,
+      "loss": 0.8423,
+      "step": 17738
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4638548666263026e-06,
+      "loss": 0.8569,
+      "step": 17739
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4633659607811884e-06,
+      "loss": 0.853,
+      "step": 17740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.462877122597458e-06,
+      "loss": 0.8325,
+      "step": 17741
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4623883520844611e-06,
+      "loss": 0.8901,
+      "step": 17742
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4618996492515508e-06,
+      "loss": 0.8442,
+      "step": 17743
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.461411014108075e-06,
+      "loss": 0.8086,
+      "step": 17744
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.460922446663382e-06,
+      "loss": 0.7949,
+      "step": 17745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4604339469268202e-06,
+      "loss": 0.8286,
+      "step": 17746
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459945514907734e-06,
+      "loss": 0.8687,
+      "step": 17747
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459457150615468e-06,
+      "loss": 0.8853,
+      "step": 17748
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.458968854059367e-06,
+      "loss": 0.8838,
+      "step": 17749
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4584806252487727e-06,
+      "loss": 0.8174,
+      "step": 17750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4579924641930242e-06,
+      "loss": 0.8638,
+      "step": 17751
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4575043709014603e-06,
+      "loss": 0.8618,
+      "step": 17752
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4570163453834197e-06,
+      "loss": 0.8408,
+      "step": 17753
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4565283876482412e-06,
+      "loss": 0.8184,
+      "step": 17754
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4560404977052567e-06,
+      "loss": 0.8291,
+      "step": 17755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4555526755638038e-06,
+      "loss": 0.8555,
+      "step": 17756
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4550649212332135e-06,
+      "loss": 0.814,
+      "step": 17757
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.454577234722816e-06,
+      "loss": 0.8198,
+      "step": 17758
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4540896160419438e-06,
+      "loss": 0.7808,
+      "step": 17759
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4536020651999238e-06,
+      "loss": 0.8223,
+      "step": 17760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4531145822060849e-06,
+      "loss": 0.8579,
+      "step": 17761
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4526271670697545e-06,
+      "loss": 0.8394,
+      "step": 17762
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4521398198002568e-06,
+      "loss": 0.8599,
+      "step": 17763
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4516525404069132e-06,
+      "loss": 0.8091,
+      "step": 17764
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4511653288990496e-06,
+      "loss": 0.8408,
+      "step": 17765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4506781852859836e-06,
+      "loss": 0.856,
+      "step": 17766
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4501911095770388e-06,
+      "loss": 0.8921,
+      "step": 17767
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4497041017815306e-06,
+      "loss": 0.8457,
+      "step": 17768
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4492171619087786e-06,
+      "loss": 0.8633,
+      "step": 17769
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4487302899680984e-06,
+      "loss": 0.8521,
+      "step": 17770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4482434859688016e-06,
+      "loss": 0.7969,
+      "step": 17771
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.447756749920204e-06,
+      "loss": 0.8193,
+      "step": 17772
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4472700818316187e-06,
+      "loss": 0.8062,
+      "step": 17773
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4467834817123538e-06,
+      "loss": 0.7405,
+      "step": 17774
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4462969495717217e-06,
+      "loss": 0.8403,
+      "step": 17775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4458104854190285e-06,
+      "loss": 0.8364,
+      "step": 17776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4453240892635805e-06,
+      "loss": 0.8442,
+      "step": 17777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4448377611146853e-06,
+      "loss": 0.8438,
+      "step": 17778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4443515009816444e-06,
+      "loss": 0.8188,
+      "step": 17779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4438653088737642e-06,
+      "loss": 0.8911,
+      "step": 17780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4433791848003425e-06,
+      "loss": 0.894,
+      "step": 17781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4428931287706831e-06,
+      "loss": 0.8276,
+      "step": 17782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4424071407940837e-06,
+      "loss": 0.8447,
+      "step": 17783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.44192122087984e-06,
+      "loss": 0.8706,
+      "step": 17784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4414353690372501e-06,
+      "loss": 0.8584,
+      "step": 17785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4409495852756106e-06,
+      "loss": 0.8398,
+      "step": 17786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4404638696042129e-06,
+      "loss": 0.895,
+      "step": 17787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4399782220323515e-06,
+      "loss": 0.8691,
+      "step": 17788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4394926425693172e-06,
+      "loss": 0.7627,
+      "step": 17789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4390071312243975e-06,
+      "loss": 0.7815,
+      "step": 17790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4385216880068842e-06,
+      "loss": 0.8599,
+      "step": 17791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4380363129260616e-06,
+      "loss": 0.8501,
+      "step": 17792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4375510059912173e-06,
+      "loss": 0.936,
+      "step": 17793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4370657672116377e-06,
+      "loss": 0.8418,
+      "step": 17794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4365805965966045e-06,
+      "loss": 0.8145,
+      "step": 17795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4360954941554001e-06,
+      "loss": 0.876,
+      "step": 17796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4356104598973024e-06,
+      "loss": 0.8545,
+      "step": 17797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.435125493831594e-06,
+      "loss": 0.8452,
+      "step": 17798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.434640595967554e-06,
+      "loss": 0.7905,
+      "step": 17799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4341557663144561e-06,
+      "loss": 0.832,
+      "step": 17800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4336710048815788e-06,
+      "loss": 0.8179,
+      "step": 17801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4331863116781952e-06,
+      "loss": 0.8521,
+      "step": 17802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4327016867135768e-06,
+      "loss": 0.8281,
+      "step": 17803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4322171299969978e-06,
+      "loss": 0.8057,
+      "step": 17804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4317326415377252e-06,
+      "loss": 0.8301,
+      "step": 17805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4312482213450307e-06,
+      "loss": 0.7485,
+      "step": 17806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4307638694281828e-06,
+      "loss": 0.8672,
+      "step": 17807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4302795857964463e-06,
+      "loss": 0.4568,
+      "step": 17808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.429795370459085e-06,
+      "loss": 0.8135,
+      "step": 17809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.429311223425366e-06,
+      "loss": 0.8447,
+      "step": 17810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4288271447045482e-06,
+      "loss": 0.8906,
+      "step": 17811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4283431343058961e-06,
+      "loss": 0.8833,
+      "step": 17812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4278591922386665e-06,
+      "loss": 0.8315,
+      "step": 17813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4273753185121213e-06,
+      "loss": 0.8047,
+      "step": 17814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.426891513135516e-06,
+      "loss": 0.8237,
+      "step": 17815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.426407776118105e-06,
+      "loss": 0.8369,
+      "step": 17816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.425924107469146e-06,
+      "loss": 0.8872,
+      "step": 17817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4254405071978894e-06,
+      "loss": 0.7773,
+      "step": 17818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.424956975313589e-06,
+      "loss": 0.8418,
+      "step": 17819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4244735118254959e-06,
+      "loss": 0.8843,
+      "step": 17820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4239901167428593e-06,
+      "loss": 0.833,
+      "step": 17821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4235067900749254e-06,
+      "loss": 0.9038,
+      "step": 17822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4230235318309443e-06,
+      "loss": 0.8701,
+      "step": 17823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4225403420201579e-06,
+      "loss": 0.3448,
+      "step": 17824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4220572206518136e-06,
+      "loss": 0.8643,
+      "step": 17825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4215741677351514e-06,
+      "loss": 0.8584,
+      "step": 17826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4210911832794139e-06,
+      "loss": 0.8833,
+      "step": 17827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4206082672938432e-06,
+      "loss": 0.8022,
+      "step": 17828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4201254197876773e-06,
+      "loss": 0.8208,
+      "step": 17829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4196426407701514e-06,
+      "loss": 0.8496,
+      "step": 17830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4191599302505056e-06,
+      "loss": 0.8379,
+      "step": 17831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4186772882379712e-06,
+      "loss": 0.8242,
+      "step": 17832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4181947147417852e-06,
+      "loss": 0.8345,
+      "step": 17833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.417712209771177e-06,
+      "loss": 0.8491,
+      "step": 17834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4172297733353813e-06,
+      "loss": 0.8091,
+      "step": 17835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4167474054436253e-06,
+      "loss": 0.4086,
+      "step": 17836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4162651061051364e-06,
+      "loss": 0.8198,
+      "step": 17837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4157828753291431e-06,
+      "loss": 0.8364,
+      "step": 17838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.415300713124873e-06,
+      "loss": 0.8394,
+      "step": 17839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4148186195015478e-06,
+      "loss": 0.814,
+      "step": 17840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4143365944683935e-06,
+      "loss": 0.4237,
+      "step": 17841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4138546380346302e-06,
+      "loss": 0.8208,
+      "step": 17842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4133727502094773e-06,
+      "loss": 0.8066,
+      "step": 17843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.412890931002157e-06,
+      "loss": 0.8457,
+      "step": 17844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4124091804218847e-06,
+      "loss": 0.8735,
+      "step": 17845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4119274984778776e-06,
+      "loss": 0.853,
+      "step": 17846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4114458851793534e-06,
+      "loss": 0.8354,
+      "step": 17847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4109643405355238e-06,
+      "loss": 0.8286,
+      "step": 17848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4104828645556023e-06,
+      "loss": 0.7974,
+      "step": 17849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4100014572487985e-06,
+      "loss": 0.8442,
+      "step": 17850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.409520118624324e-06,
+      "loss": 0.8828,
+      "step": 17851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4090388486913892e-06,
+      "loss": 0.7676,
+      "step": 17852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4085576474591977e-06,
+      "loss": 0.8481,
+      "step": 17853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4080765149369602e-06,
+      "loss": 0.8145,
+      "step": 17854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4075954511338784e-06,
+      "loss": 0.8521,
+      "step": 17855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4071144560591555e-06,
+      "loss": 0.8599,
+      "step": 17856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4066335297219958e-06,
+      "loss": 0.8647,
+      "step": 17857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4061526721315976e-06,
+      "loss": 0.9009,
+      "step": 17858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4056718832971628e-06,
+      "loss": 0.8367,
+      "step": 17859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4051911632278897e-06,
+      "loss": 0.8623,
+      "step": 17860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.404710511932974e-06,
+      "loss": 0.9033,
+      "step": 17861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4042299294216122e-06,
+      "loss": 0.7944,
+      "step": 17862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.403749415702997e-06,
+      "loss": 0.7861,
+      "step": 17863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4032689707863217e-06,
+      "loss": 0.7939,
+      "step": 17864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4027885946807806e-06,
+      "loss": 0.8242,
+      "step": 17865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4023082873955601e-06,
+      "loss": 0.8345,
+      "step": 17866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4018280489398533e-06,
+      "loss": 0.8306,
+      "step": 17867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4013478793228464e-06,
+      "loss": 0.8301,
+      "step": 17868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4008677785537234e-06,
+      "loss": 0.8735,
+      "step": 17869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4003877466416725e-06,
+      "loss": 0.8726,
+      "step": 17870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3999077835958753e-06,
+      "loss": 0.835,
+      "step": 17871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3994278894255148e-06,
+      "loss": 0.8901,
+      "step": 17872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3989480641397746e-06,
+      "loss": 0.8486,
+      "step": 17873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3984683077478322e-06,
+      "loss": 0.8086,
+      "step": 17874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3979886202588643e-06,
+      "loss": 0.7974,
+      "step": 17875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3975090016820524e-06,
+      "loss": 0.7964,
+      "step": 17876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3970294520265687e-06,
+      "loss": 0.7969,
+      "step": 17877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.39654997130159e-06,
+      "loss": 0.8579,
+      "step": 17878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3960705595162876e-06,
+      "loss": 0.3265,
+      "step": 17879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3955912166798364e-06,
+      "loss": 0.7666,
+      "step": 17880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3951119428014042e-06,
+      "loss": 0.8462,
+      "step": 17881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3946327378901603e-06,
+      "loss": 0.8594,
+      "step": 17882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.394153601955275e-06,
+      "loss": 0.7861,
+      "step": 17883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3936745350059116e-06,
+      "loss": 0.8765,
+      "step": 17884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3931955370512369e-06,
+      "loss": 0.834,
+      "step": 17885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3927166081004174e-06,
+      "loss": 0.8242,
+      "step": 17886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.392237748162613e-06,
+      "loss": 0.8105,
+      "step": 17887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3917589572469843e-06,
+      "loss": 0.8335,
+      "step": 17888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3912802353626941e-06,
+      "loss": 0.7803,
+      "step": 17889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.390801582518898e-06,
+      "loss": 0.8457,
+      "step": 17890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3903229987247568e-06,
+      "loss": 0.8047,
+      "step": 17891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3898444839894231e-06,
+      "loss": 0.8452,
+      "step": 17892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3893660383220547e-06,
+      "loss": 0.8301,
+      "step": 17893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3888876617318043e-06,
+      "loss": 0.7219,
+      "step": 17894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.388409354227821e-06,
+      "loss": 0.8101,
+      "step": 17895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3879311158192583e-06,
+      "loss": 0.8882,
+      "step": 17896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.387452946515267e-06,
+      "loss": 0.7764,
+      "step": 17897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3869748463249915e-06,
+      "loss": 0.8032,
+      "step": 17898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3864968152575825e-06,
+      "loss": 0.9048,
+      "step": 17899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3860188533221836e-06,
+      "loss": 0.8813,
+      "step": 17900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3855409605279368e-06,
+      "loss": 0.9092,
+      "step": 17901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3850631368839896e-06,
+      "loss": 0.8223,
+      "step": 17902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3845853823994787e-06,
+      "loss": 0.8652,
+      "step": 17903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3841076970835472e-06,
+      "loss": 0.8154,
+      "step": 17904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3836300809453345e-06,
+      "loss": 0.7686,
+      "step": 17905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.383152533993977e-06,
+      "loss": 0.7866,
+      "step": 17906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3826750562386116e-06,
+      "loss": 0.8887,
+      "step": 17907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3821976476883702e-06,
+      "loss": 0.7661,
+      "step": 17908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3817203083523894e-06,
+      "loss": 0.8564,
+      "step": 17909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3812430382398019e-06,
+      "loss": 0.8213,
+      "step": 17910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380765837359736e-06,
+      "loss": 0.8306,
+      "step": 17911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380288705721325e-06,
+      "loss": 0.8804,
+      "step": 17912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3798116433336944e-06,
+      "loss": 0.8228,
+      "step": 17913
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3793346502059702e-06,
+      "loss": 0.8169,
+      "step": 17914
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3788577263472814e-06,
+      "loss": 0.8301,
+      "step": 17915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.378380871766749e-06,
+      "loss": 0.8716,
+      "step": 17916
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.377904086473497e-06,
+      "loss": 0.8696,
+      "step": 17917
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.37742737047665e-06,
+      "loss": 0.8335,
+      "step": 17918
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3769507237853253e-06,
+      "loss": 0.8179,
+      "step": 17919
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3764741464086423e-06,
+      "loss": 0.7754,
+      "step": 17920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3759976383557178e-06,
+      "loss": 0.8647,
+      "step": 17921
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3755211996356687e-06,
+      "loss": 0.8882,
+      "step": 17922
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3750448302576125e-06,
+      "loss": 0.8467,
+      "step": 17923
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3745685302306588e-06,
+      "loss": 0.8506,
+      "step": 17924
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3740922995639239e-06,
+      "loss": 0.79,
+      "step": 17925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.373616138266517e-06,
+      "loss": 0.9141,
+      "step": 17926
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3731400463475463e-06,
+      "loss": 0.8496,
+      "step": 17927
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3726640238161237e-06,
+      "loss": 0.7896,
+      "step": 17928
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3721880706813523e-06,
+      "loss": 0.8691,
+      "step": 17929
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3717121869523404e-06,
+      "loss": 0.8491,
+      "step": 17930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3712363726381929e-06,
+      "loss": 0.8809,
+      "step": 17931
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3707606277480117e-06,
+      "loss": 0.8618,
+      "step": 17932
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3702849522908972e-06,
+      "loss": 0.832,
+      "step": 17933
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3698093462759526e-06,
+      "loss": 0.9014,
+      "step": 17934
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.369333809712275e-06,
+      "loss": 0.8574,
+      "step": 17935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3688583426089635e-06,
+      "loss": 0.7827,
+      "step": 17936
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3683829449751124e-06,
+      "loss": 0.8604,
+      "step": 17937
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3679076168198186e-06,
+      "loss": 0.8794,
+      "step": 17938
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.367432358152177e-06,
+      "loss": 0.8369,
+      "step": 17939
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3669571689812788e-06,
+      "loss": 0.8325,
+      "step": 17940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3664820493162135e-06,
+      "loss": 0.8159,
+      "step": 17941
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3660069991660735e-06,
+      "loss": 0.8643,
+      "step": 17942
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3655320185399445e-06,
+      "loss": 0.79,
+      "step": 17943
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3650571074469166e-06,
+      "loss": 0.791,
+      "step": 17944
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3645822658960727e-06,
+      "loss": 0.8242,
+      "step": 17945
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3641074938965005e-06,
+      "loss": 0.8135,
+      "step": 17946
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3636327914572817e-06,
+      "loss": 0.812,
+      "step": 17947
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3631581585874959e-06,
+      "loss": 0.8228,
+      "step": 17948
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3626835952962269e-06,
+      "loss": 0.8057,
+      "step": 17949
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.362209101592551e-06,
+      "loss": 0.8584,
+      "step": 17950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3617346774855472e-06,
+      "loss": 0.8169,
+      "step": 17951
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.361260322984294e-06,
+      "loss": 0.8477,
+      "step": 17952
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3607860380978648e-06,
+      "loss": 0.9067,
+      "step": 17953
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3603118228353319e-06,
+      "loss": 0.8511,
+      "step": 17954
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3598376772057708e-06,
+      "loss": 0.8438,
+      "step": 17955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3593636012182493e-06,
+      "loss": 0.8188,
+      "step": 17956
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3588895948818409e-06,
+      "loss": 0.8379,
+      "step": 17957
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3584156582056107e-06,
+      "loss": 0.8047,
+      "step": 17958
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3579417911986286e-06,
+      "loss": 0.7988,
+      "step": 17959
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3574679938699591e-06,
+      "loss": 0.8315,
+      "step": 17960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3569942662286657e-06,
+      "loss": 0.8223,
+      "step": 17961
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3565206082838128e-06,
+      "loss": 0.8027,
+      "step": 17962
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.356047020044463e-06,
+      "loss": 0.8403,
+      "step": 17963
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3555735015196747e-06,
+      "loss": 0.8506,
+      "step": 17964
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3551000527185098e-06,
+      "loss": 0.813,
+      "step": 17965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.354626673650024e-06,
+      "loss": 0.8896,
+      "step": 17966
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3541533643232734e-06,
+      "loss": 0.8452,
+      "step": 17967
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3536801247473157e-06,
+      "loss": 0.833,
+      "step": 17968
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.353206954931201e-06,
+      "loss": 0.7744,
+      "step": 17969
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.352733854883984e-06,
+      "loss": 0.8228,
+      "step": 17970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3522608246147178e-06,
+      "loss": 0.8975,
+      "step": 17971
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3517878641324494e-06,
+      "loss": 0.8555,
+      "step": 17972
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.351314973446229e-06,
+      "loss": 0.8501,
+      "step": 17973
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3508421525651006e-06,
+      "loss": 0.8301,
+      "step": 17974
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3503694014981122e-06,
+      "loss": 0.8828,
+      "step": 17975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3498967202543095e-06,
+      "loss": 0.7798,
+      "step": 17976
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3494241088427334e-06,
+      "loss": 0.814,
+      "step": 17977
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.348951567272428e-06,
+      "loss": 0.8789,
+      "step": 17978
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3484790955524323e-06,
+      "loss": 0.8081,
+      "step": 17979
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.348006693691784e-06,
+      "loss": 0.8174,
+      "step": 17980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3475343616995245e-06,
+      "loss": 0.7832,
+      "step": 17981
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3470620995846861e-06,
+      "loss": 0.8154,
+      "step": 17982
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3465899073563066e-06,
+      "loss": 0.7651,
+      "step": 17983
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3461177850234202e-06,
+      "loss": 0.8506,
+      "step": 17984
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.345645732595059e-06,
+      "loss": 0.8159,
+      "step": 17985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3451737500802531e-06,
+      "loss": 0.8789,
+      "step": 17986
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3447018374880312e-06,
+      "loss": 0.3458,
+      "step": 17987
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3442299948274234e-06,
+      "loss": 0.8672,
+      "step": 17988
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3437582221074574e-06,
+      "loss": 0.7842,
+      "step": 17989
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3432865193371575e-06,
+      "loss": 0.8521,
+      "step": 17990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3428148865255497e-06,
+      "loss": 0.7905,
+      "step": 17991
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3423433236816563e-06,
+      "loss": 0.8149,
+      "step": 17992
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3418718308144973e-06,
+      "loss": 0.8721,
+      "step": 17993
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.341400407933096e-06,
+      "loss": 0.7949,
+      "step": 17994
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3409290550464693e-06,
+      "loss": 0.842,
+      "step": 17995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3404577721636352e-06,
+      "loss": 0.854,
+      "step": 17996
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3399865592936123e-06,
+      "loss": 0.8423,
+      "step": 17997
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3395154164454138e-06,
+      "loss": 0.8623,
+      "step": 17998
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.339044343628052e-06,
+      "loss": 0.853,
+      "step": 17999
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3385733408505425e-06,
+      "loss": 0.8335,
+      "step": 18000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3381024081218935e-06,
+      "loss": 0.8589,
+      "step": 18001
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.337631545451117e-06,
+      "loss": 0.8076,
+      "step": 18002
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.337160752847219e-06,
+      "loss": 0.8481,
+      "step": 18003
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3366900303192087e-06,
+      "loss": 0.897,
+      "step": 18004
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3362193778760912e-06,
+      "loss": 0.7852,
+      "step": 18005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.335748795526869e-06,
+      "loss": 0.8213,
+      "step": 18006
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.335278283280546e-06,
+      "loss": 0.853,
+      "step": 18007
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3348078411461262e-06,
+      "loss": 0.7583,
+      "step": 18008
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3343374691326066e-06,
+      "loss": 0.8926,
+      "step": 18009
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.333867167248989e-06,
+      "loss": 0.8379,
+      "step": 18010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3333969355042696e-06,
+      "loss": 0.4136,
+      "step": 18011
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3329267739074431e-06,
+      "loss": 0.7998,
+      "step": 18012
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3324566824675073e-06,
+      "loss": 0.8457,
+      "step": 18013
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3319866611934533e-06,
+      "loss": 0.8472,
+      "step": 18014
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3315167100942762e-06,
+      "loss": 0.835,
+      "step": 18015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3310468291789636e-06,
+      "loss": 0.8291,
+      "step": 18016
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3305770184565076e-06,
+      "loss": 0.3657,
+      "step": 18017
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3301072779358965e-06,
+      "loss": 0.8442,
+      "step": 18018
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3296376076261141e-06,
+      "loss": 0.3676,
+      "step": 18019
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.329168007536148e-06,
+      "loss": 0.8564,
+      "step": 18020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.328698477674984e-06,
+      "loss": 0.8115,
+      "step": 18021
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3282290180516022e-06,
+      "loss": 0.3341,
+      "step": 18022
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3277596286749867e-06,
+      "loss": 0.7915,
+      "step": 18023
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3272903095541157e-06,
+      "loss": 0.8438,
+      "step": 18024
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3268210606979675e-06,
+      "loss": 0.8472,
+      "step": 18025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3263518821155218e-06,
+      "loss": 0.8267,
+      "step": 18026
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3258827738157526e-06,
+      "loss": 0.7993,
+      "step": 18027
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3254137358076351e-06,
+      "loss": 0.8467,
+      "step": 18028
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3249447681001453e-06,
+      "loss": 0.8105,
+      "step": 18029
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3244758707022531e-06,
+      "loss": 0.8735,
+      "step": 18030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3240070436229292e-06,
+      "loss": 0.7637,
+      "step": 18031
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3235382868711417e-06,
+      "loss": 0.8359,
+      "step": 18032
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3230696004558601e-06,
+      "loss": 0.7979,
+      "step": 18033
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3226009843860527e-06,
+      "loss": 0.7954,
+      "step": 18034
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.322132438670682e-06,
+      "loss": 0.8115,
+      "step": 18035
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3216639633187146e-06,
+      "loss": 0.772,
+      "step": 18036
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3211955583391123e-06,
+      "loss": 0.8706,
+      "step": 18037
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3207272237408342e-06,
+      "loss": 0.8882,
+      "step": 18038
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3202589595328436e-06,
+      "loss": 0.7485,
+      "step": 18039
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.319790765724096e-06,
+      "loss": 0.8604,
+      "step": 18040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3193226423235511e-06,
+      "loss": 0.855,
+      "step": 18041
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.318854589340165e-06,
+      "loss": 0.7617,
+      "step": 18042
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3183866067828914e-06,
+      "loss": 0.8613,
+      "step": 18043
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.317918694660682e-06,
+      "loss": 0.8198,
+      "step": 18044
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.317450852982492e-06,
+      "loss": 0.8691,
+      "step": 18045
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3169830817572682e-06,
+      "loss": 0.8286,
+      "step": 18046
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3165153809939634e-06,
+      "loss": 0.9106,
+      "step": 18047
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3160477507015224e-06,
+      "loss": 0.8208,
+      "step": 18048
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3155801908888932e-06,
+      "loss": 0.8203,
+      "step": 18049
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3151127015650221e-06,
+      "loss": 0.8203,
+      "step": 18050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3146452827388522e-06,
+      "loss": 0.8862,
+      "step": 18051
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.314177934419325e-06,
+      "loss": 0.894,
+      "step": 18052
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3137106566153806e-06,
+      "loss": 0.8779,
+      "step": 18053
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3132434493359603e-06,
+      "loss": 0.8574,
+      "step": 18054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.312776312590004e-06,
+      "loss": 0.8276,
+      "step": 18055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3123092463864456e-06,
+      "loss": 0.8696,
+      "step": 18056
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3118422507342239e-06,
+      "loss": 0.8198,
+      "step": 18057
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3113753256422719e-06,
+      "loss": 0.894,
+      "step": 18058
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3109084711195207e-06,
+      "loss": 0.8364,
+      "step": 18059
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3104416871749054e-06,
+      "loss": 0.8599,
+      "step": 18060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3099749738173528e-06,
+      "loss": 0.7871,
+      "step": 18061
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3095083310557942e-06,
+      "loss": 0.8311,
+      "step": 18062
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3090417588991578e-06,
+      "loss": 0.8374,
+      "step": 18063
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3085752573563692e-06,
+      "loss": 0.9033,
+      "step": 18064
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3081088264363512e-06,
+      "loss": 0.8262,
+      "step": 18065
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3076424661480303e-06,
+      "loss": 0.7722,
+      "step": 18066
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3071761765003266e-06,
+      "loss": 0.8486,
+      "step": 18067
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3067099575021624e-06,
+      "loss": 0.7905,
+      "step": 18068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3062438091624557e-06,
+      "loss": 0.854,
+      "step": 18069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.305777731490127e-06,
+      "loss": 0.854,
+      "step": 18070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3053117244940917e-06,
+      "loss": 0.8594,
+      "step": 18071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.304845788183264e-06,
+      "loss": 0.8374,
+      "step": 18072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3043799225665587e-06,
+      "loss": 0.8682,
+      "step": 18073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3039141276528905e-06,
+      "loss": 0.8271,
+      "step": 18074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3034484034511679e-06,
+      "loss": 0.7832,
+      "step": 18075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.302982749970304e-06,
+      "loss": 0.8325,
+      "step": 18076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3025171672192061e-06,
+      "loss": 0.8242,
+      "step": 18077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3020516552067797e-06,
+      "loss": 0.8418,
+      "step": 18078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.301586213941934e-06,
+      "loss": 0.8511,
+      "step": 18079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3011208434335704e-06,
+      "loss": 0.8389,
+      "step": 18080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3006555436905932e-06,
+      "loss": 0.8286,
+      "step": 18081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.300190314721907e-06,
+      "loss": 0.9077,
+      "step": 18082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2997251565364099e-06,
+      "loss": 0.8823,
+      "step": 18083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.299260069143002e-06,
+      "loss": 0.8057,
+      "step": 18084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2987950525505788e-06,
+      "loss": 0.8208,
+      "step": 18085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2983301067680387e-06,
+      "loss": 0.8818,
+      "step": 18086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2978652318042778e-06,
+      "loss": 0.8311,
+      "step": 18087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2974004276681879e-06,
+      "loss": 0.8477,
+      "step": 18088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.296935694368664e-06,
+      "loss": 0.8198,
+      "step": 18089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.296471031914595e-06,
+      "loss": 0.8433,
+      "step": 18090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2960064403148703e-06,
+      "loss": 0.8511,
+      "step": 18091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2955419195783802e-06,
+      "loss": 0.8491,
+      "step": 18092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2950774697140094e-06,
+      "loss": 0.8145,
+      "step": 18093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2946130907306448e-06,
+      "loss": 0.3595,
+      "step": 18094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2941487826371718e-06,
+      "loss": 0.7886,
+      "step": 18095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2936845454424728e-06,
+      "loss": 0.8379,
+      "step": 18096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2932203791554287e-06,
+      "loss": 0.8354,
+      "step": 18097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2927562837849188e-06,
+      "loss": 0.8145,
+      "step": 18098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2922922593398223e-06,
+      "loss": 0.8887,
+      "step": 18099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2918283058290199e-06,
+      "loss": 0.8403,
+      "step": 18100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2913644232613832e-06,
+      "loss": 0.2831,
+      "step": 18101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2909006116457913e-06,
+      "loss": 0.7939,
+      "step": 18102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2904368709911146e-06,
+      "loss": 0.8291,
+      "step": 18103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2899732013062249e-06,
+      "loss": 0.8872,
+      "step": 18104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.289509602599996e-06,
+      "loss": 0.8267,
+      "step": 18105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2890460748812939e-06,
+      "loss": 0.8403,
+      "step": 18106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2885826181589877e-06,
+      "loss": 0.8013,
+      "step": 18107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2881192324419467e-06,
+      "loss": 0.856,
+      "step": 18108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2876559177390335e-06,
+      "loss": 0.8086,
+      "step": 18109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2871926740591107e-06,
+      "loss": 0.812,
+      "step": 18110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2867295014110448e-06,
+      "loss": 0.835,
+      "step": 18111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2862663998036929e-06,
+      "loss": 0.8564,
+      "step": 18112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2858033692459187e-06,
+      "loss": 0.8311,
+      "step": 18113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2853404097465777e-06,
+      "loss": 0.8569,
+      "step": 18114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2848775213145292e-06,
+      "loss": 0.8628,
+      "step": 18115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.284414703958628e-06,
+      "loss": 0.8413,
+      "step": 18116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2839519576877274e-06,
+      "loss": 0.7991,
+      "step": 18117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.283489282510683e-06,
+      "loss": 0.8018,
+      "step": 18118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2830266784363438e-06,
+      "loss": 0.8145,
+      "step": 18119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.282564145473561e-06,
+      "loss": 0.8364,
+      "step": 18120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2821016836311855e-06,
+      "loss": 0.8169,
+      "step": 18121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2816392929180642e-06,
+      "loss": 0.8516,
+      "step": 18122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2811769733430406e-06,
+      "loss": 0.7983,
+      "step": 18123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2807147249149632e-06,
+      "loss": 0.7681,
+      "step": 18124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2802525476426725e-06,
+      "loss": 0.8481,
+      "step": 18125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2797904415350137e-06,
+      "loss": 0.8799,
+      "step": 18126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2793284066008243e-06,
+      "loss": 0.8237,
+      "step": 18127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2788664428489479e-06,
+      "loss": 0.8354,
+      "step": 18128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2784045502882192e-06,
+      "loss": 0.853,
+      "step": 18129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2779427289274749e-06,
+      "loss": 0.9004,
+      "step": 18130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2774809787755515e-06,
+      "loss": 0.8276,
+      "step": 18131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2770192998412844e-06,
+      "loss": 0.8564,
+      "step": 18132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2765576921335032e-06,
+      "loss": 0.8354,
+      "step": 18133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2760961556610424e-06,
+      "loss": 0.8252,
+      "step": 18134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2756346904327298e-06,
+      "loss": 0.9292,
+      "step": 18135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2751732964573931e-06,
+      "loss": 0.8428,
+      "step": 18136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2747119737438628e-06,
+      "loss": 0.8457,
+      "step": 18137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.274250722300961e-06,
+      "loss": 0.8672,
+      "step": 18138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.273789542137514e-06,
+      "loss": 0.7612,
+      "step": 18139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2733284332623463e-06,
+      "loss": 0.8589,
+      "step": 18140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272867395684278e-06,
+      "loss": 0.8633,
+      "step": 18141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2724064294121296e-06,
+      "loss": 0.7959,
+      "step": 18142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2719455344547193e-06,
+      "loss": 0.8394,
+      "step": 18143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.271484710820865e-06,
+      "loss": 0.8374,
+      "step": 18144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.271023958519385e-06,
+      "loss": 0.8164,
+      "step": 18145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.270563277559092e-06,
+      "loss": 0.8203,
+      "step": 18146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2701026679488016e-06,
+      "loss": 0.8628,
+      "step": 18147
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2696421296973245e-06,
+      "loss": 0.8149,
+      "step": 18148
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2691816628134701e-06,
+      "loss": 0.7979,
+      "step": 18149
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2687212673060512e-06,
+      "loss": 0.8789,
+      "step": 18150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2682609431838727e-06,
+      "loss": 0.7871,
+      "step": 18151
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2678006904557432e-06,
+      "loss": 0.8687,
+      "step": 18152
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2673405091304691e-06,
+      "loss": 0.4048,
+      "step": 18153
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2668803992168528e-06,
+      "loss": 0.8403,
+      "step": 18154
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.266420360723697e-06,
+      "loss": 0.8403,
+      "step": 18155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2659603936598019e-06,
+      "loss": 0.8047,
+      "step": 18156
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2655004980339687e-06,
+      "loss": 0.8638,
+      "step": 18157
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.265040673854997e-06,
+      "loss": 0.9023,
+      "step": 18158
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.264580921131681e-06,
+      "loss": 0.8354,
+      "step": 18159
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2641212398728187e-06,
+      "loss": 0.8447,
+      "step": 18160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2636616300872057e-06,
+      "loss": 0.8188,
+      "step": 18161
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.263202091783633e-06,
+      "loss": 0.8618,
+      "step": 18162
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.262742624970893e-06,
+      "loss": 0.8823,
+      "step": 18163
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2622832296577747e-06,
+      "loss": 0.8779,
+      "step": 18164
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2618239058530674e-06,
+      "loss": 0.8369,
+      "step": 18165
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.261364653565561e-06,
+      "loss": 0.8442,
+      "step": 18166
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2609054728040393e-06,
+      "loss": 0.7773,
+      "step": 18167
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260446363577289e-06,
+      "loss": 0.8896,
+      "step": 18168
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2599873258940925e-06,
+      "loss": 0.7876,
+      "step": 18169
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2595283597632307e-06,
+      "loss": 0.856,
+      "step": 18170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2590694651934871e-06,
+      "loss": 0.8032,
+      "step": 18171
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2586106421936378e-06,
+      "loss": 0.8042,
+      "step": 18172
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2581518907724628e-06,
+      "loss": 0.9082,
+      "step": 18173
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2576932109387401e-06,
+      "loss": 0.8291,
+      "step": 18174
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2572346027012433e-06,
+      "loss": 0.8027,
+      "step": 18175
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.256776066068745e-06,
+      "loss": 0.8423,
+      "step": 18176
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2563176010500206e-06,
+      "loss": 0.8394,
+      "step": 18177
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2558592076538383e-06,
+      "loss": 0.7983,
+      "step": 18178
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.255400885888971e-06,
+      "loss": 0.8833,
+      "step": 18179
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2549426357641842e-06,
+      "loss": 0.8013,
+      "step": 18180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2544844572882475e-06,
+      "loss": 0.8779,
+      "step": 18181
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2540263504699253e-06,
+      "loss": 0.8506,
+      "step": 18182
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2535683153179807e-06,
+      "loss": 0.8242,
+      "step": 18183
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2531103518411792e-06,
+      "loss": 0.8438,
+      "step": 18184
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2526524600482792e-06,
+      "loss": 0.8345,
+      "step": 18185
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2521946399480429e-06,
+      "loss": 0.3307,
+      "step": 18186
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2517368915492307e-06,
+      "loss": 0.8667,
+      "step": 18187
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.251279214860598e-06,
+      "loss": 0.8599,
+      "step": 18188
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2508216098908998e-06,
+      "loss": 0.8491,
+      "step": 18189
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.250364076648894e-06,
+      "loss": 0.8276,
+      "step": 18190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2499066151433303e-06,
+      "loss": 0.8545,
+      "step": 18191
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2494492253829637e-06,
+      "loss": 0.8643,
+      "step": 18192
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2489919073765421e-06,
+      "loss": 0.8179,
+      "step": 18193
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.248534661132818e-06,
+      "loss": 0.8042,
+      "step": 18194
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2480774866605373e-06,
+      "loss": 0.3259,
+      "step": 18195
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2476203839684452e-06,
+      "loss": 0.811,
+      "step": 18196
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2471633530652883e-06,
+      "loss": 0.8662,
+      "step": 18197
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2467063939598112e-06,
+      "loss": 0.8848,
+      "step": 18198
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.246249506660754e-06,
+      "loss": 0.8145,
+      "step": 18199
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2457926911768603e-06,
+      "loss": 0.7969,
+      "step": 18200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2453359475168686e-06,
+      "loss": 0.769,
+      "step": 18201
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2448792756895151e-06,
+      "loss": 0.8149,
+      "step": 18202
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.24442267570354e-06,
+      "loss": 0.873,
+      "step": 18203
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2439661475676751e-06,
+      "loss": 0.8755,
+      "step": 18204
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2435096912906574e-06,
+      "loss": 0.834,
+      "step": 18205
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.24305330688122e-06,
+      "loss": 0.9155,
+      "step": 18206
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2425969943480925e-06,
+      "loss": 0.8555,
+      "step": 18207
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2421407537000063e-06,
+      "loss": 0.8179,
+      "step": 18208
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2416845849456872e-06,
+      "loss": 0.8765,
+      "step": 18209
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2412284880938647e-06,
+      "loss": 0.8052,
+      "step": 18210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2407724631532653e-06,
+      "loss": 0.8667,
+      "step": 18211
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2403165101326108e-06,
+      "loss": 0.9004,
+      "step": 18212
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2398606290406284e-06,
+      "loss": 0.8467,
+      "step": 18213
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2394048198860364e-06,
+      "loss": 0.8438,
+      "step": 18214
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2389490826775557e-06,
+      "loss": 0.9014,
+      "step": 18215
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2384934174239066e-06,
+      "loss": 0.3213,
+      "step": 18216
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2380378241338043e-06,
+      "loss": 0.8276,
+      "step": 18217
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2375823028159667e-06,
+      "loss": 0.8677,
+      "step": 18218
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2371268534791099e-06,
+      "loss": 0.8198,
+      "step": 18219
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.236671476131946e-06,
+      "loss": 0.8657,
+      "step": 18220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2362161707831871e-06,
+      "loss": 0.896,
+      "step": 18221
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.235760937441542e-06,
+      "loss": 0.8496,
+      "step": 18222
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2353057761157223e-06,
+      "loss": 0.8184,
+      "step": 18223
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2348506868144367e-06,
+      "loss": 0.8604,
+      "step": 18224
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2343956695463892e-06,
+      "loss": 0.8687,
+      "step": 18225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2339407243202882e-06,
+      "loss": 0.8862,
+      "step": 18226
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2334858511448356e-06,
+      "loss": 0.8525,
+      "step": 18227
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2330310500287319e-06,
+      "loss": 0.8232,
+      "step": 18228
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2325763209806824e-06,
+      "loss": 0.897,
+      "step": 18229
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2321216640093824e-06,
+      "loss": 0.3381,
+      "step": 18230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.231667079123533e-06,
+      "loss": 0.8096,
+      "step": 18231
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.231212566331832e-06,
+      "loss": 0.7959,
+      "step": 18232
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2307581256429736e-06,
+      "loss": 0.875,
+      "step": 18233
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.23030375706565e-06,
+      "loss": 0.8901,
+      "step": 18234
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.229849460608557e-06,
+      "loss": 0.8599,
+      "step": 18235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2293952362803841e-06,
+      "loss": 0.8037,
+      "step": 18236
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2289410840898231e-06,
+      "loss": 0.8506,
+      "step": 18237
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2284870040455605e-06,
+      "loss": 0.8325,
+      "step": 18238
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2280329961562858e-06,
+      "loss": 0.8335,
+      "step": 18239
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2275790604306836e-06,
+      "loss": 0.8203,
+      "step": 18240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2271251968774372e-06,
+      "loss": 0.8516,
+      "step": 18241
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2266714055052314e-06,
+      "loss": 0.321,
+      "step": 18242
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2262176863227492e-06,
+      "loss": 0.8228,
+      "step": 18243
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2257640393386678e-06,
+      "loss": 0.835,
+      "step": 18244
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2253104645616686e-06,
+      "loss": 0.8193,
+      "step": 18245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.224856962000429e-06,
+      "loss": 0.7827,
+      "step": 18246
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2244035316636233e-06,
+      "loss": 0.8501,
+      "step": 18247
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.223950173559929e-06,
+      "loss": 0.8306,
+      "step": 18248
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2234968876980164e-06,
+      "loss": 0.8418,
+      "step": 18249
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2230436740865608e-06,
+      "loss": 0.8633,
+      "step": 18250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2225905327342307e-06,
+      "loss": 0.8955,
+      "step": 18251
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.222137463649697e-06,
+      "loss": 0.8369,
+      "step": 18252
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2216844668416266e-06,
+      "loss": 0.8516,
+      "step": 18253
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2212315423186855e-06,
+      "loss": 0.303,
+      "step": 18254
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.220778690089539e-06,
+      "loss": 0.8569,
+      "step": 18255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2203259101628535e-06,
+      "loss": 0.8311,
+      "step": 18256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2198732025472876e-06,
+      "loss": 0.8071,
+      "step": 18257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2194205672515054e-06,
+      "loss": 0.8706,
+      "step": 18258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.218968004284165e-06,
+      "loss": 0.7969,
+      "step": 18259
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.218515513653924e-06,
+      "loss": 0.8403,
+      "step": 18260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2180630953694411e-06,
+      "loss": 0.8408,
+      "step": 18261
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.217610749439369e-06,
+      "loss": 0.833,
+      "step": 18262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2171584758723643e-06,
+      "loss": 0.7905,
+      "step": 18263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2167062746770803e-06,
+      "loss": 0.8809,
+      "step": 18264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2162541458621668e-06,
+      "loss": 0.8809,
+      "step": 18265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2158020894362738e-06,
+      "loss": 0.8374,
+      "step": 18266
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2153501054080486e-06,
+      "loss": 0.8604,
+      "step": 18267
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2148981937861399e-06,
+      "loss": 0.8555,
+      "step": 18268
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2144463545791952e-06,
+      "loss": 0.3815,
+      "step": 18269
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2139945877958547e-06,
+      "loss": 0.8345,
+      "step": 18270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.213542893444764e-06,
+      "loss": 0.8433,
+      "step": 18271
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2130912715345662e-06,
+      "loss": 0.8643,
+      "step": 18272
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2126397220739e-06,
+      "loss": 0.915,
+      "step": 18273
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2121882450714034e-06,
+      "loss": 0.8037,
+      "step": 18274
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2117368405357132e-06,
+      "loss": 0.8311,
+      "step": 18275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.211285508475467e-06,
+      "loss": 0.8911,
+      "step": 18276
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2108342488993009e-06,
+      "loss": 0.8569,
+      "step": 18277
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2103830618158451e-06,
+      "loss": 0.7988,
+      "step": 18278
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2099319472337345e-06,
+      "loss": 0.8623,
+      "step": 18279
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.209480905161598e-06,
+      "loss": 0.7817,
+      "step": 18280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2090299356080632e-06,
+      "loss": 0.2947,
+      "step": 18281
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2085790385817614e-06,
+      "loss": 0.834,
+      "step": 18282
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2081282140913159e-06,
+      "loss": 0.8848,
+      "step": 18283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2076774621453524e-06,
+      "loss": 0.8179,
+      "step": 18284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2072267827524965e-06,
+      "loss": 0.7783,
+      "step": 18285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.206776175921369e-06,
+      "loss": 0.8018,
+      "step": 18286
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2063256416605906e-06,
+      "loss": 0.8584,
+      "step": 18287
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2058751799787794e-06,
+      "loss": 0.7925,
+      "step": 18288
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2054247908845545e-06,
+      "loss": 0.7949,
+      "step": 18289
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2049744743865349e-06,
+      "loss": 0.8765,
+      "step": 18290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.204524230493332e-06,
+      "loss": 0.8794,
+      "step": 18291
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2040740592135631e-06,
+      "loss": 0.7939,
+      "step": 18292
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2036239605558392e-06,
+      "loss": 0.8306,
+      "step": 18293
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2031739345287697e-06,
+      "loss": 0.8042,
+      "step": 18294
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2027239811409668e-06,
+      "loss": 0.8428,
+      "step": 18295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2022741004010368e-06,
+      "loss": 0.8359,
+      "step": 18296
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2018242923175883e-06,
+      "loss": 0.8916,
+      "step": 18297
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.201374556899227e-06,
+      "loss": 0.8296,
+      "step": 18298
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2009248941545565e-06,
+      "loss": 0.832,
+      "step": 18299
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.200475304092178e-06,
+      "loss": 0.8213,
+      "step": 18300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.200025786720696e-06,
+      "loss": 0.7793,
+      "step": 18301
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1995763420487066e-06,
+      "loss": 0.8857,
+      "step": 18302
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1991269700848124e-06,
+      "loss": 0.772,
+      "step": 18303
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1986776708376075e-06,
+      "loss": 0.7998,
+      "step": 18304
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1982284443156899e-06,
+      "loss": 0.8525,
+      "step": 18305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.197779290527653e-06,
+      "loss": 0.8496,
+      "step": 18306
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1973302094820881e-06,
+      "loss": 0.8447,
+      "step": 18307
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.196881201187589e-06,
+      "loss": 0.8906,
+      "step": 18308
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1964322656527472e-06,
+      "loss": 0.8403,
+      "step": 18309
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1959834028861479e-06,
+      "loss": 0.8394,
+      "step": 18310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1955346128963824e-06,
+      "loss": 0.8462,
+      "step": 18311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1950858956920342e-06,
+      "loss": 0.8198,
+      "step": 18312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1946372512816879e-06,
+      "loss": 0.8643,
+      "step": 18313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1941886796739283e-06,
+      "loss": 0.8613,
+      "step": 18314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1937401808773353e-06,
+      "loss": 0.8369,
+      "step": 18315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1932917549004903e-06,
+      "loss": 0.8428,
+      "step": 18316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1928434017519747e-06,
+      "loss": 0.8252,
+      "step": 18317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1923951214403633e-06,
+      "loss": 0.7959,
+      "step": 18318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.191946913974234e-06,
+      "loss": 0.7876,
+      "step": 18319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1914987793621592e-06,
+      "loss": 0.8262,
+      "step": 18320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1910507176127146e-06,
+      "loss": 0.8232,
+      "step": 18321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1906027287344729e-06,
+      "loss": 0.8335,
+      "step": 18322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1901548127360024e-06,
+      "loss": 0.7925,
+      "step": 18323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1897069696258756e-06,
+      "loss": 0.8286,
+      "step": 18324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1892591994126585e-06,
+      "loss": 0.8374,
+      "step": 18325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1888115021049168e-06,
+      "loss": 0.8076,
+      "step": 18326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1883638777112183e-06,
+      "loss": 0.8594,
+      "step": 18327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1879163262401233e-06,
+      "loss": 0.7998,
+      "step": 18328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.187468847700196e-06,
+      "loss": 0.8706,
+      "step": 18329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1870214420999993e-06,
+      "loss": 0.811,
+      "step": 18330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.186574109448091e-06,
+      "loss": 0.8604,
+      "step": 18331
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1861268497530286e-06,
+      "loss": 0.7822,
+      "step": 18332
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1856796630233685e-06,
+      "loss": 0.8428,
+      "step": 18333
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.185232549267667e-06,
+      "loss": 0.7988,
+      "step": 18334
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1847855084944799e-06,
+      "loss": 0.8003,
+      "step": 18335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1843385407123558e-06,
+      "loss": 0.8037,
+      "step": 18336
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1838916459298504e-06,
+      "loss": 0.8755,
+      "step": 18337
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1834448241555107e-06,
+      "loss": 0.8623,
+      "step": 18338
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1829980753978849e-06,
+      "loss": 0.8135,
+      "step": 18339
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1825513996655213e-06,
+      "loss": 0.8491,
+      "step": 18340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.182104796966964e-06,
+      "loss": 0.8149,
+      "step": 18341
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1816582673107586e-06,
+      "loss": 0.8062,
+      "step": 18342
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1812118107054483e-06,
+      "loss": 0.8457,
+      "step": 18343
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1807654271595743e-06,
+      "loss": 0.8208,
+      "step": 18344
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1803191166816742e-06,
+      "loss": 0.772,
+      "step": 18345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1798728792802899e-06,
+      "loss": 0.8838,
+      "step": 18346
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1794267149639555e-06,
+      "loss": 0.8501,
+      "step": 18347
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1789806237412106e-06,
+      "loss": 0.8579,
+      "step": 18348
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1785346056205853e-06,
+      "loss": 0.8438,
+      "step": 18349
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1780886606106167e-06,
+      "loss": 0.8198,
+      "step": 18350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.177642788719835e-06,
+      "loss": 0.8389,
+      "step": 18351
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.177196989956768e-06,
+      "loss": 0.8242,
+      "step": 18352
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.176751264329949e-06,
+      "loss": 0.8071,
+      "step": 18353
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1763056118479005e-06,
+      "loss": 0.8599,
+      "step": 18354
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1758600325191517e-06,
+      "loss": 0.7646,
+      "step": 18355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1754145263522276e-06,
+      "loss": 0.8892,
+      "step": 18356
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.17496909335565e-06,
+      "loss": 0.8345,
+      "step": 18357
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1745237335379406e-06,
+      "loss": 0.8667,
+      "step": 18358
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1740784469076217e-06,
+      "loss": 0.8638,
+      "step": 18359
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1736332334732092e-06,
+      "loss": 0.8257,
+      "step": 18360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1731880932432238e-06,
+      "loss": 0.8271,
+      "step": 18361
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1727430262261796e-06,
+      "loss": 0.8267,
+      "step": 18362
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1722980324305933e-06,
+      "loss": 0.8521,
+      "step": 18363
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1718531118649778e-06,
+      "loss": 0.8228,
+      "step": 18364
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1714082645378427e-06,
+      "loss": 0.8789,
+      "step": 18365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.170963490457701e-06,
+      "loss": 0.8613,
+      "step": 18366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1705187896330628e-06,
+      "loss": 0.814,
+      "step": 18367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1700741620724337e-06,
+      "loss": 0.8403,
+      "step": 18368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1696296077843222e-06,
+      "loss": 0.8506,
+      "step": 18369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.169185126777232e-06,
+      "loss": 0.8242,
+      "step": 18370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1687407190596656e-06,
+      "loss": 0.8662,
+      "step": 18371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.168296384640128e-06,
+      "loss": 0.8496,
+      "step": 18372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1678521235271172e-06,
+      "loss": 0.7754,
+      "step": 18373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.167407935729134e-06,
+      "loss": 0.8872,
+      "step": 18374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.166963821254678e-06,
+      "loss": 0.7949,
+      "step": 18375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.166519780112244e-06,
+      "loss": 0.8296,
+      "step": 18376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1660758123103271e-06,
+      "loss": 0.8179,
+      "step": 18377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1656319178574205e-06,
+      "loss": 0.8379,
+      "step": 18378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1651880967620176e-06,
+      "loss": 0.8491,
+      "step": 18379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1647443490326104e-06,
+      "loss": 0.8286,
+      "step": 18380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1643006746776864e-06,
+      "loss": 0.7793,
+      "step": 18381
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.163857073705734e-06,
+      "loss": 0.855,
+      "step": 18382
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1634135461252444e-06,
+      "loss": 0.834,
+      "step": 18383
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1629700919446962e-06,
+      "loss": 0.7473,
+      "step": 18384
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.162526711172578e-06,
+      "loss": 0.8569,
+      "step": 18385
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1620834038173695e-06,
+      "loss": 0.8218,
+      "step": 18386
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1616401698875535e-06,
+      "loss": 0.8936,
+      "step": 18387
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.161197009391611e-06,
+      "loss": 0.8462,
+      "step": 18388
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.160753922338017e-06,
+      "loss": 0.8271,
+      "step": 18389
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.160310908735252e-06,
+      "loss": 0.8389,
+      "step": 18390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1598679685917901e-06,
+      "loss": 0.8086,
+      "step": 18391
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.159425101916104e-06,
+      "loss": 0.8345,
+      "step": 18392
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1589823087166685e-06,
+      "loss": 0.8462,
+      "step": 18393
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1585395890019536e-06,
+      "loss": 0.8447,
+      "step": 18394
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.158096942780429e-06,
+      "loss": 0.8359,
+      "step": 18395
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1576543700605652e-06,
+      "loss": 0.2983,
+      "step": 18396
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.157211870850829e-06,
+      "loss": 0.8325,
+      "step": 18397
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1567694451596845e-06,
+      "loss": 0.8237,
+      "step": 18398
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1563270929955956e-06,
+      "loss": 0.8379,
+      "step": 18399
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.155884814367026e-06,
+      "loss": 0.8662,
+      "step": 18400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.155442609282439e-06,
+      "loss": 0.7832,
+      "step": 18401
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1550004777502915e-06,
+      "loss": 0.8306,
+      "step": 18402
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.154558419779046e-06,
+      "loss": 0.8164,
+      "step": 18403
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1541164353771568e-06,
+      "loss": 0.8105,
+      "step": 18404
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1536745245530795e-06,
+      "loss": 0.8208,
+      "step": 18405
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.153232687315271e-06,
+      "loss": 0.8516,
+      "step": 18406
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1527909236721812e-06,
+      "loss": 0.8154,
+      "step": 18407
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1523492336322639e-06,
+      "loss": 0.8784,
+      "step": 18408
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1519076172039701e-06,
+      "loss": 0.8252,
+      "step": 18409
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1514660743957478e-06,
+      "loss": 0.8091,
+      "step": 18410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1510246052160424e-06,
+      "loss": 0.8955,
+      "step": 18411
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1505832096733027e-06,
+      "loss": 0.8599,
+      "step": 18412
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1501418877759707e-06,
+      "loss": 0.8271,
+      "step": 18413
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1497006395324928e-06,
+      "loss": 0.8599,
+      "step": 18414
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1492594649513072e-06,
+      "loss": 0.8618,
+      "step": 18415
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.148818364040858e-06,
+      "loss": 0.8369,
+      "step": 18416
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1483773368095812e-06,
+      "loss": 0.8271,
+      "step": 18417
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1479363832659152e-06,
+      "loss": 0.8374,
+      "step": 18418
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1474955034182965e-06,
+      "loss": 0.8423,
+      "step": 18419
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.147054697275159e-06,
+      "loss": 0.7593,
+      "step": 18420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1466139648449365e-06,
+      "loss": 0.8145,
+      "step": 18421
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1461733061360626e-06,
+      "loss": 0.8105,
+      "step": 18422
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1457327211569657e-06,
+      "loss": 0.8247,
+      "step": 18423
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.145292209916074e-06,
+      "loss": 0.8574,
+      "step": 18424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1448517724218184e-06,
+      "loss": 0.896,
+      "step": 18425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1444114086826218e-06,
+      "loss": 0.9082,
+      "step": 18426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.143971118706912e-06,
+      "loss": 0.79,
+      "step": 18427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1435309025031098e-06,
+      "loss": 0.8438,
+      "step": 18428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1430907600796398e-06,
+      "loss": 0.8516,
+      "step": 18429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1426506914449215e-06,
+      "loss": 0.8418,
+      "step": 18430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.142210696607372e-06,
+      "loss": 0.8345,
+      "step": 18431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1417707755754109e-06,
+      "loss": 0.8535,
+      "step": 18432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.141330928357457e-06,
+      "loss": 0.8125,
+      "step": 18433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1408911549619205e-06,
+      "loss": 0.874,
+      "step": 18434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1404514553972196e-06,
+      "loss": 0.8867,
+      "step": 18435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1400118296717637e-06,
+      "loss": 0.8325,
+      "step": 18436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1395722777939627e-06,
+      "loss": 0.8018,
+      "step": 18437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.139132799772229e-06,
+      "loss": 0.8101,
+      "step": 18438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1386933956149676e-06,
+      "loss": 0.7954,
+      "step": 18439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1382540653305856e-06,
+      "loss": 0.8154,
+      "step": 18440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1378148089274904e-06,
+      "loss": 0.8262,
+      "step": 18441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.137375626414084e-06,
+      "loss": 0.833,
+      "step": 18442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1369365177987685e-06,
+      "loss": 0.8438,
+      "step": 18443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1364974830899438e-06,
+      "loss": 0.8271,
+      "step": 18444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1360585222960098e-06,
+      "loss": 0.8555,
+      "step": 18445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1356196354253668e-06,
+      "loss": 0.8491,
+      "step": 18446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.135180822486408e-06,
+      "loss": 0.8252,
+      "step": 18447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1347420834875323e-06,
+      "loss": 0.853,
+      "step": 18448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.134303418437131e-06,
+      "loss": 0.8101,
+      "step": 18449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1338648273435954e-06,
+      "loss": 0.8652,
+      "step": 18450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1334263102153193e-06,
+      "loss": 0.8608,
+      "step": 18451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1329878670606898e-06,
+      "loss": 0.918,
+      "step": 18452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1325494978880957e-06,
+      "loss": 0.8813,
+      "step": 18453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1321112027059256e-06,
+      "loss": 0.8188,
+      "step": 18454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1316729815225635e-06,
+      "loss": 0.8867,
+      "step": 18455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1312348343463924e-06,
+      "loss": 0.8467,
+      "step": 18456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1307967611857946e-06,
+      "loss": 0.8618,
+      "step": 18457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1303587620491513e-06,
+      "loss": 0.8032,
+      "step": 18458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1299208369448445e-06,
+      "loss": 0.8169,
+      "step": 18459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.129482985881249e-06,
+      "loss": 0.8638,
+      "step": 18460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1290452088667442e-06,
+      "loss": 0.7827,
+      "step": 18461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1286075059097052e-06,
+      "loss": 0.8105,
+      "step": 18462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.128169877018503e-06,
+      "loss": 0.8643,
+      "step": 18463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1277323222015142e-06,
+      "loss": 0.8711,
+      "step": 18464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1272948414671065e-06,
+      "loss": 0.792,
+      "step": 18465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1268574348236505e-06,
+      "loss": 0.8228,
+      "step": 18466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1264201022795173e-06,
+      "loss": 0.8101,
+      "step": 18467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1259828438430714e-06,
+      "loss": 0.8608,
+      "step": 18468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1255456595226765e-06,
+      "loss": 0.7896,
+      "step": 18469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1251085493267006e-06,
+      "loss": 0.7773,
+      "step": 18470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1246715132635021e-06,
+      "loss": 0.8989,
+      "step": 18471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1242345513414464e-06,
+      "loss": 0.8389,
+      "step": 18472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1237976635688896e-06,
+      "loss": 0.8149,
+      "step": 18473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1233608499541927e-06,
+      "loss": 0.8037,
+      "step": 18474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.122924110505712e-06,
+      "loss": 0.3795,
+      "step": 18475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1224874452318007e-06,
+      "loss": 0.7881,
+      "step": 18476
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1220508541408148e-06,
+      "loss": 0.8784,
+      "step": 18477
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1216143372411088e-06,
+      "loss": 0.8374,
+      "step": 18478
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1211778945410307e-06,
+      "loss": 0.7979,
+      "step": 18479
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1207415260489329e-06,
+      "loss": 0.8291,
+      "step": 18480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1203052317731623e-06,
+      "loss": 0.8057,
+      "step": 18481
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1198690117220657e-06,
+      "loss": 0.8433,
+      "step": 18482
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.11943286590399e-06,
+      "loss": 0.8047,
+      "step": 18483
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1189967943272772e-06,
+      "loss": 0.8623,
+      "step": 18484
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1185607970002731e-06,
+      "loss": 0.8721,
+      "step": 18485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1181248739313155e-06,
+      "loss": 0.7803,
+      "step": 18486
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1176890251287482e-06,
+      "loss": 0.9131,
+      "step": 18487
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.117253250600907e-06,
+      "loss": 0.8481,
+      "step": 18488
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1168175503561284e-06,
+      "loss": 0.8198,
+      "step": 18489
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1163819244027496e-06,
+      "loss": 0.7705,
+      "step": 18490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1159463727491054e-06,
+      "loss": 0.8286,
+      "step": 18491
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.115510895403526e-06,
+      "loss": 0.8911,
+      "step": 18492
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1150754923743462e-06,
+      "loss": 0.8164,
+      "step": 18493
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.114640163669894e-06,
+      "loss": 0.8579,
+      "step": 18494
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1142049092984963e-06,
+      "loss": 0.8091,
+      "step": 18495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1137697292684834e-06,
+      "loss": 0.854,
+      "step": 18496
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1133346235881787e-06,
+      "loss": 0.7871,
+      "step": 18497
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1128995922659065e-06,
+      "loss": 0.8325,
+      "step": 18498
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1124646353099921e-06,
+      "loss": 0.8433,
+      "step": 18499
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1120297527287537e-06,
+      "loss": 0.8569,
+      "step": 18500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1115949445305147e-06,
+      "loss": 0.8965,
+      "step": 18501
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1111602107235914e-06,
+      "loss": 0.8052,
+      "step": 18502
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1107255513163002e-06,
+      "loss": 0.8755,
+      "step": 18503
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1102909663169591e-06,
+      "loss": 0.7734,
+      "step": 18504
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1098564557338798e-06,
+      "loss": 0.9185,
+      "step": 18505
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1094220195753768e-06,
+      "loss": 0.8242,
+      "step": 18506
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1089876578497626e-06,
+      "loss": 0.8403,
+      "step": 18507
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1085533705653461e-06,
+      "loss": 0.8394,
+      "step": 18508
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1081191577304356e-06,
+      "loss": 0.8716,
+      "step": 18509
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1076850193533372e-06,
+      "loss": 0.8354,
+      "step": 18510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1072509554423577e-06,
+      "loss": 0.7905,
+      "step": 18511
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.106816966005803e-06,
+      "loss": 0.8525,
+      "step": 18512
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1063830510519735e-06,
+      "loss": 0.8486,
+      "step": 18513
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1059492105891733e-06,
+      "loss": 0.8584,
+      "step": 18514
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1055154446257004e-06,
+      "loss": 0.8867,
+      "step": 18515
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1050817531698532e-06,
+      "loss": 0.874,
+      "step": 18516
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.104648136229931e-06,
+      "loss": 0.4299,
+      "step": 18517
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1042145938142268e-06,
+      "loss": 0.7969,
+      "step": 18518
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1037811259310361e-06,
+      "loss": 0.8628,
+      "step": 18519
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1033477325886538e-06,
+      "loss": 0.9331,
+      "step": 18520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1029144137953696e-06,
+      "loss": 0.8896,
+      "step": 18521
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.102481169559474e-06,
+      "loss": 0.8247,
+      "step": 18522
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1020479998892537e-06,
+      "loss": 0.8809,
+      "step": 18523
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.101614904792998e-06,
+      "loss": 0.854,
+      "step": 18524
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1011818842789928e-06,
+      "loss": 0.8242,
+      "step": 18525
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1007489383555215e-06,
+      "loss": 0.8164,
+      "step": 18526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1003160670308682e-06,
+      "loss": 0.8857,
+      "step": 18527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0998832703133138e-06,
+      "loss": 0.896,
+      "step": 18528
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0994505482111374e-06,
+      "loss": 0.8677,
+      "step": 18529
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.099017900732619e-06,
+      "loss": 0.7915,
+      "step": 18530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0985853278860341e-06,
+      "loss": 0.7778,
+      "step": 18531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0981528296796607e-06,
+      "loss": 0.9199,
+      "step": 18532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0977204061217728e-06,
+      "loss": 0.8164,
+      "step": 18533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0972880572206428e-06,
+      "loss": 0.8135,
+      "step": 18534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.096855782984541e-06,
+      "loss": 0.7539,
+      "step": 18535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0964235834217395e-06,
+      "loss": 0.854,
+      "step": 18536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0959914585405058e-06,
+      "loss": 0.8799,
+      "step": 18537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0955594083491079e-06,
+      "loss": 0.8569,
+      "step": 18538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0951274328558098e-06,
+      "loss": 0.8765,
+      "step": 18539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.094695532068879e-06,
+      "loss": 0.7622,
+      "step": 18540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0942637059965766e-06,
+      "loss": 0.8467,
+      "step": 18541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0938319546471627e-06,
+      "loss": 0.7935,
+      "step": 18542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0934002780288988e-06,
+      "loss": 0.8608,
+      "step": 18543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0929686761500452e-06,
+      "loss": 0.8555,
+      "step": 18544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0925371490188563e-06,
+      "loss": 0.8418,
+      "step": 18545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.09210569664359e-06,
+      "loss": 0.8242,
+      "step": 18546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0916743190325002e-06,
+      "loss": 0.7788,
+      "step": 18547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0912430161938382e-06,
+      "loss": 0.8438,
+      "step": 18548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0908117881358577e-06,
+      "loss": 0.8789,
+      "step": 18549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0903806348668067e-06,
+      "loss": 0.8286,
+      "step": 18550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0899495563949347e-06,
+      "loss": 0.8081,
+      "step": 18551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0895185527284908e-06,
+      "loss": 0.8149,
+      "step": 18552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0890876238757188e-06,
+      "loss": 0.8013,
+      "step": 18553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0886567698448635e-06,
+      "loss": 0.8433,
+      "step": 18554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0882259906441667e-06,
+      "loss": 0.8525,
+      "step": 18555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0877952862818708e-06,
+      "loss": 0.8491,
+      "step": 18556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0873646567662165e-06,
+      "loss": 0.8242,
+      "step": 18557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0869341021054414e-06,
+      "loss": 0.8413,
+      "step": 18558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0865036223077841e-06,
+      "loss": 0.7979,
+      "step": 18559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0860732173814797e-06,
+      "loss": 0.8208,
+      "step": 18560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0856428873347608e-06,
+      "loss": 0.7856,
+      "step": 18561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0852126321758632e-06,
+      "loss": 0.7993,
+      "step": 18562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.084782451913015e-06,
+      "loss": 0.8364,
+      "step": 18563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0843523465544486e-06,
+      "loss": 0.8335,
+      "step": 18564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0839223161083933e-06,
+      "loss": 0.8442,
+      "step": 18565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.083492360583075e-06,
+      "loss": 0.8975,
+      "step": 18566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0830624799867195e-06,
+      "loss": 0.8083,
+      "step": 18567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0826326743275501e-06,
+      "loss": 0.8481,
+      "step": 18568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.08220294361379e-06,
+      "loss": 0.8203,
+      "step": 18569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.081773287853663e-06,
+      "loss": 0.8545,
+      "step": 18570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0813437070553862e-06,
+      "loss": 0.7622,
+      "step": 18571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.08091420122718e-06,
+      "loss": 0.8496,
+      "step": 18572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0804847703772614e-06,
+      "loss": 0.9097,
+      "step": 18573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0800554145138442e-06,
+      "loss": 0.7979,
+      "step": 18574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0796261336451452e-06,
+      "loss": 0.8247,
+      "step": 18575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0791969277793747e-06,
+      "loss": 0.7334,
+      "step": 18576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.078767796924745e-06,
+      "loss": 0.8843,
+      "step": 18577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.078338741089468e-06,
+      "loss": 0.873,
+      "step": 18578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0779097602817512e-06,
+      "loss": 0.855,
+      "step": 18579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0774808545097991e-06,
+      "loss": 0.877,
+      "step": 18580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.077052023781821e-06,
+      "loss": 0.4069,
+      "step": 18581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.076623268106018e-06,
+      "loss": 0.8823,
+      "step": 18582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.076194587490596e-06,
+      "loss": 0.8774,
+      "step": 18583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0757659819437532e-06,
+      "loss": 0.8789,
+      "step": 18584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0753374514736925e-06,
+      "loss": 0.9009,
+      "step": 18585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0749089960886105e-06,
+      "loss": 0.8042,
+      "step": 18586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.074480615796703e-06,
+      "loss": 0.8525,
+      "step": 18587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.074052310606169e-06,
+      "loss": 0.7993,
+      "step": 18588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0736240805251997e-06,
+      "loss": 0.8809,
+      "step": 18589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0731959255619885e-06,
+      "loss": 0.8521,
+      "step": 18590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0727678457247287e-06,
+      "loss": 0.8232,
+      "step": 18591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0723398410216085e-06,
+      "loss": 0.8579,
+      "step": 18592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0719119114608145e-06,
+      "loss": 0.8057,
+      "step": 18593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0714840570505375e-06,
+      "loss": 0.8213,
+      "step": 18594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0710562777989597e-06,
+      "loss": 0.834,
+      "step": 18595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0706285737142675e-06,
+      "loss": 0.7983,
+      "step": 18596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0702009448046412e-06,
+      "loss": 0.8335,
+      "step": 18597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0697733910782655e-06,
+      "loss": 0.8257,
+      "step": 18598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0693459125433176e-06,
+      "loss": 0.8135,
+      "step": 18599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0689185092079751e-06,
+      "loss": 0.752,
+      "step": 18600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0684911810804161e-06,
+      "loss": 0.7754,
+      "step": 18601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.068063928168817e-06,
+      "loss": 0.8472,
+      "step": 18602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.06763675048135e-06,
+      "loss": 0.8677,
+      "step": 18603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.06720964802619e-06,
+      "loss": 0.8447,
+      "step": 18604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0667826208115066e-06,
+      "loss": 0.8062,
+      "step": 18605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0663556688454685e-06,
+      "loss": 0.8174,
+      "step": 18606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0659287921362466e-06,
+      "loss": 0.8608,
+      "step": 18607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0655019906920043e-06,
+      "loss": 0.7346,
+      "step": 18608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0650752645209094e-06,
+      "loss": 0.752,
+      "step": 18609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0646486136311264e-06,
+      "loss": 0.8511,
+      "step": 18610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0642220380308155e-06,
+      "loss": 0.9458,
+      "step": 18611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0637955377281405e-06,
+      "loss": 0.8477,
+      "step": 18612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0633691127312595e-06,
+      "loss": 0.8853,
+      "step": 18613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0629427630483296e-06,
+      "loss": 0.876,
+      "step": 18614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0625164886875094e-06,
+      "loss": 0.8594,
+      "step": 18615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.062090289656953e-06,
+      "loss": 0.8291,
+      "step": 18616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0616641659648142e-06,
+      "loss": 0.9106,
+      "step": 18617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0612381176192471e-06,
+      "loss": 0.8247,
+      "step": 18618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.060812144628402e-06,
+      "loss": 0.856,
+      "step": 18619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0603862470004274e-06,
+      "loss": 0.8291,
+      "step": 18620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.059960424743471e-06,
+      "loss": 0.8325,
+      "step": 18621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.05953467786568e-06,
+      "loss": 0.8701,
+      "step": 18622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.059109006375202e-06,
+      "loss": 0.9131,
+      "step": 18623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0586834102801768e-06,
+      "loss": 0.8193,
+      "step": 18624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0582578895887507e-06,
+      "loss": 0.8232,
+      "step": 18625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.057832444309062e-06,
+      "loss": 0.8315,
+      "step": 18626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0574070744492498e-06,
+      "loss": 0.8228,
+      "step": 18627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0569817800174542e-06,
+      "loss": 0.8286,
+      "step": 18628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0565565610218092e-06,
+      "loss": 0.792,
+      "step": 18629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0561314174704513e-06,
+      "loss": 0.77,
+      "step": 18630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0557063493715158e-06,
+      "loss": 0.8457,
+      "step": 18631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0552813567331326e-06,
+      "loss": 0.856,
+      "step": 18632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0548564395634337e-06,
+      "loss": 0.8467,
+      "step": 18633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0544315978705466e-06,
+      "loss": 0.856,
+      "step": 18634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0540068316626e-06,
+      "loss": 0.8467,
+      "step": 18635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0535821409477227e-06,
+      "loss": 0.8022,
+      "step": 18636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.053157525734036e-06,
+      "loss": 0.9102,
+      "step": 18637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.052732986029667e-06,
+      "loss": 0.8364,
+      "step": 18638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.052308521842736e-06,
+      "loss": 0.8535,
+      "step": 18639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0518841331813622e-06,
+      "loss": 0.8369,
+      "step": 18640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.051459820053668e-06,
+      "loss": 0.8652,
+      "step": 18641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0510355824677677e-06,
+      "loss": 0.8765,
+      "step": 18642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0506114204317802e-06,
+      "loss": 0.7886,
+      "step": 18643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0501873339538205e-06,
+      "loss": 0.8579,
+      "step": 18644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0497633230420013e-06,
+      "loss": 0.8223,
+      "step": 18645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0493393877044334e-06,
+      "loss": 0.8286,
+      "step": 18646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0489155279492297e-06,
+      "loss": 0.8506,
+      "step": 18647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0484917437844972e-06,
+      "loss": 0.7852,
+      "step": 18648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0480680352183454e-06,
+      "loss": 0.8613,
+      "step": 18649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0476444022588783e-06,
+      "loss": 0.833,
+      "step": 18650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.047220844914203e-06,
+      "loss": 0.8794,
+      "step": 18651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0467973631924217e-06,
+      "loss": 0.8296,
+      "step": 18652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0463739571016352e-06,
+      "loss": 0.8564,
+      "step": 18653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0459506266499464e-06,
+      "loss": 0.8545,
+      "step": 18654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0455273718454518e-06,
+      "loss": 0.8486,
+      "step": 18655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0451041926962496e-06,
+      "loss": 0.8579,
+      "step": 18656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0446810892104382e-06,
+      "loss": 0.8232,
+      "step": 18657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.04425806139611e-06,
+      "loss": 0.8564,
+      "step": 18658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.043835109261357e-06,
+      "loss": 0.8086,
+      "step": 18659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0434122328142743e-06,
+      "loss": 0.8247,
+      "step": 18660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0429894320629485e-06,
+      "loss": 0.8125,
+      "step": 18661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0425667070154721e-06,
+      "loss": 0.812,
+      "step": 18662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0421440576799296e-06,
+      "loss": 0.8193,
+      "step": 18663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0417214840644091e-06,
+      "loss": 0.8579,
+      "step": 18664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.041298986176994e-06,
+      "loss": 0.8193,
+      "step": 18665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0408765640257663e-06,
+      "loss": 0.791,
+      "step": 18666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0404542176188087e-06,
+      "loss": 0.7905,
+      "step": 18667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0400319469642029e-06,
+      "loss": 0.8291,
+      "step": 18668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0396097520700244e-06,
+      "loss": 0.8857,
+      "step": 18669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0391876329443534e-06,
+      "loss": 0.8271,
+      "step": 18670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0387655895952648e-06,
+      "loss": 0.8481,
+      "step": 18671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0383436220308313e-06,
+      "loss": 0.8555,
+      "step": 18672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0379217302591282e-06,
+      "loss": 0.8101,
+      "step": 18673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0374999142882247e-06,
+      "loss": 0.8716,
+      "step": 18674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0370781741261914e-06,
+      "loss": 0.8311,
+      "step": 18675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0366565097810993e-06,
+      "loss": 0.8555,
+      "step": 18676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.036234921261014e-06,
+      "loss": 0.8525,
+      "step": 18677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0358134085740002e-06,
+      "loss": 0.7935,
+      "step": 18678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353919717281212e-06,
+      "loss": 0.8535,
+      "step": 18679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0349706107314417e-06,
+      "loss": 0.8496,
+      "step": 18680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345493255920236e-06,
+      "loss": 0.8809,
+      "step": 18681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0341281163179246e-06,
+      "loss": 0.8369,
+      "step": 18682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0337069829172048e-06,
+      "loss": 0.8188,
+      "step": 18683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033285925397921e-06,
+      "loss": 0.8472,
+      "step": 18684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032864943768127e-06,
+      "loss": 0.8433,
+      "step": 18685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0324440380358792e-06,
+      "loss": 0.8315,
+      "step": 18686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320232082092279e-06,
+      "loss": 0.7979,
+      "step": 18687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.031602454296225e-06,
+      "loss": 0.7622,
+      "step": 18688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0311817763049225e-06,
+      "loss": 0.8086,
+      "step": 18689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.030761174243366e-06,
+      "loss": 0.8203,
+      "step": 18690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0303406481196038e-06,
+      "loss": 0.7715,
+      "step": 18691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0299201979416785e-06,
+      "loss": 0.7881,
+      "step": 18692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0294998237176357e-06,
+      "loss": 0.7944,
+      "step": 18693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0290795254555197e-06,
+      "loss": 0.8623,
+      "step": 18694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0286593031633686e-06,
+      "loss": 0.8413,
+      "step": 18695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.028239156849224e-06,
+      "loss": 0.835,
+      "step": 18696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.027819086521123e-06,
+      "loss": 0.7695,
+      "step": 18697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0273990921871013e-06,
+      "loss": 0.8301,
+      "step": 18698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.026979173855196e-06,
+      "loss": 0.8335,
+      "step": 18699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0265593315334387e-06,
+      "loss": 0.752,
+      "step": 18700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0261395652298628e-06,
+      "loss": 0.8418,
+      "step": 18701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0257198749525005e-06,
+      "loss": 0.8496,
+      "step": 18702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0253002607093793e-06,
+      "loss": 0.8848,
+      "step": 18703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.024880722508526e-06,
+      "loss": 0.3657,
+      "step": 18704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0244612603579707e-06,
+      "loss": 0.7847,
+      "step": 18705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0240418742657338e-06,
+      "loss": 0.8013,
+      "step": 18706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0236225642398428e-06,
+      "loss": 0.8862,
+      "step": 18707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0232033302883171e-06,
+      "loss": 0.835,
+      "step": 18708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0227841724191794e-06,
+      "loss": 0.7842,
+      "step": 18709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0223650906404481e-06,
+      "loss": 0.8384,
+      "step": 18710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0219460849601392e-06,
+      "loss": 0.8296,
+      "step": 18711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0215271553862705e-06,
+      "loss": 0.8872,
+      "step": 18712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0211083019268575e-06,
+      "loss": 0.8359,
+      "step": 18713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0206895245899112e-06,
+      "loss": 0.894,
+      "step": 18714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0202708233834469e-06,
+      "loss": 0.8013,
+      "step": 18715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0198521983154725e-06,
+      "loss": 0.8496,
+      "step": 18716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.019433649393996e-06,
+      "loss": 0.832,
+      "step": 18717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0190151766270279e-06,
+      "loss": 0.7988,
+      "step": 18718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0185967800225716e-06,
+      "loss": 0.8462,
+      "step": 18719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.018178459588634e-06,
+      "loss": 0.7783,
+      "step": 18720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0177602153332156e-06,
+      "loss": 0.8906,
+      "step": 18721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0173420472643197e-06,
+      "loss": 0.8481,
+      "step": 18722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0169239553899478e-06,
+      "loss": 0.8589,
+      "step": 18723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0165059397180971e-06,
+      "loss": 0.8613,
+      "step": 18724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.016088000256763e-06,
+      "loss": 0.8345,
+      "step": 18725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0156701370139454e-06,
+      "loss": 0.854,
+      "step": 18726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0152523499976353e-06,
+      "loss": 0.832,
+      "step": 18727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0148346392158276e-06,
+      "loss": 0.8779,
+      "step": 18728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0144170046765123e-06,
+      "loss": 0.8594,
+      "step": 18729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0139994463876813e-06,
+      "loss": 0.854,
+      "step": 18730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0135819643573225e-06,
+      "loss": 0.8501,
+      "step": 18731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0131645585934208e-06,
+      "loss": 0.915,
+      "step": 18732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0127472291039635e-06,
+      "loss": 0.8369,
+      "step": 18733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0123299758969363e-06,
+      "loss": 0.8296,
+      "step": 18734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.011912798980319e-06,
+      "loss": 0.8477,
+      "step": 18735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0114956983620954e-06,
+      "loss": 0.8472,
+      "step": 18736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0110786740502448e-06,
+      "loss": 0.8706,
+      "step": 18737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0106617260527435e-06,
+      "loss": 0.8379,
+      "step": 18738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0102448543775706e-06,
+      "loss": 0.8179,
+      "step": 18739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0098280590326998e-06,
+      "loss": 0.8145,
+      "step": 18740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0094113400261058e-06,
+      "loss": 0.8784,
+      "step": 18741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0089946973657627e-06,
+      "loss": 0.8945,
+      "step": 18742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.00857813105964e-06,
+      "loss": 0.8867,
+      "step": 18743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0081616411157075e-06,
+      "loss": 0.8838,
+      "step": 18744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0077452275419314e-06,
+      "loss": 0.8774,
+      "step": 18745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0073288903462809e-06,
+      "loss": 0.8813,
+      "step": 18746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006912629536721e-06,
+      "loss": 0.811,
+      "step": 18747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006496445121214e-06,
+      "loss": 0.7651,
+      "step": 18748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0060803371077239e-06,
+      "loss": 0.7983,
+      "step": 18749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0056643055042103e-06,
+      "loss": 0.8926,
+      "step": 18750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0052483503186322e-06,
+      "loss": 0.8042,
+      "step": 18751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.004832471558949e-06,
+      "loss": 0.9346,
+      "step": 18752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0044166692331154e-06,
+      "loss": 0.8545,
+      "step": 18753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0040009433490866e-06,
+      "loss": 0.8291,
+      "step": 18754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0035852939148184e-06,
+      "loss": 0.8237,
+      "step": 18755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0031697209382608e-06,
+      "loss": 0.8364,
+      "step": 18756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0027542244273647e-06,
+      "loss": 0.8281,
+      "step": 18757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0023388043900778e-06,
+      "loss": 0.8325,
+      "step": 18758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0019234608343491e-06,
+      "loss": 0.7241,
+      "step": 18759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0015081937681266e-06,
+      "loss": 0.8574,
+      "step": 18760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0010930031993515e-06,
+      "loss": 0.7808,
+      "step": 18761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0006778891359697e-06,
+      "loss": 0.7988,
+      "step": 18762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0002628515859226e-06,
+      "loss": 0.8027,
+      "step": 18763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.998478905571484e-07,
+      "loss": 0.8716,
+      "step": 18764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.994330060575891e-07,
+      "loss": 0.813,
+      "step": 18765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.990181980951786e-07,
+      "loss": 0.8022,
+      "step": 18766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.98603466677855e-07,
+      "loss": 0.8242,
+      "step": 18767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.981888118135542e-07,
+      "loss": 0.8633,
+      "step": 18768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.97774233510207e-07,
+      "loss": 0.8218,
+      "step": 18769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.973597317757444e-07,
+      "loss": 0.8657,
+      "step": 18770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.969453066180983e-07,
+      "loss": 0.8135,
+      "step": 18771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.96530958045196e-07,
+      "loss": 0.8228,
+      "step": 18772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.961166860649662e-07,
+      "loss": 0.813,
+      "step": 18773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.957024906853318e-07,
+      "loss": 0.8657,
+      "step": 18774
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.952883719142203e-07,
+      "loss": 0.9048,
+      "step": 18775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.94874329759553e-07,
+      "loss": 0.8623,
+      "step": 18776
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.944603642292499e-07,
+      "loss": 0.8955,
+      "step": 18777
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.940464753312319e-07,
+      "loss": 0.8149,
+      "step": 18778
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.936326630734184e-07,
+      "loss": 0.8149,
+      "step": 18779
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.932189274637243e-07,
+      "loss": 0.8521,
+      "step": 18780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.92805268510067e-07,
+      "loss": 0.7705,
+      "step": 18781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.923916862203598e-07,
+      "loss": 0.7925,
+      "step": 18782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.919781806025136e-07,
+      "loss": 0.8125,
+      "step": 18783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.915647516644418e-07,
+      "loss": 0.8198,
+      "step": 18784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.911513994140514e-07,
+      "loss": 0.8462,
+      "step": 18785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.907381238592523e-07,
+      "loss": 0.811,
+      "step": 18786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903249250079517e-07,
+      "loss": 0.8481,
+      "step": 18787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.899118028680538e-07,
+      "loss": 0.813,
+      "step": 18788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.894987574474617e-07,
+      "loss": 0.8184,
+      "step": 18789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.890857887540777e-07,
+      "loss": 0.3787,
+      "step": 18790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.886728967958026e-07,
+      "loss": 0.8208,
+      "step": 18791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.882600815805372e-07,
+      "loss": 0.7949,
+      "step": 18792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.878473431161767e-07,
+      "loss": 0.8481,
+      "step": 18793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.874346814106206e-07,
+      "loss": 0.8564,
+      "step": 18794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.87022096471762e-07,
+      "loss": 0.8057,
+      "step": 18795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.866095883074933e-07,
+      "loss": 0.8438,
+      "step": 18796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.861971569257084e-07,
+      "loss": 0.8467,
+      "step": 18797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.857848023342958e-07,
+      "loss": 0.8574,
+      "step": 18798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.85372524541145e-07,
+      "loss": 0.8379,
+      "step": 18799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.84960323554146e-07,
+      "loss": 0.8545,
+      "step": 18800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.84548199381183e-07,
+      "loss": 0.853,
+      "step": 18801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.8413615203014e-07,
+      "loss": 0.8721,
+      "step": 18802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.837241815088995e-07,
+      "loss": 0.7822,
+      "step": 18803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.833122878253442e-07,
+      "loss": 0.853,
+      "step": 18804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.829004709873558e-07,
+      "loss": 0.8608,
+      "step": 18805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.824887310028103e-07,
+      "loss": 0.833,
+      "step": 18806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.820770678795872e-07,
+      "loss": 0.8335,
+      "step": 18807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.816654816255617e-07,
+      "loss": 0.832,
+      "step": 18808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.812539722486058e-07,
+      "loss": 0.8237,
+      "step": 18809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.808425397565963e-07,
+      "loss": 0.8291,
+      "step": 18810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.804311841574005e-07,
+      "loss": 0.8613,
+      "step": 18811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.800199054588899e-07,
+      "loss": 0.8223,
+      "step": 18812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.79608703668935e-07,
+      "loss": 0.8296,
+      "step": 18813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.791975787954012e-07,
+      "loss": 0.8418,
+      "step": 18814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.78786530846152e-07,
+      "loss": 0.8584,
+      "step": 18815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.78375559829054e-07,
+      "loss": 0.8423,
+      "step": 18816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.779646657519681e-07,
+      "loss": 0.8608,
+      "step": 18817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.775538486227575e-07,
+      "loss": 0.8711,
+      "step": 18818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.771431084492788e-07,
+      "loss": 0.8062,
+      "step": 18819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.767324452393929e-07,
+      "loss": 0.3494,
+      "step": 18820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.763218590009555e-07,
+      "loss": 0.8394,
+      "step": 18821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.759113497418198e-07,
+      "loss": 0.8418,
+      "step": 18822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.755009174698427e-07,
+      "loss": 0.8096,
+      "step": 18823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.750905621928735e-07,
+      "loss": 0.8149,
+      "step": 18824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.74680283918764e-07,
+      "loss": 0.7856,
+      "step": 18825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.742700826553657e-07,
+      "loss": 0.8872,
+      "step": 18826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.73859958410524e-07,
+      "loss": 0.835,
+      "step": 18827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.734499111920848e-07,
+      "loss": 0.834,
+      "step": 18828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.73039941007895e-07,
+      "loss": 0.8115,
+      "step": 18829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.726300478657952e-07,
+      "loss": 0.8359,
+      "step": 18830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.722202317736308e-07,
+      "loss": 0.9365,
+      "step": 18831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.718104927392386e-07,
+      "loss": 0.8955,
+      "step": 18832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.714008307704599e-07,
+      "loss": 0.3909,
+      "step": 18833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.70991245875132e-07,
+      "loss": 0.8735,
+      "step": 18834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.70581738061091e-07,
+      "loss": 0.8369,
+      "step": 18835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.701723073361697e-07,
+      "loss": 0.7729,
+      "step": 18836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.697629537082038e-07,
+      "loss": 0.9292,
+      "step": 18837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.693536771850214e-07,
+      "loss": 0.7974,
+      "step": 18838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.689444777744567e-07,
+      "loss": 0.8232,
+      "step": 18839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.685353554843347e-07,
+      "loss": 0.8516,
+      "step": 18840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.681263103224858e-07,
+      "loss": 0.8535,
+      "step": 18841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.677173422967335e-07,
+      "loss": 0.8599,
+      "step": 18842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.673084514149016e-07,
+      "loss": 0.8604,
+      "step": 18843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.66899637684814e-07,
+      "loss": 0.8486,
+      "step": 18844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.66490901114292e-07,
+      "loss": 0.8311,
+      "step": 18845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.66082241711155e-07,
+      "loss": 0.8555,
+      "step": 18846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.656736594832217e-07,
+      "loss": 0.8115,
+      "step": 18847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.65265154438309e-07,
+      "loss": 0.873,
+      "step": 18848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.648567265842306e-07,
+      "loss": 0.8296,
+      "step": 18849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.64448375928803e-07,
+      "loss": 0.7974,
+      "step": 18850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.640401024798357e-07,
+      "loss": 0.877,
+      "step": 18851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63631906245141e-07,
+      "loss": 0.834,
+      "step": 18852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.632237872325302e-07,
+      "loss": 0.8462,
+      "step": 18853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.628157454498094e-07,
+      "loss": 0.8481,
+      "step": 18854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.624077809047849e-07,
+      "loss": 0.3208,
+      "step": 18855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61999893605261e-07,
+      "loss": 0.8467,
+      "step": 18856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61592083559042e-07,
+      "loss": 0.7852,
+      "step": 18857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.611843507739317e-07,
+      "loss": 0.8076,
+      "step": 18858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.607766952577274e-07,
+      "loss": 0.9155,
+      "step": 18859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.603691170182316e-07,
+      "loss": 0.273,
+      "step": 18860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.599616160632408e-07,
+      "loss": 0.7935,
+      "step": 18861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.595541924005485e-07,
+      "loss": 0.7578,
+      "step": 18862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.591468460379527e-07,
+      "loss": 0.854,
+      "step": 18863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.587395769832448e-07,
+      "loss": 0.8374,
+      "step": 18864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.583323852442162e-07,
+      "loss": 0.7083,
+      "step": 18865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.579252708286597e-07,
+      "loss": 0.8237,
+      "step": 18866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.575182337443623e-07,
+      "loss": 0.8115,
+      "step": 18867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.571112739991112e-07,
+      "loss": 0.8999,
+      "step": 18868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.567043916006908e-07,
+      "loss": 0.8848,
+      "step": 18869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.562975865568869e-07,
+      "loss": 0.812,
+      "step": 18870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.558908588754834e-07,
+      "loss": 0.8286,
+      "step": 18871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.554842085642596e-07,
+      "loss": 0.7939,
+      "step": 18872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.55077635630997e-07,
+      "loss": 0.8462,
+      "step": 18873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.546711400834735e-07,
+      "loss": 0.7856,
+      "step": 18874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.542647219294649e-07,
+      "loss": 0.8677,
+      "step": 18875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.538583811767482e-07,
+      "loss": 0.7939,
+      "step": 18876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.534521178330952e-07,
+      "loss": 0.7583,
+      "step": 18877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.530459319062801e-07,
+      "loss": 0.8809,
+      "step": 18878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.526398234040746e-07,
+      "loss": 0.8574,
+      "step": 18879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.522337923342473e-07,
+      "loss": 0.8892,
+      "step": 18880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.518278387045643e-07,
+      "loss": 0.811,
+      "step": 18881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.514219625227961e-07,
+      "loss": 0.7773,
+      "step": 18882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.510161637967036e-07,
+      "loss": 0.8887,
+      "step": 18883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.506104425340535e-07,
+      "loss": 0.8164,
+      "step": 18884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.502047987426055e-07,
+      "loss": 0.8315,
+      "step": 18885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.497992324301226e-07,
+      "loss": 0.8486,
+      "step": 18886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.493937436043627e-07,
+      "loss": 0.8237,
+      "step": 18887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.489883322730825e-07,
+      "loss": 0.8281,
+      "step": 18888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.485829984440404e-07,
+      "loss": 0.8262,
+      "step": 18889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.481777421249877e-07,
+      "loss": 0.8652,
+      "step": 18890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.477725633236806e-07,
+      "loss": 0.8306,
+      "step": 18891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.473674620478706e-07,
+      "loss": 0.8613,
+      "step": 18892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.469624383053072e-07,
+      "loss": 0.8101,
+      "step": 18893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.465574921037385e-07,
+      "loss": 0.8525,
+      "step": 18894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.461526234509132e-07,
+      "loss": 0.8301,
+      "step": 18895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.457478323545749e-07,
+      "loss": 0.8105,
+      "step": 18896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.453431188224705e-07,
+      "loss": 0.8936,
+      "step": 18897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.449384828623409e-07,
+      "loss": 0.8521,
+      "step": 18898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.445339244819291e-07,
+      "loss": 0.8677,
+      "step": 18899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.441294436889736e-07,
+      "loss": 0.814,
+      "step": 18900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.437250404912118e-07,
+      "loss": 0.8452,
+      "step": 18901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.433207148963825e-07,
+      "loss": 0.8306,
+      "step": 18902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.429164669122209e-07,
+      "loss": 0.8428,
+      "step": 18903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.4251229654646e-07,
+      "loss": 0.8091,
+      "step": 18904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.421082038068336e-07,
+      "loss": 0.8623,
+      "step": 18905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.417041887010713e-07,
+      "loss": 0.8647,
+      "step": 18906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.413002512369024e-07,
+      "loss": 0.8286,
+      "step": 18907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.408963914220565e-07,
+      "loss": 0.8149,
+      "step": 18908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.404926092642575e-07,
+      "loss": 0.8477,
+      "step": 18909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.40088904771232e-07,
+      "loss": 0.8423,
+      "step": 18910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.396852779507043e-07,
+      "loss": 0.8501,
+      "step": 18911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.392817288103955e-07,
+      "loss": 0.8398,
+      "step": 18912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.388782573580269e-07,
+      "loss": 0.8071,
+      "step": 18913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.384748636013147e-07,
+      "loss": 0.3774,
+      "step": 18914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.380715475479791e-07,
+      "loss": 0.8442,
+      "step": 18915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.376683092057365e-07,
+      "loss": 0.4158,
+      "step": 18916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.372651485822993e-07,
+      "loss": 0.731,
+      "step": 18917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.368620656853833e-07,
+      "loss": 0.8574,
+      "step": 18918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.364590605226986e-07,
+      "loss": 0.8442,
+      "step": 18919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.360561331019541e-07,
+      "loss": 0.8008,
+      "step": 18920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.356532834308612e-07,
+      "loss": 0.8677,
+      "step": 18921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.352505115171245e-07,
+      "loss": 0.8276,
+      "step": 18922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.348478173684505e-07,
+      "loss": 0.79,
+      "step": 18923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.344452009925448e-07,
+      "loss": 0.8218,
+      "step": 18924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.340426623971089e-07,
+      "loss": 0.7959,
+      "step": 18925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.336402015898444e-07,
+      "loss": 0.8823,
+      "step": 18926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.332378185784491e-07,
+      "loss": 0.873,
+      "step": 18927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.328355133706229e-07,
+      "loss": 0.7876,
+      "step": 18928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.324332859740631e-07,
+      "loss": 0.8169,
+      "step": 18929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.320311363964635e-07,
+      "loss": 0.8706,
+      "step": 18930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.316290646455201e-07,
+      "loss": 0.8789,
+      "step": 18931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.312270707289228e-07,
+      "loss": 0.8113,
+      "step": 18932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.308251546543628e-07,
+      "loss": 0.8237,
+      "step": 18933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.304233164295306e-07,
+      "loss": 0.812,
+      "step": 18934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.300215560621123e-07,
+      "loss": 0.8257,
+      "step": 18935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.296198735597945e-07,
+      "loss": 0.3987,
+      "step": 18936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.292182689302648e-07,
+      "loss": 0.8833,
+      "step": 18937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.288167421812039e-07,
+      "loss": 0.8379,
+      "step": 18938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.284152933202922e-07,
+      "loss": 0.8247,
+      "step": 18939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.280139223552142e-07,
+      "loss": 0.897,
+      "step": 18940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.276126292936449e-07,
+      "loss": 0.8311,
+      "step": 18941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.272114141432647e-07,
+      "loss": 0.77,
+      "step": 18942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.268102769117465e-07,
+      "loss": 0.812,
+      "step": 18943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.264092176067662e-07,
+      "loss": 0.8354,
+      "step": 18944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.260082362359996e-07,
+      "loss": 0.9033,
+      "step": 18945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.256073328071125e-07,
+      "loss": 0.8096,
+      "step": 18946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.252065073277778e-07,
+      "loss": 0.8633,
+      "step": 18947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.248057598056648e-07,
+      "loss": 0.8701,
+      "step": 18948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.244050902484381e-07,
+      "loss": 0.8521,
+      "step": 18949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.240044986637653e-07,
+      "loss": 0.7915,
+      "step": 18950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.236039850593087e-07,
+      "loss": 0.8286,
+      "step": 18951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.232035494427328e-07,
+      "loss": 0.916,
+      "step": 18952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.228031918216967e-07,
+      "loss": 0.8701,
+      "step": 18953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.224029122038592e-07,
+      "loss": 0.7686,
+      "step": 18954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.220027105968809e-07,
+      "loss": 0.7646,
+      "step": 18955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.216025870084155e-07,
+      "loss": 0.793,
+      "step": 18956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.212025414461195e-07,
+      "loss": 0.8267,
+      "step": 18957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.208025739176474e-07,
+      "loss": 0.8286,
+      "step": 18958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.204026844306502e-07,
+      "loss": 0.8486,
+      "step": 18959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.200028729927774e-07,
+      "loss": 0.8643,
+      "step": 18960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.196031396116795e-07,
+      "loss": 0.8579,
+      "step": 18961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.192034842950026e-07,
+      "loss": 0.7969,
+      "step": 18962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.188039070503945e-07,
+      "loss": 0.8354,
+      "step": 18963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.184044078854982e-07,
+      "loss": 0.8345,
+      "step": 18964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.180049868079583e-07,
+      "loss": 0.8091,
+      "step": 18965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.176056438254149e-07,
+      "loss": 0.4221,
+      "step": 18966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.172063789455077e-07,
+      "loss": 0.8682,
+      "step": 18967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.168071921758764e-07,
+      "loss": 0.8154,
+      "step": 18968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.16408083524159e-07,
+      "loss": 0.8633,
+      "step": 18969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.160090529979887e-07,
+      "loss": 0.8281,
+      "step": 18970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.156101006050016e-07,
+      "loss": 0.8765,
+      "step": 18971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.152112263528301e-07,
+      "loss": 0.3796,
+      "step": 18972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.14812430249103e-07,
+      "loss": 0.8276,
+      "step": 18973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.14413712301453e-07,
+      "loss": 0.8364,
+      "step": 18974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.140150725175051e-07,
+      "loss": 0.8159,
+      "step": 18975
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.136165109048878e-07,
+      "loss": 0.8403,
+      "step": 18976
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.132180274712277e-07,
+      "loss": 0.8042,
+      "step": 18977
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.128196222241459e-07,
+      "loss": 0.8218,
+      "step": 18978
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.124212951712657e-07,
+      "loss": 0.8843,
+      "step": 18979
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.120230463202063e-07,
+      "loss": 0.834,
+      "step": 18980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.116248756785879e-07,
+      "loss": 0.8193,
+      "step": 18981
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.112267832540289e-07,
+      "loss": 0.8062,
+      "step": 18982
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.108287690541439e-07,
+      "loss": 0.835,
+      "step": 18983
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.104308330865491e-07,
+      "loss": 0.793,
+      "step": 18984
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.100329753588572e-07,
+      "loss": 0.7949,
+      "step": 18985
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.096351958786781e-07,
+      "loss": 0.8286,
+      "step": 18986
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.092374946536242e-07,
+      "loss": 0.854,
+      "step": 18987
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.088398716913027e-07,
+      "loss": 0.7991,
+      "step": 18988
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.084423269993209e-07,
+      "loss": 0.8008,
+      "step": 18989
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.080448605852859e-07,
+      "loss": 0.7905,
+      "step": 18990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.076474724568013e-07,
+      "loss": 0.832,
+      "step": 18991
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.072501626214692e-07,
+      "loss": 0.8682,
+      "step": 18992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.068529310868896e-07,
+      "loss": 0.8423,
+      "step": 18993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.064557778606631e-07,
+      "loss": 0.8003,
+      "step": 18994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.060587029503898e-07,
+      "loss": 0.7764,
+      "step": 18995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.05661706363663e-07,
+      "loss": 0.9062,
+      "step": 18996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.052647881080817e-07,
+      "loss": 0.7998,
+      "step": 18997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.048679481912371e-07,
+      "loss": 0.8691,
+      "step": 18998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.044711866207201e-07,
+      "loss": 0.7881,
+      "step": 18999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.040745034041248e-07,
+      "loss": 0.8135,
+      "step": 19000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.036778985490374e-07,
+      "loss": 0.8286,
+      "step": 19001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.032813720630468e-07,
+      "loss": 0.8687,
+      "step": 19002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.028849239537402e-07,
+      "loss": 0.7935,
+      "step": 19003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.024885542287016e-07,
+      "loss": 0.8472,
+      "step": 19004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.02092262895512e-07,
+      "loss": 0.2944,
+      "step": 19005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.016960499617566e-07,
+      "loss": 0.8677,
+      "step": 19006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.012999154350127e-07,
+      "loss": 0.8447,
+      "step": 19007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.009038593228614e-07,
+      "loss": 0.7808,
+      "step": 19008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.005078816328772e-07,
+      "loss": 0.7808,
+      "step": 19009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.001119823726384e-07,
+      "loss": 0.8857,
+      "step": 19010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.997161615497179e-07,
+      "loss": 0.8638,
+      "step": 19011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.993204191716875e-07,
+      "loss": 0.8765,
+      "step": 19012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.989247552461194e-07,
+      "loss": 0.8325,
+      "step": 19013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.985291697805843e-07,
+      "loss": 0.8994,
+      "step": 19014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.981336627826481e-07,
+      "loss": 0.8569,
+      "step": 19015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.977382342598801e-07,
+      "loss": 0.8667,
+      "step": 19016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.973428842198439e-07,
+      "loss": 0.8184,
+      "step": 19017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.969476126701021e-07,
+      "loss": 0.8223,
+      "step": 19018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.965524196182191e-07,
+      "loss": 0.7734,
+      "step": 19019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.961573050717537e-07,
+      "loss": 0.8203,
+      "step": 19020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.957622690382656e-07,
+      "loss": 0.8359,
+      "step": 19021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.953673115253142e-07,
+      "loss": 0.8037,
+      "step": 19022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.949724325404541e-07,
+      "loss": 0.8506,
+      "step": 19023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.945776320912397e-07,
+      "loss": 0.8203,
+      "step": 19024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.94182910185224e-07,
+      "loss": 0.8213,
+      "step": 19025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.937882668299586e-07,
+      "loss": 0.9028,
+      "step": 19026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.933937020329958e-07,
+      "loss": 0.8555,
+      "step": 19027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.929992158018813e-07,
+      "loss": 0.8452,
+      "step": 19028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.926048081441651e-07,
+      "loss": 0.8535,
+      "step": 19029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.922104790673913e-07,
+      "loss": 0.8125,
+      "step": 19030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.918162285791026e-07,
+      "loss": 0.8545,
+      "step": 19031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.914220566868447e-07,
+      "loss": 0.8521,
+      "step": 19032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.910279633981561e-07,
+      "loss": 0.8569,
+      "step": 19033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.906339487205773e-07,
+      "loss": 0.8281,
+      "step": 19034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.902400126616472e-07,
+      "loss": 0.8711,
+      "step": 19035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.898461552289028e-07,
+      "loss": 0.8452,
+      "step": 19036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.89452376429878e-07,
+      "loss": 0.9062,
+      "step": 19037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.890586762721049e-07,
+      "loss": 0.8418,
+      "step": 19038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.886650547631182e-07,
+      "loss": 0.8721,
+      "step": 19039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.882715119104479e-07,
+      "loss": 0.7964,
+      "step": 19040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.878780477216225e-07,
+      "loss": 0.8623,
+      "step": 19041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.874846622041705e-07,
+      "loss": 0.8071,
+      "step": 19042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.87091355365618e-07,
+      "loss": 0.8213,
+      "step": 19043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.866981272134878e-07,
+      "loss": 0.7759,
+      "step": 19044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.863049777553051e-07,
+      "loss": 0.7908,
+      "step": 19045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.859119069985894e-07,
+      "loss": 0.8022,
+      "step": 19046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.85518914950862e-07,
+      "loss": 0.8589,
+      "step": 19047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.851260016196423e-07,
+      "loss": 0.7959,
+      "step": 19048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.847331670124465e-07,
+      "loss": 0.8438,
+      "step": 19049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.843404111367887e-07,
+      "loss": 0.8149,
+      "step": 19050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.839477340001862e-07,
+      "loss": 0.8057,
+      "step": 19051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.835551356101479e-07,
+      "loss": 0.8052,
+      "step": 19052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.831626159741879e-07,
+      "loss": 0.7534,
+      "step": 19053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.82770175099813e-07,
+      "loss": 0.8027,
+      "step": 19054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.823778129945326e-07,
+      "loss": 0.8628,
+      "step": 19055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.819855296658564e-07,
+      "loss": 0.7891,
+      "step": 19056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.815933251212827e-07,
+      "loss": 0.8477,
+      "step": 19057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.812011993683206e-07,
+      "loss": 0.8237,
+      "step": 19058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.80809152414468e-07,
+      "loss": 0.856,
+      "step": 19059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.804171842672276e-07,
+      "loss": 0.8394,
+      "step": 19060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.800252949340998e-07,
+      "loss": 0.793,
+      "step": 19061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.796334844225784e-07,
+      "loss": 0.8779,
+      "step": 19062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.792417527401632e-07,
+      "loss": 0.8477,
+      "step": 19063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.788500998943466e-07,
+      "loss": 0.8135,
+      "step": 19064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.784585258926204e-07,
+      "loss": 0.7969,
+      "step": 19065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.780670307424787e-07,
+      "loss": 0.8901,
+      "step": 19066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.776756144514087e-07,
+      "loss": 0.328,
+      "step": 19067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.772842770269003e-07,
+      "loss": 0.8481,
+      "step": 19068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.768930184764412e-07,
+      "loss": 0.8652,
+      "step": 19069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.765018388075158e-07,
+      "loss": 0.8994,
+      "step": 19070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.761107380276074e-07,
+      "loss": 0.8008,
+      "step": 19071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.757197161441999e-07,
+      "loss": 0.8457,
+      "step": 19072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.753287731647719e-07,
+      "loss": 0.8325,
+      "step": 19073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.749379090968058e-07,
+      "loss": 0.8145,
+      "step": 19074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.74547123947776e-07,
+      "loss": 0.8379,
+      "step": 19075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.741564177251622e-07,
+      "loss": 0.8062,
+      "step": 19076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.737657904364377e-07,
+      "loss": 0.8374,
+      "step": 19077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.733752420890746e-07,
+      "loss": 0.8628,
+      "step": 19078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.729847726905455e-07,
+      "loss": 0.8062,
+      "step": 19079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.725943822483224e-07,
+      "loss": 0.8174,
+      "step": 19080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.72204070769872e-07,
+      "loss": 0.8281,
+      "step": 19081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.718138382626629e-07,
+      "loss": 0.8286,
+      "step": 19082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.714236847341606e-07,
+      "loss": 0.7944,
+      "step": 19083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.710336101918276e-07,
+      "loss": 0.8618,
+      "step": 19084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.706436146431297e-07,
+      "loss": 0.8911,
+      "step": 19085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.702536980955251e-07,
+      "loss": 0.811,
+      "step": 19086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.698638605564752e-07,
+      "loss": 0.8037,
+      "step": 19087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.694741020334385e-07,
+      "loss": 0.8325,
+      "step": 19088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.690844225338718e-07,
+      "loss": 0.8066,
+      "step": 19089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.686948220652292e-07,
+      "loss": 0.7998,
+      "step": 19090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.683053006349634e-07,
+      "loss": 0.8599,
+      "step": 19091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.679158582505281e-07,
+      "loss": 0.8135,
+      "step": 19092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.675264949193751e-07,
+      "loss": 0.8037,
+      "step": 19093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.671372106489511e-07,
+      "loss": 0.8506,
+      "step": 19094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.667480054467059e-07,
+      "loss": 0.814,
+      "step": 19095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.66358879320085e-07,
+      "loss": 0.813,
+      "step": 19096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.659698322765308e-07,
+      "loss": 0.8184,
+      "step": 19097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.655808643234898e-07,
+      "loss": 0.8599,
+      "step": 19098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.651919754684002e-07,
+      "loss": 0.8433,
+      "step": 19099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.64803165718704e-07,
+      "loss": 0.8823,
+      "step": 19100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.644144350818406e-07,
+      "loss": 0.7881,
+      "step": 19101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.64025783565246e-07,
+      "loss": 0.8887,
+      "step": 19102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.636372111763558e-07,
+      "loss": 0.8438,
+      "step": 19103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.632487179226023e-07,
+      "loss": 0.8335,
+      "step": 19104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.628603038114197e-07,
+      "loss": 0.7627,
+      "step": 19105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.624719688502398e-07,
+      "loss": 0.353,
+      "step": 19106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.620837130464904e-07,
+      "loss": 0.8481,
+      "step": 19107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.616955364076008e-07,
+      "loss": 0.8438,
+      "step": 19108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.61307438940997e-07,
+      "loss": 0.8276,
+      "step": 19109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.609194206541022e-07,
+      "loss": 0.8125,
+      "step": 19110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.605314815543431e-07,
+      "loss": 0.7654,
+      "step": 19111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.601436216491377e-07,
+      "loss": 0.8149,
+      "step": 19112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.597558409459089e-07,
+      "loss": 0.8408,
+      "step": 19113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.593681394520764e-07,
+      "loss": 0.8149,
+      "step": 19114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.589805171750559e-07,
+      "loss": 0.8887,
+      "step": 19115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.585929741222626e-07,
+      "loss": 0.8179,
+      "step": 19116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.582055103011128e-07,
+      "loss": 0.7754,
+      "step": 19117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.578181257190171e-07,
+      "loss": 0.8364,
+      "step": 19118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.574308203833892e-07,
+      "loss": 0.8159,
+      "step": 19119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.570435943016359e-07,
+      "loss": 0.8545,
+      "step": 19120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.566564474811689e-07,
+      "loss": 0.8457,
+      "step": 19121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.562693799293931e-07,
+      "loss": 0.7998,
+      "step": 19122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.558823916537123e-07,
+      "loss": 0.8511,
+      "step": 19123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.554954826615331e-07,
+      "loss": 0.7418,
+      "step": 19124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.551086529602548e-07,
+      "loss": 0.3669,
+      "step": 19125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.547219025572795e-07,
+      "loss": 0.8818,
+      "step": 19126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.543352314600078e-07,
+      "loss": 0.834,
+      "step": 19127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.539486396758357e-07,
+      "loss": 0.8032,
+      "step": 19128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.535621272121586e-07,
+      "loss": 0.7998,
+      "step": 19129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53175694076373e-07,
+      "loss": 0.791,
+      "step": 19130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527893402758697e-07,
+      "loss": 0.9219,
+      "step": 19131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.524030658180427e-07,
+      "loss": 0.792,
+      "step": 19132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.520168707102794e-07,
+      "loss": 0.8105,
+      "step": 19133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.516307549599711e-07,
+      "loss": 0.7529,
+      "step": 19134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512447185745037e-07,
+      "loss": 0.8447,
+      "step": 19135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.508587615612613e-07,
+      "loss": 0.8208,
+      "step": 19136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.504728839276288e-07,
+      "loss": 0.853,
+      "step": 19137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.500870856809901e-07,
+      "loss": 0.8188,
+      "step": 19138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.497013668287241e-07,
+      "loss": 0.8628,
+      "step": 19139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.493157273782116e-07,
+      "loss": 0.8179,
+      "step": 19140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.489301673368299e-07,
+      "loss": 0.8213,
+      "step": 19141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.485446867119546e-07,
+      "loss": 0.8247,
+      "step": 19142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.481592855109627e-07,
+      "loss": 0.8359,
+      "step": 19143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.477739637412247e-07,
+      "loss": 0.8452,
+      "step": 19144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.473887214101134e-07,
+      "loss": 0.875,
+      "step": 19145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.470035585250014e-07,
+      "loss": 0.8374,
+      "step": 19146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.466184750932554e-07,
+      "loss": 0.8301,
+      "step": 19147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.462334711222425e-07,
+      "loss": 0.8154,
+      "step": 19148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.458485466193284e-07,
+      "loss": 0.8867,
+      "step": 19149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.454637015918771e-07,
+      "loss": 0.7861,
+      "step": 19150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.450789360472533e-07,
+      "loss": 0.8428,
+      "step": 19151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.446942499928157e-07,
+      "loss": 0.8174,
+      "step": 19152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.44309643435926e-07,
+      "loss": 0.8486,
+      "step": 19153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.439251163839419e-07,
+      "loss": 0.8877,
+      "step": 19154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.435406688442177e-07,
+      "loss": 0.3097,
+      "step": 19155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.431563008241117e-07,
+      "loss": 0.7676,
+      "step": 19156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.427720123309752e-07,
+      "loss": 0.8604,
+      "step": 19157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.423878033721605e-07,
+      "loss": 0.9131,
+      "step": 19158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.420036739550203e-07,
+      "loss": 0.8799,
+      "step": 19159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.416196240869018e-07,
+      "loss": 0.8511,
+      "step": 19160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.412356537751526e-07,
+      "loss": 0.916,
+      "step": 19161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.408517630271179e-07,
+      "loss": 0.7734,
+      "step": 19162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.404679518501424e-07,
+      "loss": 0.8672,
+      "step": 19163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.400842202515714e-07,
+      "loss": 0.833,
+      "step": 19164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.397005682387427e-07,
+      "loss": 0.8848,
+      "step": 19165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.393169958189984e-07,
+      "loss": 0.8809,
+      "step": 19166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.389335029996787e-07,
+      "loss": 0.8721,
+      "step": 19167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.385500897881155e-07,
+      "loss": 0.8633,
+      "step": 19168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.381667561916485e-07,
+      "loss": 0.7861,
+      "step": 19169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.377835022176079e-07,
+      "loss": 0.8145,
+      "step": 19170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.374003278733279e-07,
+      "loss": 0.8149,
+      "step": 19171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.370172331661403e-07,
+      "loss": 0.8647,
+      "step": 19172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.366342181033716e-07,
+      "loss": 0.8447,
+      "step": 19173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.362512826923525e-07,
+      "loss": 0.7642,
+      "step": 19174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.358684269404077e-07,
+      "loss": 0.7959,
+      "step": 19175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.354856508548603e-07,
+      "loss": 0.8418,
+      "step": 19176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.351029544430361e-07,
+      "loss": 0.8091,
+      "step": 19177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.347203377122542e-07,
+      "loss": 0.8496,
+      "step": 19178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.343378006698361e-07,
+      "loss": 0.8374,
+      "step": 19179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.339553433231012e-07,
+      "loss": 0.7969,
+      "step": 19180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.33572965679365e-07,
+      "loss": 0.7847,
+      "step": 19181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.331906677459428e-07,
+      "loss": 0.8262,
+      "step": 19182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.328084495301502e-07,
+      "loss": 0.7905,
+      "step": 19183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.324263110392977e-07,
+      "loss": 0.8223,
+      "step": 19184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.320442522806982e-07,
+      "loss": 0.8325,
+      "step": 19185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.316622732616591e-07,
+      "loss": 0.8794,
+      "step": 19186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.312803739894897e-07,
+      "loss": 0.8071,
+      "step": 19187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.308985544714964e-07,
+      "loss": 0.8115,
+      "step": 19188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.305168147149822e-07,
+      "loss": 0.7998,
+      "step": 19189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.301351547272529e-07,
+      "loss": 0.8706,
+      "step": 19190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.297535745156077e-07,
+      "loss": 0.7979,
+      "step": 19191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.293720740873479e-07,
+      "loss": 0.8369,
+      "step": 19192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.289906534497739e-07,
+      "loss": 0.8096,
+      "step": 19193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.286093126101807e-07,
+      "loss": 0.8618,
+      "step": 19194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.282280515758639e-07,
+      "loss": 0.8501,
+      "step": 19195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.278468703541198e-07,
+      "loss": 0.8491,
+      "step": 19196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.274657689522381e-07,
+      "loss": 0.8379,
+      "step": 19197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.270847473775129e-07,
+      "loss": 0.7441,
+      "step": 19198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.267038056372306e-07,
+      "loss": 0.8545,
+      "step": 19199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.263229437386821e-07,
+      "loss": 0.8013,
+      "step": 19200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.259421616891527e-07,
+      "loss": 0.8774,
+      "step": 19201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.255614594959261e-07,
+      "loss": 0.8315,
+      "step": 19202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.251808371662861e-07,
+      "loss": 0.8013,
+      "step": 19203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.248002947075173e-07,
+      "loss": 0.7627,
+      "step": 19204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.24419832126897e-07,
+      "loss": 0.8496,
+      "step": 19205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.240394494317061e-07,
+      "loss": 0.7661,
+      "step": 19206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.236591466292216e-07,
+      "loss": 0.813,
+      "step": 19207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.232789237267169e-07,
+      "loss": 0.8589,
+      "step": 19208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.228987807314698e-07,
+      "loss": 0.8174,
+      "step": 19209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.225187176507494e-07,
+      "loss": 0.8149,
+      "step": 19210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.221387344918291e-07,
+      "loss": 0.8296,
+      "step": 19211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.217588312619801e-07,
+      "loss": 0.8716,
+      "step": 19212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.21379007968468e-07,
+      "loss": 0.9106,
+      "step": 19213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.209992646185605e-07,
+      "loss": 0.834,
+      "step": 19214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.206196012195211e-07,
+      "loss": 0.8848,
+      "step": 19215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.202400177786146e-07,
+      "loss": 0.7876,
+      "step": 19216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.198605143031041e-07,
+      "loss": 0.7896,
+      "step": 19217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.194810908002482e-07,
+      "loss": 0.8481,
+      "step": 19218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.191017472773077e-07,
+      "loss": 0.8149,
+      "step": 19219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.187224837415391e-07,
+      "loss": 0.8174,
+      "step": 19220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.183433002001967e-07,
+      "loss": 0.8291,
+      "step": 19221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.179641966605384e-07,
+      "loss": 0.8169,
+      "step": 19222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.175851731298134e-07,
+      "loss": 0.345,
+      "step": 19223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.172062296152744e-07,
+      "loss": 0.791,
+      "step": 19224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.168273661241732e-07,
+      "loss": 0.8325,
+      "step": 19225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.164485826637564e-07,
+      "loss": 0.9038,
+      "step": 19226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.160698792412703e-07,
+      "loss": 0.8755,
+      "step": 19227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.156912558639595e-07,
+      "loss": 0.8418,
+      "step": 19228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.153127125390686e-07,
+      "loss": 0.7896,
+      "step": 19229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.149342492738411e-07,
+      "loss": 0.7437,
+      "step": 19230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.145558660755143e-07,
+      "loss": 0.8262,
+      "step": 19231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.14177562951331e-07,
+      "loss": 0.793,
+      "step": 19232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.13799339908527e-07,
+      "loss": 0.8066,
+      "step": 19233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.134211969543371e-07,
+      "loss": 0.833,
+      "step": 19234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.130431340959982e-07,
+      "loss": 0.8047,
+      "step": 19235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.126651513407401e-07,
+      "loss": 0.8823,
+      "step": 19236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.122872486957966e-07,
+      "loss": 0.7837,
+      "step": 19237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.119094261683985e-07,
+      "loss": 0.8032,
+      "step": 19238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.115316837657721e-07,
+      "loss": 0.8115,
+      "step": 19239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.111540214951442e-07,
+      "loss": 0.8452,
+      "step": 19240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.107764393637418e-07,
+      "loss": 0.8032,
+      "step": 19241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.103989373787857e-07,
+      "loss": 0.8701,
+      "step": 19242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.100215155475016e-07,
+      "loss": 0.8809,
+      "step": 19243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.096441738771072e-07,
+      "loss": 0.8647,
+      "step": 19244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.092669123748242e-07,
+      "loss": 0.8286,
+      "step": 19245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.088897310478688e-07,
+      "loss": 0.8716,
+      "step": 19246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.085126299034563e-07,
+      "loss": 0.8242,
+      "step": 19247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.081356089488019e-07,
+      "loss": 0.8462,
+      "step": 19248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.077586681911198e-07,
+      "loss": 0.8867,
+      "step": 19249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.0738180763762e-07,
+      "loss": 0.8179,
+      "step": 19250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.070050272955138e-07,
+      "loss": 0.8706,
+      "step": 19251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.066283271720082e-07,
+      "loss": 0.7949,
+      "step": 19252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.062517072743103e-07,
+      "loss": 0.8188,
+      "step": 19253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.058751676096266e-07,
+      "loss": 0.8291,
+      "step": 19254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.05498708185159e-07,
+      "loss": 0.7866,
+      "step": 19255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.051223290081117e-07,
+      "loss": 0.8096,
+      "step": 19256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.047460300856835e-07,
+      "loss": 0.8584,
+      "step": 19257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.043698114250764e-07,
+      "loss": 0.792,
+      "step": 19258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.039936730334857e-07,
+      "loss": 0.8433,
+      "step": 19259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.036176149181068e-07,
+      "loss": 0.8398,
+      "step": 19260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.032416370861362e-07,
+      "loss": 0.8403,
+      "step": 19261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.02865739544767e-07,
+      "loss": 0.8037,
+      "step": 19262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.024899223011889e-07,
+      "loss": 0.8491,
+      "step": 19263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.021141853625947e-07,
+      "loss": 0.8672,
+      "step": 19264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.017385287361706e-07,
+      "loss": 0.7666,
+      "step": 19265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.013629524291033e-07,
+      "loss": 0.8105,
+      "step": 19266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.009874564485798e-07,
+      "loss": 0.7788,
+      "step": 19267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.006120408017826e-07,
+      "loss": 0.8418,
+      "step": 19268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.002367054958937e-07,
+      "loss": 0.8086,
+      "step": 19269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.998614505380959e-07,
+      "loss": 0.8447,
+      "step": 19270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.994862759355676e-07,
+      "loss": 0.8237,
+      "step": 19271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.991111816954855e-07,
+      "loss": 0.7827,
+      "step": 19272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.987361678250255e-07,
+      "loss": 0.854,
+      "step": 19273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.983612343313629e-07,
+      "loss": 0.814,
+      "step": 19274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.979863812216721e-07,
+      "loss": 0.8579,
+      "step": 19275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.976116085031222e-07,
+      "loss": 0.8013,
+      "step": 19276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.972369161828841e-07,
+      "loss": 0.9019,
+      "step": 19277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.968623042681289e-07,
+      "loss": 0.8726,
+      "step": 19278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.964877727660186e-07,
+      "loss": 0.811,
+      "step": 19279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.96113321683723e-07,
+      "loss": 0.8452,
+      "step": 19280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.957389510284025e-07,
+      "loss": 0.8525,
+      "step": 19281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.953646608072208e-07,
+      "loss": 0.7876,
+      "step": 19282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.949904510273399e-07,
+      "loss": 0.7686,
+      "step": 19283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.946163216959169e-07,
+      "loss": 0.9092,
+      "step": 19284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.942422728201121e-07,
+      "loss": 0.8618,
+      "step": 19285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.938683044070794e-07,
+      "loss": 0.7979,
+      "step": 19286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.934944164639735e-07,
+      "loss": 0.7988,
+      "step": 19287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.931206089979488e-07,
+      "loss": 0.8379,
+      "step": 19288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.927468820161554e-07,
+      "loss": 0.751,
+      "step": 19289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.923732355257435e-07,
+      "loss": 0.8154,
+      "step": 19290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.919996695338633e-07,
+      "loss": 0.7993,
+      "step": 19291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.916261840476602e-07,
+      "loss": 0.8188,
+      "step": 19292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.912527790742802e-07,
+      "loss": 0.8745,
+      "step": 19293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.908794546208653e-07,
+      "loss": 0.8115,
+      "step": 19294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.905062106945594e-07,
+      "loss": 0.8491,
+      "step": 19295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.901330473025037e-07,
+      "loss": 0.8125,
+      "step": 19296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.897599644518361e-07,
+      "loss": 0.8545,
+      "step": 19297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.893869621496958e-07,
+      "loss": 0.8369,
+      "step": 19298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.890140404032176e-07,
+      "loss": 0.8623,
+      "step": 19299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.886411992195358e-07,
+      "loss": 0.8843,
+      "step": 19300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.882684386057854e-07,
+      "loss": 0.8418,
+      "step": 19301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.878957585690955e-07,
+      "loss": 0.7944,
+      "step": 19302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.875231591165967e-07,
+      "loss": 0.8076,
+      "step": 19303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.871506402554197e-07,
+      "loss": 0.8105,
+      "step": 19304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.86778201992689e-07,
+      "loss": 0.8462,
+      "step": 19305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.864058443355299e-07,
+      "loss": 0.8398,
+      "step": 19306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.860335672910674e-07,
+      "loss": 0.8291,
+      "step": 19307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.856613708664224e-07,
+      "loss": 0.8486,
+      "step": 19308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.852892550687175e-07,
+      "loss": 0.8662,
+      "step": 19309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.849172199050692e-07,
+      "loss": 0.8389,
+      "step": 19310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.845452653825981e-07,
+      "loss": 0.7969,
+      "step": 19311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.841733915084188e-07,
+      "loss": 0.8042,
+      "step": 19312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.838015982896441e-07,
+      "loss": 0.7661,
+      "step": 19313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.834298857333894e-07,
+      "loss": 0.8164,
+      "step": 19314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.830582538467663e-07,
+      "loss": 0.8154,
+      "step": 19315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.826867026368829e-07,
+      "loss": 0.8076,
+      "step": 19316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.823152321108496e-07,
+      "loss": 0.7925,
+      "step": 19317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.819438422757724e-07,
+      "loss": 0.8281,
+      "step": 19318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.815725331387546e-07,
+      "loss": 0.8633,
+      "step": 19319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.812013047069034e-07,
+      "loss": 0.8394,
+      "step": 19320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.808301569873178e-07,
+      "loss": 0.8091,
+      "step": 19321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.804590899871e-07,
+      "loss": 0.8555,
+      "step": 19322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.800881037133501e-07,
+      "loss": 0.8042,
+      "step": 19323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.797171981731644e-07,
+      "loss": 0.8296,
+      "step": 19324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.793463733736395e-07,
+      "loss": 0.7734,
+      "step": 19325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.789756293218675e-07,
+      "loss": 0.793,
+      "step": 19326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.786049660249439e-07,
+      "loss": 0.8931,
+      "step": 19327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.782343834899603e-07,
+      "loss": 0.9014,
+      "step": 19328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.778638817240042e-07,
+      "loss": 0.853,
+      "step": 19329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.774934607341666e-07,
+      "loss": 0.8447,
+      "step": 19330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.771231205275331e-07,
+      "loss": 0.7207,
+      "step": 19331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.767528611111874e-07,
+      "loss": 0.8677,
+      "step": 19332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.763826824922161e-07,
+      "loss": 0.8921,
+      "step": 19333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.760125846776978e-07,
+      "loss": 0.8359,
+      "step": 19334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.756425676747154e-07,
+      "loss": 0.8203,
+      "step": 19335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.752726314903486e-07,
+      "loss": 0.8125,
+      "step": 19336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.749027761316735e-07,
+      "loss": 0.8281,
+      "step": 19337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.745330016057661e-07,
+      "loss": 0.8071,
+      "step": 19338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.741633079197002e-07,
+      "loss": 0.8311,
+      "step": 19339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.737936950805491e-07,
+      "loss": 0.8594,
+      "step": 19340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.734241630953854e-07,
+      "loss": 0.8306,
+      "step": 19341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.730547119712762e-07,
+      "loss": 0.813,
+      "step": 19342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.72685341715293e-07,
+      "loss": 0.8149,
+      "step": 19343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.723160523345003e-07,
+      "loss": 0.8428,
+      "step": 19344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.719468438359623e-07,
+      "loss": 0.8838,
+      "step": 19345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.715777162267451e-07,
+      "loss": 0.8301,
+      "step": 19346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.712086695139081e-07,
+      "loss": 0.7959,
+      "step": 19347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.708397037045129e-07,
+      "loss": 0.7471,
+      "step": 19348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.704708188056198e-07,
+      "loss": 0.3884,
+      "step": 19349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.70102014824285e-07,
+      "loss": 0.7891,
+      "step": 19350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.697332917675626e-07,
+      "loss": 0.8911,
+      "step": 19351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.693646496425094e-07,
+      "loss": 0.8564,
+      "step": 19352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.689960884561764e-07,
+      "loss": 0.918,
+      "step": 19353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.686276082156169e-07,
+      "loss": 0.8098,
+      "step": 19354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.682592089278773e-07,
+      "loss": 0.8745,
+      "step": 19355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.678908906000088e-07,
+      "loss": 0.8022,
+      "step": 19356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.675226532390567e-07,
+      "loss": 0.8433,
+      "step": 19357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.671544968520645e-07,
+      "loss": 0.8091,
+      "step": 19358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.667864214460779e-07,
+      "loss": 0.7881,
+      "step": 19359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.664184270281366e-07,
+      "loss": 0.7939,
+      "step": 19360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.660505136052826e-07,
+      "loss": 0.7847,
+      "step": 19361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.656826811845547e-07,
+      "loss": 0.8418,
+      "step": 19362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.6531492977299e-07,
+      "loss": 0.8828,
+      "step": 19363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.649472593776225e-07,
+      "loss": 0.8564,
+      "step": 19364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.645796700054886e-07,
+      "loss": 0.8882,
+      "step": 19365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.642121616636183e-07,
+      "loss": 0.8145,
+      "step": 19366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.63844734359046e-07,
+      "loss": 0.8154,
+      "step": 19367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.634773880987978e-07,
+      "loss": 0.7495,
+      "step": 19368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.631101228899041e-07,
+      "loss": 0.8442,
+      "step": 19369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.627429387393903e-07,
+      "loss": 0.8184,
+      "step": 19370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.623758356542799e-07,
+      "loss": 0.8442,
+      "step": 19371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.620088136415977e-07,
+      "loss": 0.8335,
+      "step": 19372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.616418727083663e-07,
+      "loss": 0.8135,
+      "step": 19373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.612750128616031e-07,
+      "loss": 0.8472,
+      "step": 19374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.609082341083301e-07,
+      "loss": 0.8784,
+      "step": 19375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.605415364555623e-07,
+      "loss": 0.8325,
+      "step": 19376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.601749199103143e-07,
+      "loss": 0.8032,
+      "step": 19377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.598083844796023e-07,
+      "loss": 0.8149,
+      "step": 19378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.594419301704364e-07,
+      "loss": 0.8481,
+      "step": 19379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.59075556989829e-07,
+      "loss": 0.8145,
+      "step": 19380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.587092649447897e-07,
+      "loss": 0.7749,
+      "step": 19381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.583430540423259e-07,
+      "loss": 0.8608,
+      "step": 19382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.579769242894436e-07,
+      "loss": 0.7271,
+      "step": 19383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.57610875693146e-07,
+      "loss": 0.8804,
+      "step": 19384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.572449082604372e-07,
+      "loss": 0.8433,
+      "step": 19385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.568790219983208e-07,
+      "loss": 0.8525,
+      "step": 19386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.565132169137934e-07,
+      "loss": 0.8218,
+      "step": 19387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.561474930138557e-07,
+      "loss": 0.8677,
+      "step": 19388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.557818503055059e-07,
+      "loss": 0.8594,
+      "step": 19389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.55416288795735e-07,
+      "loss": 0.9199,
+      "step": 19390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.550508084915404e-07,
+      "loss": 0.8281,
+      "step": 19391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.546854093999117e-07,
+      "loss": 0.8516,
+      "step": 19392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.543200915278409e-07,
+      "loss": 0.8359,
+      "step": 19393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.539548548823184e-07,
+      "loss": 0.8462,
+      "step": 19394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.535896994703295e-07,
+      "loss": 0.8735,
+      "step": 19395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.532246252988617e-07,
+      "loss": 0.854,
+      "step": 19396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.528596323748993e-07,
+      "loss": 0.8125,
+      "step": 19397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.52494720705424e-07,
+      "loss": 0.8193,
+      "step": 19398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.521298902974183e-07,
+      "loss": 0.8696,
+      "step": 19399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.517651411578613e-07,
+      "loss": 0.8682,
+      "step": 19400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.514004732937307e-07,
+      "loss": 0.812,
+      "step": 19401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.51035886712006e-07,
+      "loss": 0.8545,
+      "step": 19402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.506713814196597e-07,
+      "loss": 0.304,
+      "step": 19403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.50306957423666e-07,
+      "loss": 0.7925,
+      "step": 19404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.499426147309963e-07,
+      "loss": 0.8306,
+      "step": 19405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.495783533486212e-07,
+      "loss": 0.8745,
+      "step": 19406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.492141732835112e-07,
+      "loss": 0.8101,
+      "step": 19407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.488500745426313e-07,
+      "loss": 0.8335,
+      "step": 19408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.484860571329494e-07,
+      "loss": 0.3303,
+      "step": 19409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.481221210614292e-07,
+      "loss": 0.8853,
+      "step": 19410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.477582663350314e-07,
+      "loss": 0.8608,
+      "step": 19411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.473944929607202e-07,
+      "loss": 0.8096,
+      "step": 19412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.470308009454525e-07,
+      "loss": 0.8359,
+      "step": 19413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.46667190296187e-07,
+      "loss": 0.8433,
+      "step": 19414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.463036610198821e-07,
+      "loss": 0.7739,
+      "step": 19415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.45940213123491e-07,
+      "loss": 0.8633,
+      "step": 19416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.455768466139662e-07,
+      "loss": 0.8564,
+      "step": 19417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.452135614982615e-07,
+      "loss": 0.8022,
+      "step": 19418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.448503577833249e-07,
+      "loss": 0.8477,
+      "step": 19419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.444872354761074e-07,
+      "loss": 0.813,
+      "step": 19420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.441241945835537e-07,
+      "loss": 0.8374,
+      "step": 19421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.437612351126116e-07,
+      "loss": 0.7822,
+      "step": 19422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.433983570702241e-07,
+      "loss": 0.8691,
+      "step": 19423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.430355604633327e-07,
+      "loss": 0.8022,
+      "step": 19424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.426728452988807e-07,
+      "loss": 0.7935,
+      "step": 19425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.423102115838038e-07,
+      "loss": 0.4407,
+      "step": 19426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.419476593250425e-07,
+      "loss": 0.8345,
+      "step": 19427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.415851885295333e-07,
+      "loss": 0.8809,
+      "step": 19428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.412227992042098e-07,
+      "loss": 0.8477,
+      "step": 19429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.408604913560041e-07,
+      "loss": 0.8022,
+      "step": 19430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.404982649918502e-07,
+      "loss": 0.7969,
+      "step": 19431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.401361201186752e-07,
+      "loss": 0.8086,
+      "step": 19432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.397740567434097e-07,
+      "loss": 0.7998,
+      "step": 19433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.394120748729789e-07,
+      "loss": 0.834,
+      "step": 19434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.390501745143103e-07,
+      "loss": 0.811,
+      "step": 19435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.386883556743262e-07,
+      "loss": 0.8389,
+      "step": 19436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.38326618359948e-07,
+      "loss": 0.7974,
+      "step": 19437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.379649625780966e-07,
+      "loss": 0.8735,
+      "step": 19438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.376033883356931e-07,
+      "loss": 0.8228,
+      "step": 19439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.372418956396521e-07,
+      "loss": 0.8843,
+      "step": 19440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.368804844968919e-07,
+      "loss": 0.8208,
+      "step": 19441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.365191549143263e-07,
+      "loss": 0.8203,
+      "step": 19442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.361579068988667e-07,
+      "loss": 0.8071,
+      "step": 19443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.357967404574257e-07,
+      "loss": 0.8574,
+      "step": 19444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.354356555969122e-07,
+      "loss": 0.8511,
+      "step": 19445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.350746523242342e-07,
+      "loss": 0.7529,
+      "step": 19446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.347137306463004e-07,
+      "loss": 0.8477,
+      "step": 19447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.343528905700137e-07,
+      "loss": 0.8276,
+      "step": 19448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.339921321022786e-07,
+      "loss": 0.8027,
+      "step": 19449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.336314552499946e-07,
+      "loss": 0.7944,
+      "step": 19450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.332708600200644e-07,
+      "loss": 0.8179,
+      "step": 19451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.32910346419387e-07,
+      "loss": 0.8047,
+      "step": 19452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.325499144548575e-07,
+      "loss": 0.8052,
+      "step": 19453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.321895641333742e-07,
+      "loss": 0.7729,
+      "step": 19454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3182929546183e-07,
+      "loss": 0.8149,
+      "step": 19455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.314691084471154e-07,
+      "loss": 0.8257,
+      "step": 19456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.311090030961249e-07,
+      "loss": 0.7881,
+      "step": 19457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.307489794157446e-07,
+      "loss": 0.8228,
+      "step": 19458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.303890374128636e-07,
+      "loss": 0.8838,
+      "step": 19459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.300291770943701e-07,
+      "loss": 0.8726,
+      "step": 19460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.296693984671465e-07,
+      "loss": 0.7832,
+      "step": 19461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.293097015380773e-07,
+      "loss": 0.3524,
+      "step": 19462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.289500863140414e-07,
+      "loss": 0.8662,
+      "step": 19463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.285905528019211e-07,
+      "loss": 0.811,
+      "step": 19464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.282311010085952e-07,
+      "loss": 0.7964,
+      "step": 19465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.278717309409389e-07,
+      "loss": 0.8262,
+      "step": 19466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.275124426058295e-07,
+      "loss": 0.8442,
+      "step": 19467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.271532360101396e-07,
+      "loss": 0.8008,
+      "step": 19468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.267941111607407e-07,
+      "loss": 0.8135,
+      "step": 19469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.264350680645049e-07,
+      "loss": 0.8677,
+      "step": 19470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.260761067283001e-07,
+      "loss": 0.7837,
+      "step": 19471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.257172271589935e-07,
+      "loss": 0.8066,
+      "step": 19472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.25358429363453e-07,
+      "loss": 0.7336,
+      "step": 19473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.24999713348542e-07,
+      "loss": 0.8579,
+      "step": 19474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.246410791211216e-07,
+      "loss": 0.8213,
+      "step": 19475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.24282526688056e-07,
+      "loss": 0.7891,
+      "step": 19476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.239240560562022e-07,
+      "loss": 0.8589,
+      "step": 19477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.235656672324204e-07,
+      "loss": 0.8696,
+      "step": 19478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.232073602235651e-07,
+      "loss": 0.8306,
+      "step": 19479
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.228491350364936e-07,
+      "loss": 0.853,
+      "step": 19480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.224909916780581e-07,
+      "loss": 0.8115,
+      "step": 19481
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.221329301551088e-07,
+      "loss": 0.8887,
+      "step": 19482
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.217749504744986e-07,
+      "loss": 0.8501,
+      "step": 19483
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.214170526430758e-07,
+      "loss": 0.8594,
+      "step": 19484
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.210592366676855e-07,
+      "loss": 0.7871,
+      "step": 19485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.207015025551767e-07,
+      "loss": 0.3901,
+      "step": 19486
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.203438503123911e-07,
+      "loss": 0.8135,
+      "step": 19487
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.199862799461704e-07,
+      "loss": 0.8384,
+      "step": 19488
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.196287914633576e-07,
+      "loss": 0.8271,
+      "step": 19489
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1927138487079e-07,
+      "loss": 0.8599,
+      "step": 19490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.189140601753081e-07,
+      "loss": 0.8013,
+      "step": 19491
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.185568173837443e-07,
+      "loss": 0.8364,
+      "step": 19492
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.18199656502937e-07,
+      "loss": 0.8604,
+      "step": 19493
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.178425775397174e-07,
+      "loss": 0.7666,
+      "step": 19494
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.174855805009162e-07,
+      "loss": 0.8408,
+      "step": 19495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.17128665393364e-07,
+      "loss": 0.873,
+      "step": 19496
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.167718322238909e-07,
+      "loss": 0.8276,
+      "step": 19497
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.164150809993209e-07,
+      "loss": 0.8447,
+      "step": 19498
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.160584117264812e-07,
+      "loss": 0.8896,
+      "step": 19499
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.15701824412195e-07,
+      "loss": 0.8232,
+      "step": 19500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.153453190632831e-07,
+      "loss": 0.8291,
+      "step": 19501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.14988895686568e-07,
+      "loss": 0.8135,
+      "step": 19502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.146325542888665e-07,
+      "loss": 0.8804,
+      "step": 19503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.142762948769977e-07,
+      "loss": 0.4113,
+      "step": 19504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.13920117457777e-07,
+      "loss": 0.9219,
+      "step": 19505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.13564022038018e-07,
+      "loss": 0.8809,
+      "step": 19506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.132080086245347e-07,
+      "loss": 0.8535,
+      "step": 19507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.128520772241376e-07,
+      "loss": 0.8237,
+      "step": 19508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.124962278436348e-07,
+      "loss": 0.8276,
+      "step": 19509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.121404604898363e-07,
+      "loss": 0.7925,
+      "step": 19510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.117847751695467e-07,
+      "loss": 0.8706,
+      "step": 19511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.114291718895716e-07,
+      "loss": 0.8877,
+      "step": 19512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.110736506567156e-07,
+      "loss": 0.812,
+      "step": 19513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.107182114777794e-07,
+      "loss": 0.8032,
+      "step": 19514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.103628543595626e-07,
+      "loss": 0.8848,
+      "step": 19515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.100075793088629e-07,
+      "loss": 0.7651,
+      "step": 19516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.096523863324784e-07,
+      "loss": 0.3361,
+      "step": 19517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.092972754372057e-07,
+      "loss": 0.8799,
+      "step": 19518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.089422466298363e-07,
+      "loss": 0.8896,
+      "step": 19519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.08587299917165e-07,
+      "loss": 0.854,
+      "step": 19520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.082324353059805e-07,
+      "loss": 0.8232,
+      "step": 19521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.078776528030718e-07,
+      "loss": 0.7954,
+      "step": 19522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.075229524152283e-07,
+      "loss": 0.8159,
+      "step": 19523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.07168334149233e-07,
+      "loss": 0.8911,
+      "step": 19524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.068137980118728e-07,
+      "loss": 0.8296,
+      "step": 19525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.06459344009931e-07,
+      "loss": 0.3545,
+      "step": 19526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.061049721501872e-07,
+      "loss": 0.8579,
+      "step": 19527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.057506824394223e-07,
+      "loss": 0.7339,
+      "step": 19528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.053964748844122e-07,
+      "loss": 0.8003,
+      "step": 19529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.05042349491935e-07,
+      "loss": 0.8647,
+      "step": 19530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.046883062687671e-07,
+      "loss": 0.8496,
+      "step": 19531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.043343452216788e-07,
+      "loss": 0.856,
+      "step": 19532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.039804663574451e-07,
+      "loss": 0.8662,
+      "step": 19533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.036266696828348e-07,
+      "loss": 0.8896,
+      "step": 19534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.032729552046153e-07,
+      "loss": 0.8013,
+      "step": 19535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.029193229295562e-07,
+      "loss": 0.8359,
+      "step": 19536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.025657728644208e-07,
+      "loss": 0.8062,
+      "step": 19537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.022123050159734e-07,
+      "loss": 0.8403,
+      "step": 19538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.018589193909791e-07,
+      "loss": 0.8984,
+      "step": 19539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.015056159961964e-07,
+      "loss": 0.9028,
+      "step": 19540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.011523948383836e-07,
+      "loss": 0.9062,
+      "step": 19541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.007992559243004e-07,
+      "loss": 0.8687,
+      "step": 19542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.004461992607015e-07,
+      "loss": 0.7886,
+      "step": 19543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.000932248543429e-07,
+      "loss": 0.7817,
+      "step": 19544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.997403327119761e-07,
+      "loss": 0.8423,
+      "step": 19545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.993875228403541e-07,
+      "loss": 0.8213,
+      "step": 19546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.990347952462251e-07,
+      "loss": 0.8286,
+      "step": 19547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.986821499363372e-07,
+      "loss": 0.8589,
+      "step": 19548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.983295869174378e-07,
+      "loss": 0.873,
+      "step": 19549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.979771061962731e-07,
+      "loss": 0.8101,
+      "step": 19550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.976247077795845e-07,
+      "loss": 0.8608,
+      "step": 19551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.972723916741165e-07,
+      "loss": 0.8506,
+      "step": 19552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.96920157886607e-07,
+      "loss": 0.8647,
+      "step": 19553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.965680064237951e-07,
+      "loss": 0.8276,
+      "step": 19554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.962159372924193e-07,
+      "loss": 0.8564,
+      "step": 19555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.958639504992138e-07,
+      "loss": 0.7432,
+      "step": 19556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.955120460509135e-07,
+      "loss": 0.749,
+      "step": 19557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.951602239542515e-07,
+      "loss": 0.8354,
+      "step": 19558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.94808484215958e-07,
+      "loss": 0.8232,
+      "step": 19559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.944568268427626e-07,
+      "loss": 0.7817,
+      "step": 19560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.941052518413921e-07,
+      "loss": 0.3958,
+      "step": 19561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.937537592185728e-07,
+      "loss": 0.8232,
+      "step": 19562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.93402348981031e-07,
+      "loss": 0.8579,
+      "step": 19563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.930510211354873e-07,
+      "loss": 0.8984,
+      "step": 19564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.92699775688666e-07,
+      "loss": 0.8198,
+      "step": 19565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.923486126472856e-07,
+      "loss": 0.8403,
+      "step": 19566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.919975320180627e-07,
+      "loss": 0.8169,
+      "step": 19567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.916465338077172e-07,
+      "loss": 0.8062,
+      "step": 19568
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.912956180229608e-07,
+      "loss": 0.7935,
+      "step": 19569
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.909447846705098e-07,
+      "loss": 0.8276,
+      "step": 19570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.905940337570755e-07,
+      "loss": 0.7896,
+      "step": 19571
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.902433652893686e-07,
+      "loss": 0.8511,
+      "step": 19572
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.898927792740972e-07,
+      "loss": 0.8018,
+      "step": 19573
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.895422757179682e-07,
+      "loss": 0.7769,
+      "step": 19574
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.891918546276877e-07,
+      "loss": 0.8569,
+      "step": 19575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.88841516009961e-07,
+      "loss": 0.8091,
+      "step": 19576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.884912598714883e-07,
+      "loss": 0.8545,
+      "step": 19577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.88141086218973e-07,
+      "loss": 0.814,
+      "step": 19578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.877909950591138e-07,
+      "loss": 0.8564,
+      "step": 19579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.87440986398607e-07,
+      "loss": 0.8174,
+      "step": 19580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.870910602441505e-07,
+      "loss": 0.8364,
+      "step": 19581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.867412166024373e-07,
+      "loss": 0.7798,
+      "step": 19582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.86391455480162e-07,
+      "loss": 0.8701,
+      "step": 19583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.860417768840161e-07,
+      "loss": 0.8296,
+      "step": 19584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.85692180820689e-07,
+      "loss": 0.8291,
+      "step": 19585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.853426672968682e-07,
+      "loss": 0.8735,
+      "step": 19586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.849932363192424e-07,
+      "loss": 0.853,
+      "step": 19587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.846438878944938e-07,
+      "loss": 0.8535,
+      "step": 19588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.842946220293095e-07,
+      "loss": 0.8359,
+      "step": 19589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.839454387303685e-07,
+      "loss": 0.7544,
+      "step": 19590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.835963380043537e-07,
+      "loss": 0.8032,
+      "step": 19591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.832473198579426e-07,
+      "loss": 0.7661,
+      "step": 19592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.828983842978115e-07,
+      "loss": 0.8945,
+      "step": 19593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.825495313306385e-07,
+      "loss": 0.8198,
+      "step": 19594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.822007609630954e-07,
+      "loss": 0.8433,
+      "step": 19595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.818520732018558e-07,
+      "loss": 0.8809,
+      "step": 19596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.815034680535915e-07,
+      "loss": 0.8394,
+      "step": 19597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.811549455249711e-07,
+      "loss": 0.8213,
+      "step": 19598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.808065056226609e-07,
+      "loss": 0.8555,
+      "step": 19599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.804581483533295e-07,
+      "loss": 0.8643,
+      "step": 19600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.801098737236395e-07,
+      "loss": 0.8745,
+      "step": 19601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.797616817402563e-07,
+      "loss": 0.8525,
+      "step": 19602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.794135724098383e-07,
+      "loss": 0.9023,
+      "step": 19603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.790655457390488e-07,
+      "loss": 0.8569,
+      "step": 19604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.787176017345438e-07,
+      "loss": 0.833,
+      "step": 19605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.783697404029799e-07,
+      "loss": 0.8511,
+      "step": 19606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.780219617510125e-07,
+      "loss": 0.8286,
+      "step": 19607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.77674265785297e-07,
+      "loss": 0.7837,
+      "step": 19608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.773266525124827e-07,
+      "loss": 0.7998,
+      "step": 19609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.769791219392224e-07,
+      "loss": 0.7773,
+      "step": 19610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.766316740721634e-07,
+      "loss": 0.8486,
+      "step": 19611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.762843089179527e-07,
+      "loss": 0.5076,
+      "step": 19612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.759370264832371e-07,
+      "loss": 0.8252,
+      "step": 19613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.755898267746597e-07,
+      "loss": 0.7988,
+      "step": 19614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.752427097988623e-07,
+      "loss": 0.8296,
+      "step": 19615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.748956755624886e-07,
+      "loss": 0.8413,
+      "step": 19616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.745487240721748e-07,
+      "loss": 0.8145,
+      "step": 19617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.742018553345626e-07,
+      "loss": 0.8887,
+      "step": 19618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.738550693562829e-07,
+      "loss": 0.8423,
+      "step": 19619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.735083661439729e-07,
+      "loss": 0.7964,
+      "step": 19620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.731617457042666e-07,
+      "loss": 0.9048,
+      "step": 19621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.728152080437939e-07,
+      "loss": 0.8394,
+      "step": 19622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.724687531691842e-07,
+      "loss": 0.8779,
+      "step": 19623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.721223810870681e-07,
+      "loss": 0.8379,
+      "step": 19624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.717760918040705e-07,
+      "loss": 0.8276,
+      "step": 19625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.71429885326817e-07,
+      "loss": 0.7905,
+      "step": 19626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.710837616619298e-07,
+      "loss": 0.8223,
+      "step": 19627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.707377208160315e-07,
+      "loss": 0.4314,
+      "step": 19628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.703917627957435e-07,
+      "loss": 0.8159,
+      "step": 19629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.700458876076832e-07,
+      "loss": 0.8535,
+      "step": 19630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.697000952584687e-07,
+      "loss": 0.8171,
+      "step": 19631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.693543857547153e-07,
+      "loss": 0.8198,
+      "step": 19632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.690087591030358e-07,
+      "loss": 0.854,
+      "step": 19633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.686632153100448e-07,
+      "loss": 0.8428,
+      "step": 19634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.683177543823505e-07,
+      "loss": 0.8726,
+      "step": 19635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.679723763265633e-07,
+      "loss": 0.7803,
+      "step": 19636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.676270811492925e-07,
+      "loss": 0.8188,
+      "step": 19637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.67281868857142e-07,
+      "loss": 0.7764,
+      "step": 19638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.669367394567172e-07,
+      "loss": 0.8594,
+      "step": 19639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.665916929546195e-07,
+      "loss": 0.8345,
+      "step": 19640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.662467293574515e-07,
+      "loss": 0.8203,
+      "step": 19641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.659018486718133e-07,
+      "loss": 0.8403,
+      "step": 19642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.655570509043019e-07,
+      "loss": 0.8086,
+      "step": 19643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.652123360615154e-07,
+      "loss": 0.8267,
+      "step": 19644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.648677041500473e-07,
+      "loss": 0.8418,
+      "step": 19645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.645231551764902e-07,
+      "loss": 0.8433,
+      "step": 19646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.641786891474383e-07,
+      "loss": 0.7983,
+      "step": 19647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.638343060694796e-07,
+      "loss": 0.8408,
+      "step": 19648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.634900059492034e-07,
+      "loss": 0.75,
+      "step": 19649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.63145788793198e-07,
+      "loss": 0.8647,
+      "step": 19650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.628016546080479e-07,
+      "loss": 0.958,
+      "step": 19651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.624576034003355e-07,
+      "loss": 0.8203,
+      "step": 19652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.621136351766449e-07,
+      "loss": 0.8423,
+      "step": 19653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.617697499435554e-07,
+      "loss": 0.8103,
+      "step": 19654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.614259477076479e-07,
+      "loss": 0.8794,
+      "step": 19655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.610822284754975e-07,
+      "loss": 0.8145,
+      "step": 19656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.60738592253683e-07,
+      "loss": 0.8394,
+      "step": 19657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.603950390487762e-07,
+      "loss": 0.7959,
+      "step": 19658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.600515688673503e-07,
+      "loss": 0.8809,
+      "step": 19659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.597081817159773e-07,
+      "loss": 0.8403,
+      "step": 19660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.593648776012257e-07,
+      "loss": 0.7974,
+      "step": 19661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.590216565296636e-07,
+      "loss": 0.8154,
+      "step": 19662
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.586785185078587e-07,
+      "loss": 0.8179,
+      "step": 19663
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.583354635423755e-07,
+      "loss": 0.8086,
+      "step": 19664
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.579924916397745e-07,
+      "loss": 0.8628,
+      "step": 19665
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.576496028066209e-07,
+      "loss": 0.8574,
+      "step": 19666
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.573067970494724e-07,
+      "loss": 0.8311,
+      "step": 19667
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.569640743748885e-07,
+      "loss": 0.8101,
+      "step": 19668
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.566214347894246e-07,
+      "loss": 0.8438,
+      "step": 19669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.562788782996382e-07,
+      "loss": 0.835,
+      "step": 19670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.55936404912082e-07,
+      "loss": 0.8306,
+      "step": 19671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.555940146333062e-07,
+      "loss": 0.8115,
+      "step": 19672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.552517074698633e-07,
+      "loss": 0.8384,
+      "step": 19673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.549094834283026e-07,
+      "loss": 0.791,
+      "step": 19674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.545673425151694e-07,
+      "loss": 0.7881,
+      "step": 19675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.542252847370111e-07,
+      "loss": 0.8164,
+      "step": 19676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.538833101003722e-07,
+      "loss": 0.9048,
+      "step": 19677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.535414186117922e-07,
+      "loss": 0.7944,
+      "step": 19678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.531996102778154e-07,
+      "loss": 0.8794,
+      "step": 19679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.528578851049789e-07,
+      "loss": 0.7783,
+      "step": 19680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.525162430998211e-07,
+      "loss": 0.8433,
+      "step": 19681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.521746842688798e-07,
+      "loss": 0.8535,
+      "step": 19682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.518332086186874e-07,
+      "loss": 0.8359,
+      "step": 19683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.514918161557781e-07,
+      "loss": 0.8213,
+      "step": 19684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.511505068866813e-07,
+      "loss": 0.8447,
+      "step": 19685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.50809280817929e-07,
+      "loss": 0.8208,
+      "step": 19686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.50468137956049e-07,
+      "loss": 0.8584,
+      "step": 19687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.501270783075663e-07,
+      "loss": 0.7988,
+      "step": 19688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.497861018790086e-07,
+      "loss": 0.8447,
+      "step": 19689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.494452086768982e-07,
+      "loss": 0.7896,
+      "step": 19690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.49104398707755e-07,
+      "loss": 0.8047,
+      "step": 19691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.487636719781021e-07,
+      "loss": 0.7983,
+      "step": 19692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.484230284944554e-07,
+      "loss": 0.8794,
+      "step": 19693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.480824682633336e-07,
+      "loss": 0.8196,
+      "step": 19694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.47741991291253e-07,
+      "loss": 0.8433,
+      "step": 19695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.47401597584727e-07,
+      "loss": 0.8569,
+      "step": 19696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.470612871502668e-07,
+      "loss": 0.874,
+      "step": 19697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.467210599943824e-07,
+      "loss": 0.8042,
+      "step": 19698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.46380916123584e-07,
+      "loss": 0.8745,
+      "step": 19699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.460408555443804e-07,
+      "loss": 0.771,
+      "step": 19700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.457008782632751e-07,
+      "loss": 0.8154,
+      "step": 19701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.453609842867741e-07,
+      "loss": 0.812,
+      "step": 19702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.450211736213802e-07,
+      "loss": 0.8169,
+      "step": 19703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.446814462735929e-07,
+      "loss": 0.8394,
+      "step": 19704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.443418022499132e-07,
+      "loss": 0.8276,
+      "step": 19705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.440022415568375e-07,
+      "loss": 0.7935,
+      "step": 19706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.436627642008631e-07,
+      "loss": 0.8081,
+      "step": 19707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.433233701884861e-07,
+      "loss": 0.8481,
+      "step": 19708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.429840595261983e-07,
+      "loss": 0.8701,
+      "step": 19709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.426448322204898e-07,
+      "loss": 0.8452,
+      "step": 19710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.423056882778533e-07,
+      "loss": 0.8213,
+      "step": 19711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.419666277047748e-07,
+      "loss": 0.8711,
+      "step": 19712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.416276505077428e-07,
+      "loss": 0.8276,
+      "step": 19713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.412887566932413e-07,
+      "loss": 0.8604,
+      "step": 19714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.409499462677554e-07,
+      "loss": 0.8667,
+      "step": 19715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.406112192377656e-07,
+      "loss": 0.8335,
+      "step": 19716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.402725756097522e-07,
+      "loss": 0.812,
+      "step": 19717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.399340153901939e-07,
+      "loss": 0.833,
+      "step": 19718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.395955385855701e-07,
+      "loss": 0.834,
+      "step": 19719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.392571452023532e-07,
+      "loss": 0.8379,
+      "step": 19720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.389188352470205e-07,
+      "loss": 0.8364,
+      "step": 19721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.385806087260421e-07,
+      "loss": 0.854,
+      "step": 19722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.382424656458885e-07,
+      "loss": 0.8301,
+      "step": 19723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.379044060130312e-07,
+      "loss": 0.8276,
+      "step": 19724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.375664298339351e-07,
+      "loss": 0.8057,
+      "step": 19725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.372285371150682e-07,
+      "loss": 0.8291,
+      "step": 19726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.368907278628939e-07,
+      "loss": 0.8276,
+      "step": 19727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.365530020838751e-07,
+      "loss": 0.8813,
+      "step": 19728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.362153597844756e-07,
+      "loss": 0.8164,
+      "step": 19729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.358778009711502e-07,
+      "loss": 0.8188,
+      "step": 19730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.355403256503595e-07,
+      "loss": 0.9028,
+      "step": 19731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.352029338285615e-07,
+      "loss": 0.7827,
+      "step": 19732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.348656255122077e-07,
+      "loss": 0.7793,
+      "step": 19733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.345284007077546e-07,
+      "loss": 0.8252,
+      "step": 19734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.341912594216509e-07,
+      "loss": 0.8213,
+      "step": 19735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.338542016603494e-07,
+      "loss": 0.7871,
+      "step": 19736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.335172274302976e-07,
+      "loss": 0.8198,
+      "step": 19737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.331803367379402e-07,
+      "loss": 0.7979,
+      "step": 19738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.328435295897245e-07,
+      "loss": 0.8306,
+      "step": 19739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.325068059920946e-07,
+      "loss": 0.8481,
+      "step": 19740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.321701659514912e-07,
+      "loss": 0.8398,
+      "step": 19741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.318336094743565e-07,
+      "loss": 0.7954,
+      "step": 19742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.314971365671279e-07,
+      "loss": 0.8574,
+      "step": 19743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.311607472362418e-07,
+      "loss": 0.9014,
+      "step": 19744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.308244414881365e-07,
+      "loss": 0.8809,
+      "step": 19745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.304882193292433e-07,
+      "loss": 0.8862,
+      "step": 19746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.301520807659961e-07,
+      "loss": 0.7837,
+      "step": 19747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.298160258048263e-07,
+      "loss": 0.7715,
+      "step": 19748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.294800544521628e-07,
+      "loss": 0.8428,
+      "step": 19749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.291441667144333e-07,
+      "loss": 0.8379,
+      "step": 19750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.288083625980618e-07,
+      "loss": 0.8008,
+      "step": 19751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.284726421094751e-07,
+      "loss": 0.8662,
+      "step": 19752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.281370052550961e-07,
+      "loss": 0.8511,
+      "step": 19753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.278014520413444e-07,
+      "loss": 0.8105,
+      "step": 19754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.274659824746416e-07,
+      "loss": 0.8057,
+      "step": 19755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.27130596561405e-07,
+      "loss": 0.8125,
+      "step": 19756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.267952943080496e-07,
+      "loss": 0.835,
+      "step": 19757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.264600757209926e-07,
+      "loss": 0.8145,
+      "step": 19758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.26124940806645e-07,
+      "loss": 0.8315,
+      "step": 19759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.257898895714198e-07,
+      "loss": 0.7698,
+      "step": 19760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.254549220217282e-07,
+      "loss": 0.873,
+      "step": 19761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.251200381639771e-07,
+      "loss": 0.7832,
+      "step": 19762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.247852380045738e-07,
+      "loss": 0.8481,
+      "step": 19763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.244505215499225e-07,
+      "loss": 0.7827,
+      "step": 19764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.241158888064275e-07,
+      "loss": 0.8799,
+      "step": 19765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.237813397804921e-07,
+      "loss": 0.8237,
+      "step": 19766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.234468744785144e-07,
+      "loss": 0.8398,
+      "step": 19767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.231124929068961e-07,
+      "loss": 0.7827,
+      "step": 19768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.227781950720329e-07,
+      "loss": 0.8926,
+      "step": 19769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.22443980980319e-07,
+      "loss": 0.8164,
+      "step": 19770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.221098506381512e-07,
+      "loss": 0.7861,
+      "step": 19771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.217758040519201e-07,
+      "loss": 0.8286,
+      "step": 19772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.214418412280166e-07,
+      "loss": 0.8149,
+      "step": 19773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.211079621728311e-07,
+      "loss": 0.8105,
+      "step": 19774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.207741668927514e-07,
+      "loss": 0.8384,
+      "step": 19775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.204404553941612e-07,
+      "loss": 0.8477,
+      "step": 19776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.20106827683447e-07,
+      "loss": 0.8037,
+      "step": 19777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.197732837669906e-07,
+      "loss": 0.8408,
+      "step": 19778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.194398236511751e-07,
+      "loss": 0.8491,
+      "step": 19779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.191064473423769e-07,
+      "loss": 0.7803,
+      "step": 19780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.187731548469772e-07,
+      "loss": 0.7749,
+      "step": 19781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.184399461713509e-07,
+      "loss": 0.8535,
+      "step": 19782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.181068213218716e-07,
+      "loss": 0.7666,
+      "step": 19783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.177737803049139e-07,
+      "loss": 0.8726,
+      "step": 19784
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.174408231268509e-07,
+      "loss": 0.8867,
+      "step": 19785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.17107949794049e-07,
+      "loss": 0.8569,
+      "step": 19786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.167751603128802e-07,
+      "loss": 0.8364,
+      "step": 19787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.164424546897102e-07,
+      "loss": 0.6912,
+      "step": 19788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.161098329309018e-07,
+      "loss": 0.8062,
+      "step": 19789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.15777295042822e-07,
+      "loss": 0.8491,
+      "step": 19790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.154448410318298e-07,
+      "loss": 0.7725,
+      "step": 19791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.151124709042872e-07,
+      "loss": 0.8774,
+      "step": 19792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.147801846665536e-07,
+      "loss": 0.8115,
+      "step": 19793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.144479823249855e-07,
+      "loss": 0.855,
+      "step": 19794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.141158638859385e-07,
+      "loss": 0.8242,
+      "step": 19795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.137838293557652e-07,
+      "loss": 0.8477,
+      "step": 19796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.134518787408189e-07,
+      "loss": 0.8691,
+      "step": 19797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.131200120474512e-07,
+      "loss": 0.8584,
+      "step": 19798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.127882292820097e-07,
+      "loss": 0.8027,
+      "step": 19799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.12456530450844e-07,
+      "loss": 0.8018,
+      "step": 19800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.121249155602987e-07,
+      "loss": 0.8105,
+      "step": 19801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.117933846167173e-07,
+      "loss": 0.8286,
+      "step": 19802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.114619376264442e-07,
+      "loss": 0.8223,
+      "step": 19803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.111305745958191e-07,
+      "loss": 0.812,
+      "step": 19804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.107992955311815e-07,
+      "loss": 0.7979,
+      "step": 19805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.104681004388713e-07,
+      "loss": 0.3254,
+      "step": 19806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.101369893252235e-07,
+      "loss": 0.8198,
+      "step": 19807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.098059621965724e-07,
+      "loss": 0.8018,
+      "step": 19808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.094750190592502e-07,
+      "loss": 0.8638,
+      "step": 19809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.091441599195896e-07,
+      "loss": 0.876,
+      "step": 19810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.088133847839217e-07,
+      "loss": 0.8003,
+      "step": 19811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.084826936585719e-07,
+      "loss": 0.8291,
+      "step": 19812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.081520865498696e-07,
+      "loss": 0.8652,
+      "step": 19813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.078215634641388e-07,
+      "loss": 0.8315,
+      "step": 19814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.074911244077014e-07,
+      "loss": 0.8545,
+      "step": 19815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.071607693868814e-07,
+      "loss": 0.7935,
+      "step": 19816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.06830498407997e-07,
+      "loss": 0.8149,
+      "step": 19817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.065003114773682e-07,
+      "loss": 0.7979,
+      "step": 19818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.061702086013133e-07,
+      "loss": 0.7842,
+      "step": 19819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.058401897861454e-07,
+      "loss": 0.8159,
+      "step": 19820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.055102550381781e-07,
+      "loss": 0.8623,
+      "step": 19821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.051804043637255e-07,
+      "loss": 0.8926,
+      "step": 19822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.048506377690966e-07,
+      "loss": 0.8296,
+      "step": 19823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.045209552606013e-07,
+      "loss": 0.8564,
+      "step": 19824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.041913568445457e-07,
+      "loss": 0.2897,
+      "step": 19825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.038618425272375e-07,
+      "loss": 0.8389,
+      "step": 19826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.035324123149794e-07,
+      "loss": 0.8267,
+      "step": 19827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.032030662140736e-07,
+      "loss": 0.811,
+      "step": 19828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.028738042308224e-07,
+      "loss": 0.402,
+      "step": 19829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.025446263715234e-07,
+      "loss": 0.7822,
+      "step": 19830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.022155326424756e-07,
+      "loss": 0.812,
+      "step": 19831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.018865230499749e-07,
+      "loss": 0.8413,
+      "step": 19832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.015575976003163e-07,
+      "loss": 0.8618,
+      "step": 19833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.012287562997903e-07,
+      "loss": 0.8081,
+      "step": 19834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.008999991546915e-07,
+      "loss": 0.8149,
+      "step": 19835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.005713261713065e-07,
+      "loss": 0.8584,
+      "step": 19836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.002427373559256e-07,
+      "loss": 0.7856,
+      "step": 19837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.999142327148333e-07,
+      "loss": 0.7715,
+      "step": 19838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.995858122543158e-07,
+      "loss": 0.8999,
+      "step": 19839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.992574759806574e-07,
+      "loss": 0.8696,
+      "step": 19840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.989292239001365e-07,
+      "loss": 0.833,
+      "step": 19841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.986010560190342e-07,
+      "loss": 0.8335,
+      "step": 19842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.98272972343631e-07,
+      "loss": 0.8066,
+      "step": 19843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.979449728802012e-07,
+      "loss": 0.854,
+      "step": 19844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.976170576350215e-07,
+      "loss": 0.8208,
+      "step": 19845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.972892266143638e-07,
+      "loss": 0.7827,
+      "step": 19846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.969614798245022e-07,
+      "loss": 0.8291,
+      "step": 19847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.966338172717056e-07,
+      "loss": 0.8101,
+      "step": 19848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.96306238962242e-07,
+      "loss": 0.8022,
+      "step": 19849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.959787449023796e-07,
+      "loss": 0.8843,
+      "step": 19850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.956513350983845e-07,
+      "loss": 0.8403,
+      "step": 19851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.953240095565189e-07,
+      "loss": 0.8672,
+      "step": 19852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.949967682830465e-07,
+      "loss": 0.8701,
+      "step": 19853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.946696112842276e-07,
+      "loss": 0.7798,
+      "step": 19854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.943425385663204e-07,
+      "loss": 0.385,
+      "step": 19855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.940155501355832e-07,
+      "loss": 0.8472,
+      "step": 19856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.93688645998271e-07,
+      "loss": 0.75,
+      "step": 19857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.933618261606383e-07,
+      "loss": 0.8159,
+      "step": 19858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.930350906289384e-07,
+      "loss": 0.8706,
+      "step": 19859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.927084394094223e-07,
+      "loss": 0.8301,
+      "step": 19860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.923818725083386e-07,
+      "loss": 0.9004,
+      "step": 19861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.920553899319342e-07,
+      "loss": 0.7578,
+      "step": 19862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.917289916864566e-07,
+      "loss": 0.8677,
+      "step": 19863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.914026777781507e-07,
+      "loss": 0.739,
+      "step": 19864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.910764482132575e-07,
+      "loss": 0.8076,
+      "step": 19865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.907503029980205e-07,
+      "loss": 0.8291,
+      "step": 19866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.904242421386785e-07,
+      "loss": 0.8647,
+      "step": 19867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.900982656414684e-07,
+      "loss": 0.832,
+      "step": 19868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.897723735126281e-07,
+      "loss": 0.8159,
+      "step": 19869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.894465657583914e-07,
+      "loss": 0.8281,
+      "step": 19870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.89120842384992e-07,
+      "loss": 0.8018,
+      "step": 19871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.887952033986627e-07,
+      "loss": 0.8682,
+      "step": 19872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.884696488056324e-07,
+      "loss": 0.8384,
+      "step": 19873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.881441786121289e-07,
+      "loss": 0.3639,
+      "step": 19874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.878187928243789e-07,
+      "loss": 0.7637,
+      "step": 19875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.87493491448608e-07,
+      "loss": 0.7466,
+      "step": 19876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.871682744910412e-07,
+      "loss": 0.8149,
+      "step": 19877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.86843141957898e-07,
+      "loss": 0.7798,
+      "step": 19878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.865180938554005e-07,
+      "loss": 0.9355,
+      "step": 19879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.861931301897667e-07,
+      "loss": 0.8198,
+      "step": 19880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.858682509672125e-07,
+      "loss": 0.8833,
+      "step": 19881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.855434561939554e-07,
+      "loss": 0.854,
+      "step": 19882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.852187458762077e-07,
+      "loss": 0.7666,
+      "step": 19883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.848941200201819e-07,
+      "loss": 0.8252,
+      "step": 19884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.845695786320898e-07,
+      "loss": 0.8257,
+      "step": 19885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8424512171814e-07,
+      "loss": 0.8462,
+      "step": 19886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.839207492845378e-07,
+      "loss": 0.8115,
+      "step": 19887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.835964613374922e-07,
+      "loss": 0.8735,
+      "step": 19888
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.832722578832039e-07,
+      "loss": 0.8086,
+      "step": 19889
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.82948138927879e-07,
+      "loss": 0.8013,
+      "step": 19890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.826241044777154e-07,
+      "loss": 0.8687,
+      "step": 19891
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.823001545389145e-07,
+      "loss": 0.8755,
+      "step": 19892
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.819762891176727e-07,
+      "loss": 0.8774,
+      "step": 19893
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.816525082201858e-07,
+      "loss": 0.8916,
+      "step": 19894
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8132881185265e-07,
+      "loss": 0.7861,
+      "step": 19895
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.810052000212557e-07,
+      "loss": 0.8848,
+      "step": 19896
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.806816727321956e-07,
+      "loss": 0.8071,
+      "step": 19897
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.803582299916599e-07,
+      "loss": 0.8481,
+      "step": 19898
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.800348718058357e-07,
+      "loss": 0.7993,
+      "step": 19899
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.797115981809087e-07,
+      "loss": 0.3293,
+      "step": 19900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.79388409123065e-07,
+      "loss": 0.8115,
+      "step": 19901
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.790653046384865e-07,
+      "loss": 0.8208,
+      "step": 19902
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.787422847333557e-07,
+      "loss": 0.7537,
+      "step": 19903
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.784193494138512e-07,
+      "loss": 0.8115,
+      "step": 19904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.780964986861537e-07,
+      "loss": 0.8516,
+      "step": 19905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.777737325564375e-07,
+      "loss": 0.8608,
+      "step": 19906
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.774510510308778e-07,
+      "loss": 0.8374,
+      "step": 19907
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.77128454115648e-07,
+      "loss": 0.8403,
+      "step": 19908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.76805941816922e-07,
+      "loss": 0.7993,
+      "step": 19909
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.76483514140867e-07,
+      "loss": 0.7837,
+      "step": 19910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.761611710936544e-07,
+      "loss": 0.8301,
+      "step": 19911
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.758389126814495e-07,
+      "loss": 0.7988,
+      "step": 19912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.755167389104166e-07,
+      "loss": 0.8579,
+      "step": 19913
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.751946497867223e-07,
+      "loss": 0.7988,
+      "step": 19914
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.748726453165254e-07,
+      "loss": 0.8032,
+      "step": 19915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.745507255059879e-07,
+      "loss": 0.9082,
+      "step": 19916
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.742288903612697e-07,
+      "loss": 0.8418,
+      "step": 19917
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.739071398885271e-07,
+      "loss": 0.8374,
+      "step": 19918
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.735854740939157e-07,
+      "loss": 0.8628,
+      "step": 19919
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.732638929835882e-07,
+      "loss": 0.8354,
+      "step": 19920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.729423965636977e-07,
+      "loss": 0.8813,
+      "step": 19921
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.726209848403969e-07,
+      "loss": 0.8896,
+      "step": 19922
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.72299657819832e-07,
+      "loss": 0.8091,
+      "step": 19923
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.719784155081531e-07,
+      "loss": 0.8789,
+      "step": 19924
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.716572579115049e-07,
+      "loss": 0.8511,
+      "step": 19925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.713361850360299e-07,
+      "loss": 0.7769,
+      "step": 19926
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.710151968878741e-07,
+      "loss": 0.8516,
+      "step": 19927
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.706942934731752e-07,
+      "loss": 0.8008,
+      "step": 19928
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.703734747980743e-07,
+      "loss": 0.8091,
+      "step": 19929
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.700527408687107e-07,
+      "loss": 0.8081,
+      "step": 19930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.697320916912186e-07,
+      "loss": 0.8003,
+      "step": 19931
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.694115272717326e-07,
+      "loss": 0.854,
+      "step": 19932
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.690910476163847e-07,
+      "loss": 0.79,
+      "step": 19933
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.687706527313075e-07,
+      "loss": 0.8076,
+      "step": 19934
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.684503426226312e-07,
+      "loss": 0.7974,
+      "step": 19935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.681301172964826e-07,
+      "loss": 0.8301,
+      "step": 19936
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.678099767589896e-07,
+      "loss": 0.7954,
+      "step": 19937
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.674899210162755e-07,
+      "loss": 0.833,
+      "step": 19938
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.671699500744632e-07,
+      "loss": 0.8022,
+      "step": 19939
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.668500639396763e-07,
+      "loss": 0.8105,
+      "step": 19940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.665302626180318e-07,
+      "loss": 0.2659,
+      "step": 19941
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.662105461156503e-07,
+      "loss": 0.7944,
+      "step": 19942
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.658909144386487e-07,
+      "loss": 0.7886,
+      "step": 19943
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.655713675931412e-07,
+      "loss": 0.8208,
+      "step": 19944
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.652519055852401e-07,
+      "loss": 0.9004,
+      "step": 19945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.649325284210594e-07,
+      "loss": 0.7803,
+      "step": 19946
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.646132361067069e-07,
+      "loss": 0.8203,
+      "step": 19947
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.64294028648294e-07,
+      "loss": 0.8511,
+      "step": 19948
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.639749060519251e-07,
+      "loss": 0.8037,
+      "step": 19949
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.636558683237065e-07,
+      "loss": 0.9106,
+      "step": 19950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.633369154697438e-07,
+      "loss": 0.8276,
+      "step": 19951
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.63018047496135e-07,
+      "loss": 0.8418,
+      "step": 19952
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.626992644089829e-07,
+      "loss": 0.814,
+      "step": 19953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.623805662143878e-07,
+      "loss": 0.3328,
+      "step": 19954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.620619529184435e-07,
+      "loss": 0.8784,
+      "step": 19955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.617434245272485e-07,
+      "loss": 0.9312,
+      "step": 19956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.614249810468953e-07,
+      "loss": 0.7705,
+      "step": 19957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.611066224834766e-07,
+      "loss": 0.854,
+      "step": 19958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.607883488430837e-07,
+      "loss": 0.833,
+      "step": 19959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.604701601318036e-07,
+      "loss": 0.897,
+      "step": 19960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.601520563557266e-07,
+      "loss": 0.8057,
+      "step": 19961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.598340375209355e-07,
+      "loss": 0.8306,
+      "step": 19962
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.595161036335168e-07,
+      "loss": 0.8618,
+      "step": 19963
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.591982546995528e-07,
+      "loss": 0.8247,
+      "step": 19964
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.588804907251244e-07,
+      "loss": 0.8618,
+      "step": 19965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.585628117163089e-07,
+      "loss": 0.8374,
+      "step": 19966
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.582452176791875e-07,
+      "loss": 0.8027,
+      "step": 19967
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.579277086198326e-07,
+      "loss": 0.8188,
+      "step": 19968
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.576102845443221e-07,
+      "loss": 0.3584,
+      "step": 19969
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.572929454587261e-07,
+      "loss": 0.8091,
+      "step": 19970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.569756913691176e-07,
+      "loss": 0.8584,
+      "step": 19971
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.566585222815657e-07,
+      "loss": 0.8408,
+      "step": 19972
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.56341438202137e-07,
+      "loss": 0.3719,
+      "step": 19973
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.56024439136899e-07,
+      "loss": 0.7603,
+      "step": 19974
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.557075250919175e-07,
+      "loss": 0.8838,
+      "step": 19975
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.55390696073253e-07,
+      "loss": 0.834,
+      "step": 19976
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.550739520869697e-07,
+      "loss": 0.8286,
+      "step": 19977
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.547572931391254e-07,
+      "loss": 0.877,
+      "step": 19978
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.544407192357787e-07,
+      "loss": 0.8418,
+      "step": 19979
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.541242303829869e-07,
+      "loss": 0.8267,
+      "step": 19980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.538078265868036e-07,
+      "loss": 0.7969,
+      "step": 19981
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.534915078532826e-07,
+      "loss": 0.8237,
+      "step": 19982
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.531752741884772e-07,
+      "loss": 0.8369,
+      "step": 19983
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.528591255984355e-07,
+      "loss": 0.8521,
+      "step": 19984
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.525430620892075e-07,
+      "loss": 0.7739,
+      "step": 19985
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.522270836668376e-07,
+      "loss": 0.8003,
+      "step": 19986
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.519111903373725e-07,
+      "loss": 0.8594,
+      "step": 19987
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.515953821068565e-07,
+      "loss": 0.8589,
+      "step": 19988
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.512796589813296e-07,
+      "loss": 0.7583,
+      "step": 19989
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.509640209668344e-07,
+      "loss": 0.8765,
+      "step": 19990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.50648468069408e-07,
+      "loss": 0.8589,
+      "step": 19991
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.503330002950868e-07,
+      "loss": 0.8926,
+      "step": 19992
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.500176176499084e-07,
+      "loss": 0.8467,
+      "step": 19993
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.497023201399037e-07,
+      "loss": 0.7622,
+      "step": 19994
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.493871077711066e-07,
+      "loss": 0.832,
+      "step": 19995
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.490719805495487e-07,
+      "loss": 0.8354,
+      "step": 19996
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.487569384812574e-07,
+      "loss": 0.8906,
+      "step": 19997
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.484419815722603e-07,
+      "loss": 0.8545,
+      "step": 19998
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.481271098285818e-07,
+      "loss": 0.8057,
+      "step": 19999
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.478123232562471e-07,
+      "loss": 0.8555,
+      "step": 20000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.474976218612793e-07,
+      "loss": 0.854,
+      "step": 20001
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.471830056496968e-07,
+      "loss": 0.8418,
+      "step": 20002
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.468684746275221e-07,
+      "loss": 0.8613,
+      "step": 20003
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.465540288007703e-07,
+      "loss": 0.8975,
+      "step": 20004
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.462396681754567e-07,
+      "loss": 0.8267,
+      "step": 20005
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.459253927575975e-07,
+      "loss": 0.8179,
+      "step": 20006
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.456112025532029e-07,
+      "loss": 0.8071,
+      "step": 20007
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.452970975682859e-07,
+      "loss": 0.8364,
+      "step": 20008
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.449830778088555e-07,
+      "loss": 0.8726,
+      "step": 20009
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.446691432809193e-07,
+      "loss": 0.876,
+      "step": 20010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.443552939904823e-07,
+      "loss": 0.8599,
+      "step": 20011
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.440415299435509e-07,
+      "loss": 0.8506,
+      "step": 20012
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.437278511461253e-07,
+      "loss": 0.8481,
+      "step": 20013
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.434142576042096e-07,
+      "loss": 0.7671,
+      "step": 20014
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.431007493238005e-07,
+      "loss": 0.9077,
+      "step": 20015
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.427873263108985e-07,
+      "loss": 0.8618,
+      "step": 20016
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.42473988571498e-07,
+      "loss": 0.8364,
+      "step": 20017
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.421607361115938e-07,
+      "loss": 0.8618,
+      "step": 20018
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.418475689371794e-07,
+      "loss": 0.7954,
+      "step": 20019
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.415344870542471e-07,
+      "loss": 0.8438,
+      "step": 20020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.412214904687851e-07,
+      "loss": 0.8491,
+      "step": 20021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.40908579186783e-07,
+      "loss": 0.8271,
+      "step": 20022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.405957532142264e-07,
+      "loss": 0.811,
+      "step": 20023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.40283012557099e-07,
+      "loss": 0.8135,
+      "step": 20024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.39970357221386e-07,
+      "loss": 0.8159,
+      "step": 20025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.396577872130676e-07,
+      "loss": 0.3433,
+      "step": 20026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.39345302538124e-07,
+      "loss": 0.8193,
+      "step": 20027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.390329032025354e-07,
+      "loss": 0.8716,
+      "step": 20028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.387205892122766e-07,
+      "loss": 0.8174,
+      "step": 20029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.384083605733231e-07,
+      "loss": 0.814,
+      "step": 20030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.38096217291647e-07,
+      "loss": 0.8359,
+      "step": 20031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.37784159373222e-07,
+      "loss": 0.7891,
+      "step": 20032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.374721868240179e-07,
+      "loss": 0.8057,
+      "step": 20033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.371602996500025e-07,
+      "loss": 0.8516,
+      "step": 20034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.368484978571437e-07,
+      "loss": 0.793,
+      "step": 20035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.365367814514066e-07,
+      "loss": 0.9277,
+      "step": 20036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.362251504387533e-07,
+      "loss": 0.8345,
+      "step": 20037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.359136048251479e-07,
+      "loss": 0.8413,
+      "step": 20038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.356021446165482e-07,
+      "loss": 0.8569,
+      "step": 20039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.352907698189152e-07,
+      "loss": 0.8511,
+      "step": 20040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.349794804382063e-07,
+      "loss": 0.8208,
+      "step": 20041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.346682764803756e-07,
+      "loss": 0.8101,
+      "step": 20042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.343571579513773e-07,
+      "loss": 0.8198,
+      "step": 20043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.340461248571626e-07,
+      "loss": 0.8735,
+      "step": 20044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.33735177203683e-07,
+      "loss": 0.8496,
+      "step": 20045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.334243149968882e-07,
+      "loss": 0.8184,
+      "step": 20046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.331135382427239e-07,
+      "loss": 0.8496,
+      "step": 20047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.328028469471375e-07,
+      "loss": 0.7798,
+      "step": 20048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.324922411160721e-07,
+      "loss": 0.811,
+      "step": 20049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.321817207554691e-07,
+      "loss": 0.7998,
+      "step": 20050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.31871285871271e-07,
+      "loss": 0.8179,
+      "step": 20051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.315609364694158e-07,
+      "loss": 0.8481,
+      "step": 20052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.312506725558403e-07,
+      "loss": 0.8159,
+      "step": 20053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.309404941364826e-07,
+      "loss": 0.8032,
+      "step": 20054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.306304012172758e-07,
+      "loss": 0.793,
+      "step": 20055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.303203938041513e-07,
+      "loss": 0.8267,
+      "step": 20056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.300104719030419e-07,
+      "loss": 0.8125,
+      "step": 20057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.297006355198753e-07,
+      "loss": 0.7949,
+      "step": 20058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.293908846605805e-07,
+      "loss": 0.749,
+      "step": 20059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.290812193310818e-07,
+      "loss": 0.8379,
+      "step": 20060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.287716395373049e-07,
+      "loss": 0.8257,
+      "step": 20061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.284621452851741e-07,
+      "loss": 0.7896,
+      "step": 20062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.281527365806072e-07,
+      "loss": 0.8374,
+      "step": 20063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.278434134295257e-07,
+      "loss": 0.8506,
+      "step": 20064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.27534175837846e-07,
+      "loss": 0.8433,
+      "step": 20065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.272250238114857e-07,
+      "loss": 0.8545,
+      "step": 20066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.269159573563592e-07,
+      "loss": 0.8228,
+      "step": 20067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.266069764783788e-07,
+      "loss": 0.8389,
+      "step": 20068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.262980811834567e-07,
+      "loss": 0.8037,
+      "step": 20069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.259892714775022e-07,
+      "loss": 0.8564,
+      "step": 20070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.256805473664217e-07,
+      "loss": 0.8413,
+      "step": 20071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.253719088561243e-07,
+      "loss": 0.7959,
+      "step": 20072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.250633559525126e-07,
+      "loss": 0.8564,
+      "step": 20073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.247548886614906e-07,
+      "loss": 0.7676,
+      "step": 20074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.244465069889604e-07,
+      "loss": 0.8037,
+      "step": 20075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.241382109408211e-07,
+      "loss": 0.8467,
+      "step": 20076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.238300005229702e-07,
+      "loss": 0.876,
+      "step": 20077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.235218757413052e-07,
+      "loss": 0.8364,
+      "step": 20078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.232138366017203e-07,
+      "loss": 0.7798,
+      "step": 20079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.229058831101102e-07,
+      "loss": 0.7695,
+      "step": 20080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.225980152723648e-07,
+      "loss": 0.896,
+      "step": 20081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.222902330943758e-07,
+      "loss": 0.8726,
+      "step": 20082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.219825365820302e-07,
+      "loss": 0.8682,
+      "step": 20083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.216749257412146e-07,
+      "loss": 0.8379,
+      "step": 20084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.213674005778146e-07,
+      "loss": 0.7715,
+      "step": 20085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.210599610977146e-07,
+      "loss": 0.8516,
+      "step": 20086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.207526073067948e-07,
+      "loss": 0.8569,
+      "step": 20087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.204453392109366e-07,
+      "loss": 0.8228,
+      "step": 20088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.20138156816018e-07,
+      "loss": 0.8433,
+      "step": 20089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.19831060127915e-07,
+      "loss": 0.8706,
+      "step": 20090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.195240491525044e-07,
+      "loss": 0.8511,
+      "step": 20091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.192171238956584e-07,
+      "loss": 0.7925,
+      "step": 20092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.189102843632493e-07,
+      "loss": 0.853,
+      "step": 20093
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.186035305611492e-07,
+      "loss": 0.8418,
+      "step": 20094
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.18296862495225e-07,
+      "loss": 0.8271,
+      "step": 20095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.179902801713443e-07,
+      "loss": 0.8643,
+      "step": 20096
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.176837835953713e-07,
+      "loss": 0.8091,
+      "step": 20097
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.173773727731707e-07,
+      "loss": 0.8677,
+      "step": 20098
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.170710477106061e-07,
+      "loss": 0.8457,
+      "step": 20099
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.167648084135351e-07,
+      "loss": 0.8569,
+      "step": 20100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.164586548878192e-07,
+      "loss": 0.9136,
+      "step": 20101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.161525871393141e-07,
+      "loss": 0.7722,
+      "step": 20102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.158466051738753e-07,
+      "loss": 0.8638,
+      "step": 20103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.155407089973574e-07,
+      "loss": 0.8418,
+      "step": 20104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.152348986156119e-07,
+      "loss": 0.8403,
+      "step": 20105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.149291740344897e-07,
+      "loss": 0.8701,
+      "step": 20106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.146235352598411e-07,
+      "loss": 0.8232,
+      "step": 20107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.14317982297512e-07,
+      "loss": 0.8774,
+      "step": 20108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.140125151533487e-07,
+      "loss": 0.8291,
+      "step": 20109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.137071338331939e-07,
+      "loss": 0.853,
+      "step": 20110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.134018383428912e-07,
+      "loss": 0.7686,
+      "step": 20111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.130966286882827e-07,
+      "loss": 0.792,
+      "step": 20112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.127915048752047e-07,
+      "loss": 0.8179,
+      "step": 20113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.124864669094976e-07,
+      "loss": 0.8145,
+      "step": 20114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.121815147969955e-07,
+      "loss": 0.7871,
+      "step": 20115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.118766485435323e-07,
+      "loss": 0.8579,
+      "step": 20116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.115718681549414e-07,
+      "loss": 0.8081,
+      "step": 20117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.112671736370534e-07,
+      "loss": 0.7817,
+      "step": 20118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.109625649956973e-07,
+      "loss": 0.8428,
+      "step": 20119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.106580422367025e-07,
+      "loss": 0.8208,
+      "step": 20120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.103536053658931e-07,
+      "loss": 0.8408,
+      "step": 20121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.100492543890934e-07,
+      "loss": 0.8496,
+      "step": 20122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.097449893121281e-07,
+      "loss": 0.8315,
+      "step": 20123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.094408101408155e-07,
+      "loss": 0.7788,
+      "step": 20124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.091367168809774e-07,
+      "loss": 0.8203,
+      "step": 20125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.088327095384299e-07,
+      "loss": 0.8657,
+      "step": 20126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.085287881189905e-07,
+      "loss": 0.8252,
+      "step": 20127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.082249526284733e-07,
+      "loss": 0.8008,
+      "step": 20128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.079212030726899e-07,
+      "loss": 0.8765,
+      "step": 20129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.076175394574534e-07,
+      "loss": 0.8022,
+      "step": 20130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.073139617885714e-07,
+      "loss": 0.8784,
+      "step": 20131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.070104700718531e-07,
+      "loss": 0.771,
+      "step": 20132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.067070643131056e-07,
+      "loss": 0.8535,
+      "step": 20133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.064037445181324e-07,
+      "loss": 0.7788,
+      "step": 20134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.061005106927358e-07,
+      "loss": 0.877,
+      "step": 20135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.057973628427187e-07,
+      "loss": 0.7793,
+      "step": 20136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.054943009738795e-07,
+      "loss": 0.8999,
+      "step": 20137
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.05191325092017e-07,
+      "loss": 0.8291,
+      "step": 20138
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.048884352029271e-07,
+      "loss": 0.8545,
+      "step": 20139
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.045856313124053e-07,
+      "loss": 0.8623,
+      "step": 20140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.042829134262445e-07,
+      "loss": 0.8574,
+      "step": 20141
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.039802815502354e-07,
+      "loss": 0.8389,
+      "step": 20142
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.036777356901679e-07,
+      "loss": 0.8398,
+      "step": 20143
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.033752758518312e-07,
+      "loss": 0.8359,
+      "step": 20144
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.030729020410108e-07,
+      "loss": 0.791,
+      "step": 20145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.027706142634931e-07,
+      "loss": 0.8306,
+      "step": 20146
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.024684125250601e-07,
+      "loss": 0.8779,
+      "step": 20147
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.021662968314927e-07,
+      "loss": 0.3163,
+      "step": 20148
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.018642671885726e-07,
+      "loss": 0.8115,
+      "step": 20149
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.015623236020766e-07,
+      "loss": 0.8599,
+      "step": 20150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.01260466077782e-07,
+      "loss": 0.7495,
+      "step": 20151
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.009586946214645e-07,
+      "loss": 0.7905,
+      "step": 20152
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.006570092388968e-07,
+      "loss": 0.3496,
+      "step": 20153
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.003554099358504e-07,
+      "loss": 0.8564,
+      "step": 20154
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.000538967180951e-07,
+      "loss": 0.8394,
+      "step": 20155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.997524695913997e-07,
+      "loss": 0.8628,
+      "step": 20156
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.994511285615317e-07,
+      "loss": 0.8213,
+      "step": 20157
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.991498736342549e-07,
+      "loss": 0.8608,
+      "step": 20158
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.98848704815334e-07,
+      "loss": 0.834,
+      "step": 20159
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.985476221105306e-07,
+      "loss": 0.8657,
+      "step": 20160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.98246625525604e-07,
+      "loss": 0.876,
+      "step": 20161
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.979457150663142e-07,
+      "loss": 0.832,
+      "step": 20162
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.976448907384157e-07,
+      "loss": 0.3875,
+      "step": 20163
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.973441525476657e-07,
+      "loss": 0.897,
+      "step": 20164
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.97043500499818e-07,
+      "loss": 0.8179,
+      "step": 20165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.967429346006242e-07,
+      "loss": 0.7896,
+      "step": 20166
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.964424548558344e-07,
+      "loss": 0.7803,
+      "step": 20167
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.961420612711959e-07,
+      "loss": 0.8716,
+      "step": 20168
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.958417538524574e-07,
+      "loss": 0.8154,
+      "step": 20169
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.955415326053648e-07,
+      "loss": 0.8252,
+      "step": 20170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.952413975356596e-07,
+      "loss": 0.8257,
+      "step": 20171
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.94941348649085e-07,
+      "loss": 0.8457,
+      "step": 20172
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.946413859513833e-07,
+      "loss": 0.8467,
+      "step": 20173
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.943415094482895e-07,
+      "loss": 0.978,
+      "step": 20174
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.940417191455437e-07,
+      "loss": 0.793,
+      "step": 20175
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.937420150488797e-07,
+      "loss": 0.8501,
+      "step": 20176
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.934423971640317e-07,
+      "loss": 0.8999,
+      "step": 20177
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.93142865496733e-07,
+      "loss": 0.875,
+      "step": 20178
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.928434200527127e-07,
+      "loss": 0.7964,
+      "step": 20179
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.925440608377008e-07,
+      "loss": 0.8149,
+      "step": 20180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.922447878574238e-07,
+      "loss": 0.8311,
+      "step": 20181
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.91945601117606e-07,
+      "loss": 0.8257,
+      "step": 20182
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.91646500623974e-07,
+      "loss": 0.8174,
+      "step": 20183
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.913474863822481e-07,
+      "loss": 0.875,
+      "step": 20184
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.910485583981489e-07,
+      "loss": 0.8267,
+      "step": 20185
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.907497166773978e-07,
+      "loss": 0.3419,
+      "step": 20186
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.904509612257097e-07,
+      "loss": 0.8057,
+      "step": 20187
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.901522920488e-07,
+      "loss": 0.8159,
+      "step": 20188
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.898537091523847e-07,
+      "loss": 0.8296,
+      "step": 20189
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.895552125421737e-07,
+      "loss": 0.8594,
+      "step": 20190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.892568022238803e-07,
+      "loss": 0.834,
+      "step": 20191
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.889584782032114e-07,
+      "loss": 0.8765,
+      "step": 20192
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.886602404858765e-07,
+      "loss": 0.8198,
+      "step": 20193
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.883620890775797e-07,
+      "loss": 0.7454,
+      "step": 20194
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.88064023984025e-07,
+      "loss": 0.8467,
+      "step": 20195
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.877660452109162e-07,
+      "loss": 0.8135,
+      "step": 20196
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.874681527639524e-07,
+      "loss": 0.835,
+      "step": 20197
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.871703466488337e-07,
+      "loss": 0.875,
+      "step": 20198
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.868726268712587e-07,
+      "loss": 0.8965,
+      "step": 20199
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.865749934369224e-07,
+      "loss": 0.8125,
+      "step": 20200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.862774463515174e-07,
+      "loss": 0.8076,
+      "step": 20201
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.859799856207387e-07,
+      "loss": 0.8276,
+      "step": 20202
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.856826112502749e-07,
+      "loss": 0.8765,
+      "step": 20203
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.853853232458172e-07,
+      "loss": 0.7979,
+      "step": 20204
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.850881216130521e-07,
+      "loss": 0.8188,
+      "step": 20205
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.847910063576666e-07,
+      "loss": 0.8379,
+      "step": 20206
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.844939774853441e-07,
+      "loss": 0.8423,
+      "step": 20207
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.841970350017666e-07,
+      "loss": 0.8462,
+      "step": 20208
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.839001789126152e-07,
+      "loss": 0.8667,
+      "step": 20209
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.836034092235714e-07,
+      "loss": 0.814,
+      "step": 20210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.833067259403107e-07,
+      "loss": 0.793,
+      "step": 20211
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.830101290685102e-07,
+      "loss": 0.8052,
+      "step": 20212
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.827136186138437e-07,
+      "loss": 0.8525,
+      "step": 20213
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.824171945819834e-07,
+      "loss": 0.8799,
+      "step": 20214
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.821208569786018e-07,
+      "loss": 0.7622,
+      "step": 20215
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.818246058093668e-07,
+      "loss": 0.8335,
+      "step": 20216
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.815284410799465e-07,
+      "loss": 0.874,
+      "step": 20217
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.812323627960085e-07,
+      "loss": 0.8203,
+      "step": 20218
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.809363709632158e-07,
+      "loss": 0.8276,
+      "step": 20219
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.806404655872321e-07,
+      "loss": 0.8408,
+      "step": 20220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.803446466737166e-07,
+      "loss": 0.8115,
+      "step": 20221
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.800489142283298e-07,
+      "loss": 0.8247,
+      "step": 20222
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797532682567313e-07,
+      "loss": 0.8262,
+      "step": 20223
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.794577087645746e-07,
+      "loss": 0.8276,
+      "step": 20224
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.791622357575165e-07,
+      "loss": 0.834,
+      "step": 20225
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.788668492412085e-07,
+      "loss": 0.8384,
+      "step": 20226
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.785715492213011e-07,
+      "loss": 0.8022,
+      "step": 20227
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.782763357034464e-07,
+      "loss": 0.813,
+      "step": 20228
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.779812086932894e-07,
+      "loss": 0.793,
+      "step": 20229
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.776861681964772e-07,
+      "loss": 0.8193,
+      "step": 20230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.773912142186566e-07,
+      "loss": 0.7993,
+      "step": 20231
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.770963467654688e-07,
+      "loss": 0.7778,
+      "step": 20232
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7680156584255457e-07,
+      "loss": 0.7783,
+      "step": 20233
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7650687145555375e-07,
+      "loss": 0.8149,
+      "step": 20234
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7621226361010407e-07,
+      "loss": 0.8184,
+      "step": 20235
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.759177423118433e-07,
+      "loss": 0.7334,
+      "step": 20236
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7562330756640463e-07,
+      "loss": 0.7495,
+      "step": 20237
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7532895937942247e-07,
+      "loss": 0.8149,
+      "step": 20238
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7503469775652735e-07,
+      "loss": 0.8184,
+      "step": 20239
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.74740522703348e-07,
+      "loss": 0.7954,
+      "step": 20240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7444643422551394e-07,
+      "loss": 0.8389,
+      "step": 20241
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7415243232865003e-07,
+      "loss": 0.79,
+      "step": 20242
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7385851701838237e-07,
+      "loss": 0.8335,
+      "step": 20243
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.735646883003348e-07,
+      "loss": 0.853,
+      "step": 20244
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.732709461801266e-07,
+      "loss": 0.8589,
+      "step": 20245
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7297729066337837e-07,
+      "loss": 0.8389,
+      "step": 20246
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.726837217557084e-07,
+      "loss": 0.8208,
+      "step": 20247
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7239023946273275e-07,
+      "loss": 0.8384,
+      "step": 20248
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7209684379006695e-07,
+      "loss": 0.8398,
+      "step": 20249
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7180353474332306e-07,
+      "loss": 0.3363,
+      "step": 20250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.715103123281134e-07,
+      "loss": 0.7998,
+      "step": 20251
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.712171765500484e-07,
+      "loss": 0.8477,
+      "step": 20252
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.709241274147336e-07,
+      "loss": 0.321,
+      "step": 20253
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.706311649277773e-07,
+      "loss": 0.8018,
+      "step": 20254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7033828909478553e-07,
+      "loss": 0.792,
+      "step": 20255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7004549992135885e-07,
+      "loss": 0.8276,
+      "step": 20256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.697527974131011e-07,
+      "loss": 0.917,
+      "step": 20257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6946018157561113e-07,
+      "loss": 0.8208,
+      "step": 20258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6916765241448617e-07,
+      "loss": 0.6996,
+      "step": 20259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.688752099353244e-07,
+      "loss": 0.8164,
+      "step": 20260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6858285414371985e-07,
+      "loss": 0.8145,
+      "step": 20261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6829058504526515e-07,
+      "loss": 0.3464,
+      "step": 20262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.679984026455542e-07,
+      "loss": 0.8774,
+      "step": 20263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.677063069501747e-07,
+      "loss": 0.8218,
+      "step": 20264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6741429796471616e-07,
+      "loss": 0.8955,
+      "step": 20265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.671223756947635e-07,
+      "loss": 0.8052,
+      "step": 20266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.668305401459022e-07,
+      "loss": 0.8462,
+      "step": 20267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6653879132371736e-07,
+      "loss": 0.8438,
+      "step": 20268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6624712923378834e-07,
+      "loss": 0.8457,
+      "step": 20269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6595555388169733e-07,
+      "loss": 0.7988,
+      "step": 20270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.65664065273021e-07,
+      "loss": 0.8447,
+      "step": 20271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.653726634133354e-07,
+      "loss": 0.9082,
+      "step": 20272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.650813483082173e-07,
+      "loss": 0.8296,
+      "step": 20273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.647901199632382e-07,
+      "loss": 0.832,
+      "step": 20274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6449897838397095e-07,
+      "loss": 0.8657,
+      "step": 20275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.642079235759861e-07,
+      "loss": 0.3439,
+      "step": 20276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.639169555448508e-07,
+      "loss": 0.8569,
+      "step": 20277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.636260742961329e-07,
+      "loss": 0.8291,
+      "step": 20278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.633352798353952e-07,
+      "loss": 0.791,
+      "step": 20279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6304457216820254e-07,
+      "loss": 0.8677,
+      "step": 20280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6275395130011724e-07,
+      "loss": 0.7744,
+      "step": 20281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.624634172366982e-07,
+      "loss": 0.8545,
+      "step": 20282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6217296998350423e-07,
+      "loss": 0.8916,
+      "step": 20283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.618826095460938e-07,
+      "loss": 0.8101,
+      "step": 20284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.615923359300184e-07,
+      "loss": 0.8364,
+      "step": 20285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6130214914083426e-07,
+      "loss": 0.3513,
+      "step": 20286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.610120491840908e-07,
+      "loss": 0.7881,
+      "step": 20287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6072203606533974e-07,
+      "loss": 0.8335,
+      "step": 20288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.604321097901299e-07,
+      "loss": 0.9023,
+      "step": 20289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.601422703640068e-07,
+      "loss": 0.9072,
+      "step": 20290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5985251779251616e-07,
+      "loss": 0.3518,
+      "step": 20291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5956285208120176e-07,
+      "loss": 0.8271,
+      "step": 20292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5927327323560367e-07,
+      "loss": 0.8252,
+      "step": 20293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.589837812612646e-07,
+      "loss": 0.8096,
+      "step": 20294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.586943761637203e-07,
+      "loss": 0.8931,
+      "step": 20295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.584050579485083e-07,
+      "loss": 0.8394,
+      "step": 20296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.581158266211655e-07,
+      "loss": 0.8374,
+      "step": 20297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5782668218722403e-07,
+      "loss": 0.8584,
+      "step": 20298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.575376246522162e-07,
+      "loss": 0.8413,
+      "step": 20299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.572486540216703e-07,
+      "loss": 0.8267,
+      "step": 20300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.569597703011158e-07,
+      "loss": 0.8442,
+      "step": 20301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5667097349608116e-07,
+      "loss": 0.8184,
+      "step": 20302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.563822636120896e-07,
+      "loss": 0.8379,
+      "step": 20303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5609364065466565e-07,
+      "loss": 0.8467,
+      "step": 20304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.558051046293305e-07,
+      "loss": 0.8438,
+      "step": 20305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.555166555416035e-07,
+      "loss": 0.8809,
+      "step": 20306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.552282933970054e-07,
+      "loss": 0.8452,
+      "step": 20307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5494001820105103e-07,
+      "loss": 0.8608,
+      "step": 20308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5465182995925563e-07,
+      "loss": 0.8452,
+      "step": 20309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.543637286771346e-07,
+      "loss": 0.7881,
+      "step": 20310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5407571436019813e-07,
+      "loss": 0.8354,
+      "step": 20311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.537877870139562e-07,
+      "loss": 0.8638,
+      "step": 20312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5349994664391825e-07,
+      "loss": 0.8066,
+      "step": 20313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5321219325559053e-07,
+      "loss": 0.8682,
+      "step": 20314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5292452685447854e-07,
+      "loss": 0.8481,
+      "step": 20315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.526369474460851e-07,
+      "loss": 0.7988,
+      "step": 20316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5234945503591365e-07,
+      "loss": 0.8706,
+      "step": 20317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5206204962946365e-07,
+      "loss": 0.7578,
+      "step": 20318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.517747312322318e-07,
+      "loss": 0.8325,
+      "step": 20319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.514874998497165e-07,
+      "loss": 0.8252,
+      "step": 20320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.512003554874139e-07,
+      "loss": 0.8394,
+      "step": 20321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5091329815081576e-07,
+      "loss": 0.8403,
+      "step": 20322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.506263278454154e-07,
+      "loss": 0.8003,
+      "step": 20323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.503394445767023e-07,
+      "loss": 0.8472,
+      "step": 20324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.500526483501644e-07,
+      "loss": 0.7974,
+      "step": 20325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.497659391712894e-07,
+      "loss": 0.8711,
+      "step": 20326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4947931704556194e-07,
+      "loss": 0.8584,
+      "step": 20327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.491927819784658e-07,
+      "loss": 0.7954,
+      "step": 20328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4890633397548343e-07,
+      "loss": 0.7988,
+      "step": 20329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.486199730420954e-07,
+      "loss": 0.8096,
+      "step": 20330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.483336991837783e-07,
+      "loss": 0.8174,
+      "step": 20331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.480475124060102e-07,
+      "loss": 0.7264,
+      "step": 20332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.477614127142654e-07,
+      "loss": 0.3369,
+      "step": 20333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4747540011401913e-07,
+      "loss": 0.8154,
+      "step": 20334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.471894746107419e-07,
+      "loss": 0.8711,
+      "step": 20335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4690363620990495e-07,
+      "loss": 0.8477,
+      "step": 20336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.466178849169761e-07,
+      "loss": 0.8184,
+      "step": 20337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.463322207374221e-07,
+      "loss": 0.8774,
+      "step": 20338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.460466436767086e-07,
+      "loss": 0.8179,
+      "step": 20339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4576115374029837e-07,
+      "loss": 0.8306,
+      "step": 20340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.454757509336538e-07,
+      "loss": 0.7944,
+      "step": 20341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.45190435262236e-07,
+      "loss": 0.8359,
+      "step": 20342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4490520673150295e-07,
+      "loss": 0.8081,
+      "step": 20343
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.446200653469107e-07,
+      "loss": 0.8882,
+      "step": 20344
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.443350111139144e-07,
+      "loss": 0.7817,
+      "step": 20345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4405004403796805e-07,
+      "loss": 0.8125,
+      "step": 20346
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4376516412452495e-07,
+      "loss": 0.8657,
+      "step": 20347
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4348037137903255e-07,
+      "loss": 0.9023,
+      "step": 20348
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.43195665806942e-07,
+      "loss": 0.8394,
+      "step": 20349
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4291104741369886e-07,
+      "loss": 0.8457,
+      "step": 20350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.426265162047477e-07,
+      "loss": 0.8145,
+      "step": 20351
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.423420721855337e-07,
+      "loss": 0.7715,
+      "step": 20352
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.420577153614969e-07,
+      "loss": 0.834,
+      "step": 20353
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4177344573807843e-07,
+      "loss": 0.8779,
+      "step": 20354
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4148926332071795e-07,
+      "loss": 0.8345,
+      "step": 20355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.41205168114851e-07,
+      "loss": 0.8301,
+      "step": 20356
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.409211601259117e-07,
+      "loss": 0.8003,
+      "step": 20357
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.406372393593361e-07,
+      "loss": 0.8462,
+      "step": 20358
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4035340582055386e-07,
+      "loss": 0.7852,
+      "step": 20359
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.400696595149972e-07,
+      "loss": 0.7905,
+      "step": 20360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3978600044809196e-07,
+      "loss": 0.8555,
+      "step": 20361
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3950242862526803e-07,
+      "loss": 0.7979,
+      "step": 20362
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3921894405194897e-07,
+      "loss": 0.8262,
+      "step": 20363
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.389355467335571e-07,
+      "loss": 0.7456,
+      "step": 20364
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.386522366755169e-07,
+      "loss": 0.8589,
+      "step": 20365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3836901388324635e-07,
+      "loss": 0.8325,
+      "step": 20366
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3808587836216445e-07,
+      "loss": 0.8486,
+      "step": 20367
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.378028301176895e-07,
+      "loss": 0.7373,
+      "step": 20368
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3751986915523516e-07,
+      "loss": 0.8164,
+      "step": 20369
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3723699548021524e-07,
+      "loss": 0.8262,
+      "step": 20370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3695420909804166e-07,
+      "loss": 0.876,
+      "step": 20371
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.366715100141239e-07,
+      "loss": 0.8994,
+      "step": 20372
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.363888982338721e-07,
+      "loss": 0.8291,
+      "step": 20373
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3610637376269137e-07,
+      "loss": 0.8315,
+      "step": 20374
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.35823936605988e-07,
+      "loss": 0.8579,
+      "step": 20375
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.355415867691648e-07,
+      "loss": 0.856,
+      "step": 20376
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.352593242576231e-07,
+      "loss": 0.8887,
+      "step": 20377
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3497714907676403e-07,
+      "loss": 0.8472,
+      "step": 20378
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3469506123198615e-07,
+      "loss": 0.8633,
+      "step": 20379
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3441306072868505e-07,
+      "loss": 0.7471,
+      "step": 20380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3413114757225706e-07,
+      "loss": 0.8091,
+      "step": 20381
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.338493217680956e-07,
+      "loss": 0.7788,
+      "step": 20382
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.335675833215908e-07,
+      "loss": 0.8159,
+      "step": 20383
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.33285932238135e-07,
+      "loss": 0.8262,
+      "step": 20384
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3300436852311446e-07,
+      "loss": 0.8188,
+      "step": 20385
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3272289218191767e-07,
+      "loss": 0.8228,
+      "step": 20386
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3244150321992917e-07,
+      "loss": 0.8604,
+      "step": 20387
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.321602016425325e-07,
+      "loss": 0.8755,
+      "step": 20388
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.318789874551094e-07,
+      "loss": 0.8188,
+      "step": 20389
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3159786066303886e-07,
+      "loss": 0.8374,
+      "step": 20390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3131682127170004e-07,
+      "loss": 0.7461,
+      "step": 20391
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.310358692864708e-07,
+      "loss": 0.814,
+      "step": 20392
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.30755004712724e-07,
+      "loss": 0.8086,
+      "step": 20393
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3047422755583435e-07,
+      "loss": 0.8657,
+      "step": 20394
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3019353782117524e-07,
+      "loss": 0.8071,
+      "step": 20395
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2991293551411297e-07,
+      "loss": 0.7959,
+      "step": 20396
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2963242064001874e-07,
+      "loss": 0.349,
+      "step": 20397
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.293519932042578e-07,
+      "loss": 0.7974,
+      "step": 20398
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2907165321219523e-07,
+      "loss": 0.7915,
+      "step": 20399
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.287914006691957e-07,
+      "loss": 0.8364,
+      "step": 20400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2851123558061927e-07,
+      "loss": 0.3248,
+      "step": 20401
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2823115795182837e-07,
+      "loss": 0.8105,
+      "step": 20402
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2795116778817815e-07,
+      "loss": 0.8032,
+      "step": 20403
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2767126509502656e-07,
+      "loss": 0.8296,
+      "step": 20404
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.273914498777293e-07,
+      "loss": 0.853,
+      "step": 20405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2711172214163875e-07,
+      "loss": 0.8662,
+      "step": 20406
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.268320818921062e-07,
+      "loss": 0.7787,
+      "step": 20407
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.265525291344841e-07,
+      "loss": 0.8135,
+      "step": 20408
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.262730638741186e-07,
+      "loss": 0.8545,
+      "step": 20409
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.259936861163566e-07,
+      "loss": 0.8623,
+      "step": 20410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2571439586654215e-07,
+      "loss": 0.3898,
+      "step": 20411
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.254351931300199e-07,
+      "loss": 0.8262,
+      "step": 20412
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2515607791213165e-07,
+      "loss": 0.8472,
+      "step": 20413
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.248770502182159e-07,
+      "loss": 0.8564,
+      "step": 20414
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2459811005361227e-07,
+      "loss": 0.8101,
+      "step": 20415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.243192574236571e-07,
+      "loss": 0.7715,
+      "step": 20416
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2404049233368383e-07,
+      "loss": 0.8198,
+      "step": 20417
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2376181478902765e-07,
+      "loss": 0.8057,
+      "step": 20418
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2348322479501827e-07,
+      "loss": 0.7822,
+      "step": 20419
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2320472235698686e-07,
+      "loss": 0.8086,
+      "step": 20420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.22926307480262e-07,
+      "loss": 0.8096,
+      "step": 20421
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2264798017016883e-07,
+      "loss": 0.854,
+      "step": 20422
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.223697404320326e-07,
+      "loss": 0.8081,
+      "step": 20423
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2209158827117737e-07,
+      "loss": 0.7944,
+      "step": 20424
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2181352369292325e-07,
+      "loss": 0.8242,
+      "step": 20425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.215355467025911e-07,
+      "loss": 0.855,
+      "step": 20426
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.212576573054983e-07,
+      "loss": 0.8335,
+      "step": 20427
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.209798555069627e-07,
+      "loss": 0.2941,
+      "step": 20428
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.207021413122975e-07,
+      "loss": 0.7678,
+      "step": 20429
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.20424514726816e-07,
+      "loss": 0.8193,
+      "step": 20430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2014697575583074e-07,
+      "loss": 0.3033,
+      "step": 20431
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.198695244046502e-07,
+      "loss": 0.8149,
+      "step": 20432
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1959216067858246e-07,
+      "loss": 0.8486,
+      "step": 20433
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.193148845829359e-07,
+      "loss": 0.8457,
+      "step": 20434
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1903769612301357e-07,
+      "loss": 0.8472,
+      "step": 20435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.187605953041185e-07,
+      "loss": 0.8291,
+      "step": 20436
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1848358213155247e-07,
+      "loss": 0.8628,
+      "step": 20437
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1820665661061465e-07,
+      "loss": 0.8335,
+      "step": 20438
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1792981874660466e-07,
+      "loss": 0.7881,
+      "step": 20439
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1765306854481603e-07,
+      "loss": 0.812,
+      "step": 20440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1737640601054676e-07,
+      "loss": 0.8384,
+      "step": 20441
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.170998311490876e-07,
+      "loss": 0.8198,
+      "step": 20442
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.168233439657299e-07,
+      "loss": 0.8364,
+      "step": 20443
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.165469444657633e-07,
+      "loss": 0.8496,
+      "step": 20444
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1627063265447745e-07,
+      "loss": 0.8379,
+      "step": 20445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.15994408537157e-07,
+      "loss": 0.8491,
+      "step": 20446
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.157182721190872e-07,
+      "loss": 0.8071,
+      "step": 20447
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1544222340555105e-07,
+      "loss": 0.8574,
+      "step": 20448
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.151662624018288e-07,
+      "loss": 0.8291,
+      "step": 20449
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.148903891132017e-07,
+      "loss": 0.8286,
+      "step": 20450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1461460354494553e-07,
+      "loss": 0.8433,
+      "step": 20451
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.143389057023378e-07,
+      "loss": 0.7671,
+      "step": 20452
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1406329559065373e-07,
+      "loss": 0.8159,
+      "step": 20453
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.137877732151652e-07,
+      "loss": 0.897,
+      "step": 20454
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.13512338581144e-07,
+      "loss": 0.8701,
+      "step": 20455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.132369916938578e-07,
+      "loss": 0.8384,
+      "step": 20456
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.129617325585761e-07,
+      "loss": 0.8643,
+      "step": 20457
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.126865611805653e-07,
+      "loss": 0.3073,
+      "step": 20458
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.124114775650889e-07,
+      "loss": 0.8276,
+      "step": 20459
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.121364817174106e-07,
+      "loss": 0.8315,
+      "step": 20460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.118615736427911e-07,
+      "loss": 0.8574,
+      "step": 20461
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1158675334648843e-07,
+      "loss": 0.8496,
+      "step": 20462
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1131202083376287e-07,
+      "loss": 0.4069,
+      "step": 20463
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.110373761098685e-07,
+      "loss": 0.7837,
+      "step": 20464
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1076281918006e-07,
+      "loss": 0.812,
+      "step": 20465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.104883500495921e-07,
+      "loss": 0.814,
+      "step": 20466
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1021396872371333e-07,
+      "loss": 0.8564,
+      "step": 20467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0993967520767455e-07,
+      "loss": 0.79,
+      "step": 20468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.09665469506722e-07,
+      "loss": 0.8052,
+      "step": 20469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.093913516261022e-07,
+      "loss": 0.3905,
+      "step": 20470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.091173215710603e-07,
+      "loss": 0.3547,
+      "step": 20471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0884337934683826e-07,
+      "loss": 0.3557,
+      "step": 20472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0856952495867743e-07,
+      "loss": 0.8584,
+      "step": 20473
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.082957584118169e-07,
+      "loss": 0.8315,
+      "step": 20474
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0802207971149364e-07,
+      "loss": 0.7876,
+      "step": 20475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.077484888629446e-07,
+      "loss": 0.8389,
+      "step": 20476
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.074749858714022e-07,
+      "loss": 0.8262,
+      "step": 20477
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.072015707421006e-07,
+      "loss": 0.7915,
+      "step": 20478
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0692824348027127e-07,
+      "loss": 0.8276,
+      "step": 20479
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0665500409114265e-07,
+      "loss": 0.8369,
+      "step": 20480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0638185257994065e-07,
+      "loss": 0.832,
+      "step": 20481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0610878895189385e-07,
+      "loss": 0.7915,
+      "step": 20482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0583581321222367e-07,
+      "loss": 0.8892,
+      "step": 20483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.055629253661547e-07,
+      "loss": 0.8584,
+      "step": 20484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.052901254189062e-07,
+      "loss": 0.7812,
+      "step": 20485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.05017413375699e-07,
+      "loss": 0.7749,
+      "step": 20486
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.047447892417494e-07,
+      "loss": 0.8716,
+      "step": 20487
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0447225302227224e-07,
+      "loss": 0.8594,
+      "step": 20488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0419980472248264e-07,
+      "loss": 0.7964,
+      "step": 20489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0392744434759324e-07,
+      "loss": 0.875,
+      "step": 20490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.036551719028142e-07,
+      "loss": 0.8579,
+      "step": 20491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0338298739335536e-07,
+      "loss": 0.8237,
+      "step": 20492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.03110890824423e-07,
+      "loss": 0.8203,
+      "step": 20493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0283888220122246e-07,
+      "loss": 0.7925,
+      "step": 20494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.02566961528959e-07,
+      "loss": 0.8613,
+      "step": 20495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.022951288128335e-07,
+      "loss": 0.8398,
+      "step": 20496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.020233840580473e-07,
+      "loss": 0.7935,
+      "step": 20497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0175172726980014e-07,
+      "loss": 0.7822,
+      "step": 20498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.014801584532885e-07,
+      "loss": 0.8145,
+      "step": 20499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.012086776137075e-07,
+      "loss": 0.7871,
+      "step": 20500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.009372847562504e-07,
+      "loss": 0.8472,
+      "step": 20501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.006659798861107e-07,
+      "loss": 0.7559,
+      "step": 20502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0039476300847866e-07,
+      "loss": 0.8672,
+      "step": 20503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.001236341285425e-07,
+      "loss": 0.8281,
+      "step": 20504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.998525932514907e-07,
+      "loss": 0.8052,
+      "step": 20505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.995816403825076e-07,
+      "loss": 0.874,
+      "step": 20506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.993107755267761e-07,
+      "loss": 0.813,
+      "step": 20507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9903999868947995e-07,
+      "loss": 0.8613,
+      "step": 20508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.987693098757983e-07,
+      "loss": 0.8613,
+      "step": 20509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.984987090909098e-07,
+      "loss": 0.8677,
+      "step": 20510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9822819633999363e-07,
+      "loss": 0.3545,
+      "step": 20511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.979577716282224e-07,
+      "loss": 0.8188,
+      "step": 20512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9768743496077287e-07,
+      "loss": 0.8584,
+      "step": 20513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.974171863428128e-07,
+      "loss": 0.8428,
+      "step": 20514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9714702577951514e-07,
+      "loss": 0.8457,
+      "step": 20515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9687695327604857e-07,
+      "loss": 0.7847,
+      "step": 20516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9660696883757896e-07,
+      "loss": 0.8184,
+      "step": 20517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.963370724692722e-07,
+      "loss": 0.4441,
+      "step": 20518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.960672641762925e-07,
+      "loss": 0.8242,
+      "step": 20519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9579754396380064e-07,
+      "loss": 0.8608,
+      "step": 20520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.955279118369576e-07,
+      "loss": 0.813,
+      "step": 20521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9525836780092033e-07,
+      "loss": 0.7886,
+      "step": 20522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.949889118608463e-07,
+      "loss": 0.8472,
+      "step": 20523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9471954402189206e-07,
+      "loss": 0.8594,
+      "step": 20524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9445026428920896e-07,
+      "loss": 0.8472,
+      "step": 20525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9418107266795067e-07,
+      "loss": 0.854,
+      "step": 20526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9391196916326645e-07,
+      "loss": 0.7749,
+      "step": 20527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9364295378030384e-07,
+      "loss": 0.812,
+      "step": 20528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.933740265242109e-07,
+      "loss": 0.8745,
+      "step": 20529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.931051874001313e-07,
+      "loss": 0.8179,
+      "step": 20530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9283643641320867e-07,
+      "loss": 0.8579,
+      "step": 20531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.925677735685857e-07,
+      "loss": 0.896,
+      "step": 20532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.92299198871402e-07,
+      "loss": 0.8242,
+      "step": 20533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.920307123267958e-07,
+      "loss": 0.8027,
+      "step": 20534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9176231393990183e-07,
+      "loss": 0.8516,
+      "step": 20535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9149400371585654e-07,
+      "loss": 0.7827,
+      "step": 20536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.912257816597942e-07,
+      "loss": 0.8188,
+      "step": 20537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.909576477768445e-07,
+      "loss": 0.8652,
+      "step": 20538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9068960207213846e-07,
+      "loss": 0.8159,
+      "step": 20539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9042164455080356e-07,
+      "loss": 0.7939,
+      "step": 20540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9015377521796574e-07,
+      "loss": 0.8467,
+      "step": 20541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.898859940787514e-07,
+      "loss": 0.8008,
+      "step": 20542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8961830113828204e-07,
+      "loss": 0.8926,
+      "step": 20543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8935069640167913e-07,
+      "loss": 0.8252,
+      "step": 20544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8908317987406355e-07,
+      "loss": 0.8447,
+      "step": 20545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.888157515605534e-07,
+      "loss": 0.813,
+      "step": 20546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8854841146626244e-07,
+      "loss": 0.7964,
+      "step": 20547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8828115959630876e-07,
+      "loss": 0.8638,
+      "step": 20548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.880139959558027e-07,
+      "loss": 0.5186,
+      "step": 20549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8774692054985685e-07,
+      "loss": 0.7979,
+      "step": 20550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.874799333835799e-07,
+      "loss": 0.8623,
+      "step": 20551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.872130344620811e-07,
+      "loss": 0.8403,
+      "step": 20552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.869462237904659e-07,
+      "loss": 0.8105,
+      "step": 20553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.866795013738378e-07,
+      "loss": 0.8125,
+      "step": 20554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.864128672173001e-07,
+      "loss": 0.8652,
+      "step": 20555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8614632132595585e-07,
+      "loss": 0.8066,
+      "step": 20556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.858798637049016e-07,
+      "loss": 0.8145,
+      "step": 20557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.856134943592377e-07,
+      "loss": 0.8394,
+      "step": 20558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8534721329405886e-07,
+      "loss": 0.8262,
+      "step": 20559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8508102051445896e-07,
+      "loss": 0.875,
+      "step": 20560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.848149160255321e-07,
+      "loss": 0.8442,
+      "step": 20561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8454889983236765e-07,
+      "loss": 0.8496,
+      "step": 20562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8428297194005593e-07,
+      "loss": 0.7573,
+      "step": 20563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.840171323536851e-07,
+      "loss": 0.9126,
+      "step": 20564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8375138107834055e-07,
+      "loss": 0.8281,
+      "step": 20565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.834857181191065e-07,
+      "loss": 0.8418,
+      "step": 20566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8322014348106395e-07,
+      "loss": 0.8066,
+      "step": 20567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8295465716929603e-07,
+      "loss": 0.8408,
+      "step": 20568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8268925918888147e-07,
+      "loss": 0.833,
+      "step": 20569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.824239495448961e-07,
+      "loss": 0.8325,
+      "step": 20570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.821587282424183e-07,
+      "loss": 0.7979,
+      "step": 20571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.81893595286521e-07,
+      "loss": 0.834,
+      "step": 20572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8162855068227534e-07,
+      "loss": 0.8071,
+      "step": 20573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.813635944347538e-07,
+      "loss": 0.8223,
+      "step": 20574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.810987265490246e-07,
+      "loss": 0.8555,
+      "step": 20575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8083394703015477e-07,
+      "loss": 0.8589,
+      "step": 20576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.805692558832114e-07,
+      "loss": 0.8096,
+      "step": 20577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.80304653113257e-07,
+      "loss": 0.8369,
+      "step": 20578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8004013872535495e-07,
+      "loss": 0.8198,
+      "step": 20579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7977571272456383e-07,
+      "loss": 0.7607,
+      "step": 20580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.795113751159446e-07,
+      "loss": 0.8091,
+      "step": 20581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.792471259045538e-07,
+      "loss": 0.8726,
+      "step": 20582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.789829650954463e-07,
+      "loss": 0.8208,
+      "step": 20583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7871889269367747e-07,
+      "loss": 0.7671,
+      "step": 20584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.784549087042988e-07,
+      "loss": 0.8252,
+      "step": 20585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7819101313235906e-07,
+      "loss": 0.8203,
+      "step": 20586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7792720598290923e-07,
+      "loss": 0.8311,
+      "step": 20587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.776634872609947e-07,
+      "loss": 0.7725,
+      "step": 20588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.773998569716614e-07,
+      "loss": 0.3396,
+      "step": 20589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.771363151199542e-07,
+      "loss": 0.8301,
+      "step": 20590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7687286171091355e-07,
+      "loss": 0.8052,
+      "step": 20591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.766094967495798e-07,
+      "loss": 0.8657,
+      "step": 20592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.763462202409929e-07,
+      "loss": 0.8584,
+      "step": 20593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.760830321901876e-07,
+      "loss": 0.8447,
+      "step": 20594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7581993260220104e-07,
+      "loss": 0.8188,
+      "step": 20595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7555692148206536e-07,
+      "loss": 0.8779,
+      "step": 20596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.752939988348142e-07,
+      "loss": 0.8735,
+      "step": 20597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7503116466547586e-07,
+      "loss": 0.8872,
+      "step": 20598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7476841897907844e-07,
+      "loss": 0.8716,
+      "step": 20599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.745057617806508e-07,
+      "loss": 0.8638,
+      "step": 20600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.742431930752161e-07,
+      "loss": 0.8057,
+      "step": 20601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.739807128677986e-07,
+      "loss": 0.8359,
+      "step": 20602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.737183211634199e-07,
+      "loss": 0.8389,
+      "step": 20603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.734560179671004e-07,
+      "loss": 0.8442,
+      "step": 20604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.731938032838567e-07,
+      "loss": 0.9121,
+      "step": 20605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7293167711870747e-07,
+      "loss": 0.8081,
+      "step": 20606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7266963947666544e-07,
+      "loss": 0.8218,
+      "step": 20607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7240769036274606e-07,
+      "loss": 0.8306,
+      "step": 20608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7214582978195855e-07,
+      "loss": 0.8604,
+      "step": 20609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7188405773931515e-07,
+      "loss": 0.833,
+      "step": 20610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7162237423982284e-07,
+      "loss": 0.5386,
+      "step": 20611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.713607792884871e-07,
+      "loss": 0.8076,
+      "step": 20612
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.710992728903129e-07,
+      "loss": 0.8179,
+      "step": 20613
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7083785505030503e-07,
+      "loss": 0.8672,
+      "step": 20614
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.705765257734628e-07,
+      "loss": 0.7803,
+      "step": 20615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7031528506478785e-07,
+      "loss": 0.8115,
+      "step": 20616
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7005413292927674e-07,
+      "loss": 0.8462,
+      "step": 20617
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6979306937192486e-07,
+      "loss": 0.8486,
+      "step": 20618
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.695320943977293e-07,
+      "loss": 0.8252,
+      "step": 20619
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6927120801168e-07,
+      "loss": 0.8071,
+      "step": 20620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.690104102187697e-07,
+      "loss": 0.8047,
+      "step": 20621
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6874970102398934e-07,
+      "loss": 0.8643,
+      "step": 20622
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.684890804323232e-07,
+      "loss": 0.8794,
+      "step": 20623
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.682285484487619e-07,
+      "loss": 0.8809,
+      "step": 20624
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6796810507828515e-07,
+      "loss": 0.8047,
+      "step": 20625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.67707750325878e-07,
+      "loss": 0.8135,
+      "step": 20626
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6744748419652144e-07,
+      "loss": 0.8169,
+      "step": 20627
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.671873066951942e-07,
+      "loss": 0.8247,
+      "step": 20628
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.66927217826874e-07,
+      "loss": 0.8521,
+      "step": 20629
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.66667217596538e-07,
+      "loss": 0.8369,
+      "step": 20630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6640730600915884e-07,
+      "loss": 0.77,
+      "step": 20631
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.661474830697098e-07,
+      "loss": 0.7944,
+      "step": 20632
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.658877487831608e-07,
+      "loss": 0.8262,
+      "step": 20633
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6562810315448115e-07,
+      "loss": 0.8208,
+      "step": 20634
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6536854618864023e-07,
+      "loss": 0.8091,
+      "step": 20635
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.651090778906008e-07,
+      "loss": 0.8408,
+      "step": 20636
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.648496982653299e-07,
+      "loss": 0.8774,
+      "step": 20637
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.645904073177881e-07,
+      "loss": 0.8579,
+      "step": 20638
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.643312050529352e-07,
+      "loss": 0.7744,
+      "step": 20639
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.640720914757323e-07,
+      "loss": 0.7744,
+      "step": 20640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.638130665911349e-07,
+      "loss": 0.8408,
+      "step": 20641
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6355413040409903e-07,
+      "loss": 0.8442,
+      "step": 20642
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.632952829195796e-07,
+      "loss": 0.7935,
+      "step": 20643
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.630365241425282e-07,
+      "loss": 0.8491,
+      "step": 20644
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.627778540778948e-07,
+      "loss": 0.8013,
+      "step": 20645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.625192727306276e-07,
+      "loss": 0.8521,
+      "step": 20646
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6226078010567433e-07,
+      "loss": 0.8096,
+      "step": 20647
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.620023762079816e-07,
+      "loss": 0.7144,
+      "step": 20648
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6174406104249106e-07,
+      "loss": 0.7388,
+      "step": 20649
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6148583461414644e-07,
+      "loss": 0.7974,
+      "step": 20650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.612276969278872e-07,
+      "loss": 0.8594,
+      "step": 20651
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6096964798865095e-07,
+      "loss": 0.814,
+      "step": 20652
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.607116878013767e-07,
+      "loss": 0.8271,
+      "step": 20653
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.604538163709975e-07,
+      "loss": 0.8188,
+      "step": 20654
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6019603370244794e-07,
+      "loss": 0.7817,
+      "step": 20655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5993833980066005e-07,
+      "loss": 0.8594,
+      "step": 20656
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5968073467056387e-07,
+      "loss": 0.918,
+      "step": 20657
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5942321831708705e-07,
+      "loss": 0.8242,
+      "step": 20658
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.591657907451568e-07,
+      "loss": 0.7969,
+      "step": 20659
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5890845195969746e-07,
+      "loss": 0.7925,
+      "step": 20660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.586512019656335e-07,
+      "loss": 0.8242,
+      "step": 20661
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.583940407678854e-07,
+      "loss": 0.8315,
+      "step": 20662
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5813696837137425e-07,
+      "loss": 0.7534,
+      "step": 20663
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.578799847810177e-07,
+      "loss": 0.9097,
+      "step": 20664
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5762309000173137e-07,
+      "loss": 0.8335,
+      "step": 20665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.573662840384312e-07,
+      "loss": 0.8218,
+      "step": 20666
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.571095668960295e-07,
+      "loss": 0.813,
+      "step": 20667
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.568529385794378e-07,
+      "loss": 0.8652,
+      "step": 20668
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5659639909356725e-07,
+      "loss": 0.8789,
+      "step": 20669
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5633994844332444e-07,
+      "loss": 0.8242,
+      "step": 20670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.560835866336149e-07,
+      "loss": 0.7778,
+      "step": 20671
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.558273136693452e-07,
+      "loss": 0.7988,
+      "step": 20672
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.555711295554165e-07,
+      "loss": 0.8779,
+      "step": 20673
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5531503429673154e-07,
+      "loss": 0.8242,
+      "step": 20674
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5505902789818857e-07,
+      "loss": 0.8955,
+      "step": 20675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.548031103646865e-07,
+      "loss": 0.8311,
+      "step": 20676
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5454728170112087e-07,
+      "loss": 0.8535,
+      "step": 20677
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5429154191238547e-07,
+      "loss": 0.8066,
+      "step": 20678
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.540358910033731e-07,
+      "loss": 0.8418,
+      "step": 20679
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.537803289789765e-07,
+      "loss": 0.7563,
+      "step": 20680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5352485584408293e-07,
+      "loss": 0.8735,
+      "step": 20681
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5326947160358115e-07,
+      "loss": 0.8364,
+      "step": 20682
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5301417626235735e-07,
+      "loss": 0.8184,
+      "step": 20683
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.527589698252937e-07,
+      "loss": 0.8569,
+      "step": 20684
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.525038522972757e-07,
+      "loss": 0.8071,
+      "step": 20685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5224882368318116e-07,
+      "loss": 0.835,
+      "step": 20686
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.519938839878906e-07,
+      "loss": 0.853,
+      "step": 20687
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.517390332162818e-07,
+      "loss": 0.8193,
+      "step": 20688
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5148427137323083e-07,
+      "loss": 0.8193,
+      "step": 20689
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.512295984636105e-07,
+      "loss": 0.7749,
+      "step": 20690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5097501449229243e-07,
+      "loss": 0.8252,
+      "step": 20691
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.507205194641489e-07,
+      "loss": 0.8281,
+      "step": 20692
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.504661133840487e-07,
+      "loss": 0.8247,
+      "step": 20693
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.502117962568574e-07,
+      "loss": 0.8984,
+      "step": 20694
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.499575680874429e-07,
+      "loss": 0.8564,
+      "step": 20695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.497034288806672e-07,
+      "loss": 0.8447,
+      "step": 20696
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4944937864139274e-07,
+      "loss": 0.8115,
+      "step": 20697
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.491954173744805e-07,
+      "loss": 0.761,
+      "step": 20698
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.489415450847877e-07,
+      "loss": 0.4419,
+      "step": 20699
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.486877617771722e-07,
+      "loss": 0.8491,
+      "step": 20700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4843406745649066e-07,
+      "loss": 0.8726,
+      "step": 20701
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4818046212759473e-07,
+      "loss": 0.8369,
+      "step": 20702
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4792694579533715e-07,
+      "loss": 0.8081,
+      "step": 20703
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4767351846456744e-07,
+      "loss": 0.8174,
+      "step": 20704
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.474201801401339e-07,
+      "loss": 0.7788,
+      "step": 20705
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4716693082688493e-07,
+      "loss": 0.79,
+      "step": 20706
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.469137705296638e-07,
+      "loss": 0.8452,
+      "step": 20707
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.466606992533156e-07,
+      "loss": 0.8193,
+      "step": 20708
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.464077170026808e-07,
+      "loss": 0.7573,
+      "step": 20709
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4615482378259845e-07,
+      "loss": 0.8047,
+      "step": 20710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4590201959790903e-07,
+      "loss": 0.8423,
+      "step": 20711
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.45649304453447e-07,
+      "loss": 0.8076,
+      "step": 20712
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.45396678354048e-07,
+      "loss": 0.814,
+      "step": 20713
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.451441413045459e-07,
+      "loss": 0.8281,
+      "step": 20714
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4489169330977177e-07,
+      "loss": 0.8857,
+      "step": 20715
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4463933437455465e-07,
+      "loss": 0.7686,
+      "step": 20716
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4438706450372337e-07,
+      "loss": 0.7939,
+      "step": 20717
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.44134883702103e-07,
+      "loss": 0.8154,
+      "step": 20718
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4388279197452024e-07,
+      "loss": 0.813,
+      "step": 20719
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.436307893257956e-07,
+      "loss": 0.8213,
+      "step": 20720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4337887576075257e-07,
+      "loss": 0.3995,
+      "step": 20721
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4312705128420944e-07,
+      "loss": 0.8408,
+      "step": 20722
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4287531590098345e-07,
+      "loss": 0.8896,
+      "step": 20723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.426236696158908e-07,
+      "loss": 0.874,
+      "step": 20724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.423721124337481e-07,
+      "loss": 0.8252,
+      "step": 20725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.42120644359365e-07,
+      "loss": 0.8657,
+      "step": 20726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.418692653975553e-07,
+      "loss": 0.8408,
+      "step": 20727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4161797555312625e-07,
+      "loss": 0.8647,
+      "step": 20728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.413667748308852e-07,
+      "loss": 0.8511,
+      "step": 20729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.411156632356399e-07,
+      "loss": 0.8555,
+      "step": 20730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.408646407721927e-07,
+      "loss": 0.8433,
+      "step": 20731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.406137074453469e-07,
+      "loss": 0.8398,
+      "step": 20732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4036286325990377e-07,
+      "loss": 0.8228,
+      "step": 20733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.40112108220661e-07,
+      "loss": 0.8398,
+      "step": 20734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3986144233241815e-07,
+      "loss": 0.8389,
+      "step": 20735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.39610865599968e-07,
+      "loss": 0.8091,
+      "step": 20736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3936037802810576e-07,
+      "loss": 0.8184,
+      "step": 20737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.391099796216241e-07,
+      "loss": 0.8726,
+      "step": 20738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.388596703853131e-07,
+      "loss": 0.8267,
+      "step": 20739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3860945032396173e-07,
+      "loss": 0.7769,
+      "step": 20740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.383593194423568e-07,
+      "loss": 0.8682,
+      "step": 20741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3810927774528436e-07,
+      "loss": 0.8569,
+      "step": 20742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3785932523752786e-07,
+      "loss": 0.8037,
+      "step": 20743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3760946192386844e-07,
+      "loss": 0.8398,
+      "step": 20744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.373596878090868e-07,
+      "loss": 0.8354,
+      "step": 20745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.371100028979624e-07,
+      "loss": 0.8101,
+      "step": 20746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.368604071952708e-07,
+      "loss": 0.8511,
+      "step": 20747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3661090070578886e-07,
+      "loss": 0.8926,
+      "step": 20748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.363614834342888e-07,
+      "loss": 0.8633,
+      "step": 20749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3611215538554176e-07,
+      "loss": 0.8633,
+      "step": 20750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.35862916564319e-07,
+      "loss": 0.877,
+      "step": 20751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3561376697538785e-07,
+      "loss": 0.875,
+      "step": 20752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3536470662351607e-07,
+      "loss": 0.8203,
+      "step": 20753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3511573551346824e-07,
+      "loss": 0.8594,
+      "step": 20754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3486685365000715e-07,
+      "loss": 0.8218,
+      "step": 20755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3461806103789463e-07,
+      "loss": 0.8237,
+      "step": 20756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.343693576818896e-07,
+      "loss": 0.8247,
+      "step": 20757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.341207435867511e-07,
+      "loss": 0.7618,
+      "step": 20758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3387221875723577e-07,
+      "loss": 0.8452,
+      "step": 20759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.336237831980971e-07,
+      "loss": 0.8389,
+      "step": 20760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.333754369140896e-07,
+      "loss": 0.793,
+      "step": 20761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3312717990996337e-07,
+      "loss": 0.8516,
+      "step": 20762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.328790121904679e-07,
+      "loss": 0.3511,
+      "step": 20763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.326309337603523e-07,
+      "loss": 0.793,
+      "step": 20764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.323829446243604e-07,
+      "loss": 0.8179,
+      "step": 20765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3213504478723843e-07,
+      "loss": 0.8315,
+      "step": 20766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3188723425372937e-07,
+      "loss": 0.8623,
+      "step": 20767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3163951302857376e-07,
+      "loss": 0.8276,
+      "step": 20768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3139188111651065e-07,
+      "loss": 0.8511,
+      "step": 20769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.311443385222768e-07,
+      "loss": 0.7324,
+      "step": 20770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3089688525060836e-07,
+      "loss": 0.8315,
+      "step": 20771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.306495213062416e-07,
+      "loss": 0.8501,
+      "step": 20772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3040224669390664e-07,
+      "loss": 0.771,
+      "step": 20773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3015506141833577e-07,
+      "loss": 0.8057,
+      "step": 20774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.299079654842574e-07,
+      "loss": 0.8345,
+      "step": 20775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.296609588963978e-07,
+      "loss": 0.8271,
+      "step": 20776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.294140416594843e-07,
+      "loss": 0.853,
+      "step": 20777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2916721377823977e-07,
+      "loss": 0.8354,
+      "step": 20778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2892047525738656e-07,
+      "loss": 0.8301,
+      "step": 20779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.286738261016459e-07,
+      "loss": 0.8794,
+      "step": 20780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2842726631573616e-07,
+      "loss": 0.8408,
+      "step": 20781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.281807959043737e-07,
+      "loss": 0.8145,
+      "step": 20782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.279344148722752e-07,
+      "loss": 0.8979,
+      "step": 20783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2768812322415246e-07,
+      "loss": 0.8652,
+      "step": 20784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2744192096472e-07,
+      "loss": 0.8677,
+      "step": 20785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.271958080986853e-07,
+      "loss": 0.7681,
+      "step": 20786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2694978463075956e-07,
+      "loss": 0.8486,
+      "step": 20787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.267038505656478e-07,
+      "loss": 0.8652,
+      "step": 20788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2645800590805467e-07,
+      "loss": 0.834,
+      "step": 20789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.262122506626847e-07,
+      "loss": 0.8086,
+      "step": 20790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2596658483424026e-07,
+      "loss": 0.8877,
+      "step": 20791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2572100842741927e-07,
+      "loss": 0.835,
+      "step": 20792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.254755214469224e-07,
+      "loss": 0.3713,
+      "step": 20793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.252301238974448e-07,
+      "loss": 0.8018,
+      "step": 20794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.249848157836805e-07,
+      "loss": 0.8384,
+      "step": 20795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2473959711032456e-07,
+      "loss": 0.8491,
+      "step": 20796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.244944678820661e-07,
+      "loss": 0.7761,
+      "step": 20797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2424942810359684e-07,
+      "loss": 0.8174,
+      "step": 20798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2400447777960476e-07,
+      "loss": 0.8877,
+      "step": 20799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2375961691477496e-07,
+      "loss": 0.7671,
+      "step": 20800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.235148455137932e-07,
+      "loss": 0.8325,
+      "step": 20801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.232701635813401e-07,
+      "loss": 0.402,
+      "step": 20802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.230255711220992e-07,
+      "loss": 0.8721,
+      "step": 20803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.227810681407495e-07,
+      "loss": 0.875,
+      "step": 20804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2253665464196724e-07,
+      "loss": 0.8433,
+      "step": 20805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.22292330630431e-07,
+      "loss": 0.8271,
+      "step": 20806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.22048096110813e-07,
+      "loss": 0.8394,
+      "step": 20807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.218039510877863e-07,
+      "loss": 0.7715,
+      "step": 20808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2155989556602206e-07,
+      "loss": 0.8359,
+      "step": 20809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2131592955018885e-07,
+      "loss": 0.8584,
+      "step": 20810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2107205304495515e-07,
+      "loss": 0.856,
+      "step": 20811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2082826605498605e-07,
+      "loss": 0.8013,
+      "step": 20812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2058456858494627e-07,
+      "loss": 0.835,
+      "step": 20813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.203409606394969e-07,
+      "loss": 0.8091,
+      "step": 20814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2009744222329885e-07,
+      "loss": 0.8252,
+      "step": 20815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1985401334101165e-07,
+      "loss": 0.8496,
+      "step": 20816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.196106739972926e-07,
+      "loss": 0.8242,
+      "step": 20817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.193674241967959e-07,
+      "loss": 0.8325,
+      "step": 20818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1912426394417716e-07,
+      "loss": 0.8003,
+      "step": 20819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1888119324408715e-07,
+      "loss": 0.813,
+      "step": 20820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.186382121011755e-07,
+      "loss": 0.8286,
+      "step": 20821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.183953205200929e-07,
+      "loss": 0.855,
+      "step": 20822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.18152518505484e-07,
+      "loss": 0.7539,
+      "step": 20823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.179098060619951e-07,
+      "loss": 0.8022,
+      "step": 20824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.176671831942707e-07,
+      "loss": 0.8105,
+      "step": 20825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1742464990695166e-07,
+      "loss": 0.8696,
+      "step": 20826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.17182206204677e-07,
+      "loss": 0.7939,
+      "step": 20827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.169398520920863e-07,
+      "loss": 0.8838,
+      "step": 20828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1669758757381543e-07,
+      "loss": 0.8623,
+      "step": 20829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.164554126545005e-07,
+      "loss": 0.8862,
+      "step": 20830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.16213327338773e-07,
+      "loss": 0.8398,
+      "step": 20831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.159713316312657e-07,
+      "loss": 0.8047,
+      "step": 20832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1572942553660836e-07,
+      "loss": 0.833,
+      "step": 20833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1548760905942776e-07,
+      "loss": 0.7808,
+      "step": 20834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1524588220435195e-07,
+      "loss": 0.8198,
+      "step": 20835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.150042449760038e-07,
+      "loss": 0.8389,
+      "step": 20836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.147626973790069e-07,
+      "loss": 0.8179,
+      "step": 20837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1452123941798364e-07,
+      "loss": 0.8008,
+      "step": 20838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.14279871097552e-07,
+      "loss": 0.7964,
+      "step": 20839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1403859242232994e-07,
+      "loss": 0.8423,
+      "step": 20840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.137974033969343e-07,
+      "loss": 0.8315,
+      "step": 20841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.135563040259781e-07,
+      "loss": 0.8145,
+      "step": 20842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1331529431407604e-07,
+      "loss": 0.813,
+      "step": 20843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1307437426583597e-07,
+      "loss": 0.8159,
+      "step": 20844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1283354388586927e-07,
+      "loss": 0.8896,
+      "step": 20845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.125928031787845e-07,
+      "loss": 0.8862,
+      "step": 20846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1235215214918456e-07,
+      "loss": 0.8286,
+      "step": 20847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.121115908016742e-07,
+      "loss": 0.7837,
+      "step": 20848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1187111914085743e-07,
+      "loss": 0.8516,
+      "step": 20849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1163073717133287e-07,
+      "loss": 0.833,
+      "step": 20850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.113904448977012e-07,
+      "loss": 0.9023,
+      "step": 20851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1115024232455773e-07,
+      "loss": 0.8311,
+      "step": 20852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.109101294564998e-07,
+      "loss": 0.8115,
+      "step": 20853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.106701062981199e-07,
+      "loss": 0.8369,
+      "step": 20854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.104301728540093e-07,
+      "loss": 0.7812,
+      "step": 20855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1019032912876e-07,
+      "loss": 0.7993,
+      "step": 20856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0995057512696046e-07,
+      "loss": 0.8286,
+      "step": 20857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.097109108531959e-07,
+      "loss": 0.7861,
+      "step": 20858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0947133631205383e-07,
+      "loss": 0.8594,
+      "step": 20859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.092318515081161e-07,
+      "loss": 0.7935,
+      "step": 20860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.08992456445964e-07,
+      "loss": 0.7935,
+      "step": 20861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0875315113017944e-07,
+      "loss": 0.7852,
+      "step": 20862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0851393556533825e-07,
+      "loss": 0.835,
+      "step": 20863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0827480975601896e-07,
+      "loss": 0.8384,
+      "step": 20864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.080357737067963e-07,
+      "loss": 0.8354,
+      "step": 20865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.077968274222426e-07,
+      "loss": 0.8345,
+      "step": 20866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.075579709069304e-07,
+      "loss": 0.8892,
+      "step": 20867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.073192041654272e-07,
+      "loss": 0.8589,
+      "step": 20868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.070805272023025e-07,
+      "loss": 0.8101,
+      "step": 20869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0684194002212287e-07,
+      "loss": 0.8696,
+      "step": 20870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.066034426294523e-07,
+      "loss": 0.8511,
+      "step": 20871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0636503502885386e-07,
+      "loss": 0.8604,
+      "step": 20872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0612671722488884e-07,
+      "loss": 0.8687,
+      "step": 20873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.058884892221159e-07,
+      "loss": 0.8149,
+      "step": 20874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0565035102509355e-07,
+      "loss": 0.8081,
+      "step": 20875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.054123026383765e-07,
+      "loss": 0.835,
+      "step": 20876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0517434406652005e-07,
+      "loss": 0.8081,
+      "step": 20877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0493647531407765e-07,
+      "loss": 0.8091,
+      "step": 20878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0469869638559857e-07,
+      "loss": 0.873,
+      "step": 20879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0446100728563186e-07,
+      "loss": 0.8638,
+      "step": 20880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.042234080187251e-07,
+      "loss": 0.792,
+      "step": 20881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.039858985894245e-07,
+      "loss": 0.8408,
+      "step": 20882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.037484790022738e-07,
+      "loss": 0.8081,
+      "step": 20883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.035111492618148e-07,
+      "loss": 0.8105,
+      "step": 20884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.032739093725884e-07,
+      "loss": 0.8516,
+      "step": 20885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0303675933913366e-07,
+      "loss": 0.8057,
+      "step": 20886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.027996991659865e-07,
+      "loss": 0.8013,
+      "step": 20887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.025627288576838e-07,
+      "loss": 0.8423,
+      "step": 20888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.023258484187569e-07,
+      "loss": 0.8179,
+      "step": 20889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0208905785373996e-07,
+      "loss": 0.8525,
+      "step": 20890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.018523571671628e-07,
+      "loss": 0.3281,
+      "step": 20891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0161574636355326e-07,
+      "loss": 0.7727,
+      "step": 20892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.013792254474374e-07,
+      "loss": 0.7827,
+      "step": 20893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.011427944233419e-07,
+      "loss": 0.7788,
+      "step": 20894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0090645329578894e-07,
+      "loss": 0.875,
+      "step": 20895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.006702020693003e-07,
+      "loss": 0.8491,
+      "step": 20896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.004340407483958e-07,
+      "loss": 0.7783,
+      "step": 20897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.001979693375945e-07,
+      "loss": 0.7891,
+      "step": 20898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.999619878414117e-07,
+      "loss": 0.7871,
+      "step": 20899
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9972609626436165e-07,
+      "loss": 0.8384,
+      "step": 20900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9949029461095904e-07,
+      "loss": 0.9121,
+      "step": 20901
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9925458288571295e-07,
+      "loss": 0.8354,
+      "step": 20902
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.990189610931343e-07,
+      "loss": 0.8228,
+      "step": 20903
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9878342923773174e-07,
+      "loss": 0.8765,
+      "step": 20904
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9854798732401047e-07,
+      "loss": 0.8511,
+      "step": 20905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9831263535647414e-07,
+      "loss": 0.8525,
+      "step": 20906
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9807737333962636e-07,
+      "loss": 0.8398,
+      "step": 20907
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.978422012779669e-07,
+      "loss": 0.8179,
+      "step": 20908
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9760711917599715e-07,
+      "loss": 0.8457,
+      "step": 20909
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9737212703821184e-07,
+      "loss": 0.8496,
+      "step": 20910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9713722486910955e-07,
+      "loss": 0.8276,
+      "step": 20911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.969024126731823e-07,
+      "loss": 0.7378,
+      "step": 20912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.966676904549226e-07,
+      "loss": 0.8496,
+      "step": 20913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9643305821882185e-07,
+      "loss": 0.8018,
+      "step": 20914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.961985159693687e-07,
+      "loss": 0.8296,
+      "step": 20915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9596406371105005e-07,
+      "loss": 0.8418,
+      "step": 20916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.957297014483518e-07,
+      "loss": 0.749,
+      "step": 20917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9549542918575756e-07,
+      "loss": 0.8325,
+      "step": 20918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.952612469277488e-07,
+      "loss": 0.8354,
+      "step": 20919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.950271546788064e-07,
+      "loss": 0.8789,
+      "step": 20920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.947931524434078e-07,
+      "loss": 0.3752,
+      "step": 20921
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.945592402260311e-07,
+      "loss": 0.8599,
+      "step": 20922
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9432541803115114e-07,
+      "loss": 0.7607,
+      "step": 20923
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9409168586324143e-07,
+      "loss": 0.8232,
+      "step": 20924
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.938580437267735e-07,
+      "loss": 0.833,
+      "step": 20925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9362449162621653e-07,
+      "loss": 0.8276,
+      "step": 20926
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9339102956603917e-07,
+      "loss": 0.8066,
+      "step": 20927
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.931576575507089e-07,
+      "loss": 0.8608,
+      "step": 20928
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.929243755846889e-07,
+      "loss": 0.8364,
+      "step": 20929
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9269118367244385e-07,
+      "loss": 0.8623,
+      "step": 20930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9245808181843414e-07,
+      "loss": 0.8521,
+      "step": 20931
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.922250700271184e-07,
+      "loss": 0.7979,
+      "step": 20932
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9199214830295697e-07,
+      "loss": 0.8262,
+      "step": 20933
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.917593166504035e-07,
+      "loss": 0.811,
+      "step": 20934
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.915265750739138e-07,
+      "loss": 0.8091,
+      "step": 20935
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9129392357794106e-07,
+      "loss": 0.8843,
+      "step": 20936
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.910613621669356e-07,
+      "loss": 0.8213,
+      "step": 20937
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.908288908453466e-07,
+      "loss": 0.8921,
+      "step": 20938
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9059650961762054e-07,
+      "loss": 0.9395,
+      "step": 20939
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.90364218488205e-07,
+      "loss": 0.8926,
+      "step": 20940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.901320174615435e-07,
+      "loss": 0.8647,
+      "step": 20941
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.898999065420782e-07,
+      "loss": 0.8667,
+      "step": 20942
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8966788573425054e-07,
+      "loss": 0.8467,
+      "step": 20943
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.894359550424986e-07,
+      "loss": 0.748,
+      "step": 20944
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.892041144712587e-07,
+      "loss": 0.8271,
+      "step": 20945
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8897236402496865e-07,
+      "loss": 0.4004,
+      "step": 20946
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.887407037080597e-07,
+      "loss": 0.8413,
+      "step": 20947
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.885091335249657e-07,
+      "loss": 0.8296,
+      "step": 20948
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8827765348011685e-07,
+      "loss": 0.8457,
+      "step": 20949
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8804626357794073e-07,
+      "loss": 0.7676,
+      "step": 20950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8781496382286444e-07,
+      "loss": 0.8325,
+      "step": 20951
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.875837542193144e-07,
+      "loss": 0.855,
+      "step": 20952
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8735263477171204e-07,
+      "loss": 0.8184,
+      "step": 20953
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8712160548448055e-07,
+      "loss": 0.8086,
+      "step": 20954
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8689066636203854e-07,
+      "loss": 0.7852,
+      "step": 20955
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8665981740880534e-07,
+      "loss": 0.3286,
+      "step": 20956
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8642905862919844e-07,
+      "loss": 0.8291,
+      "step": 20957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8619839002762994e-07,
+      "loss": 0.8574,
+      "step": 20958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8596781160851404e-07,
+      "loss": 0.813,
+      "step": 20959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8573732337626283e-07,
+      "loss": 0.8433,
+      "step": 20960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.855069253352849e-07,
+      "loss": 0.7368,
+      "step": 20961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.852766174899896e-07,
+      "loss": 0.8228,
+      "step": 20962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.850463998447811e-07,
+      "loss": 0.8008,
+      "step": 20963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.848162724040654e-07,
+      "loss": 0.8403,
+      "step": 20964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8458623517224503e-07,
+      "loss": 0.8457,
+      "step": 20965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.843562881537193e-07,
+      "loss": 0.8994,
+      "step": 20966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.841264313528902e-07,
+      "loss": 0.8379,
+      "step": 20967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8389666477415255e-07,
+      "loss": 0.8086,
+      "step": 20968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8366698842190285e-07,
+      "loss": 0.8804,
+      "step": 20969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.834374023005371e-07,
+      "loss": 0.8052,
+      "step": 20970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8320790641444605e-07,
+      "loss": 0.8013,
+      "step": 20971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8297850076801967e-07,
+      "loss": 0.7734,
+      "step": 20972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.827491853656489e-07,
+      "loss": 0.8916,
+      "step": 20973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8251996021171846e-07,
+      "loss": 0.8032,
+      "step": 20974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8229082531061545e-07,
+      "loss": 0.8384,
+      "step": 20975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8206178066672306e-07,
+      "loss": 0.8174,
+      "step": 20976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.818328262844239e-07,
+      "loss": 0.8301,
+      "step": 20977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8160396216809714e-07,
+      "loss": 0.8545,
+      "step": 20978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.813751883221216e-07,
+      "loss": 0.7974,
+      "step": 20979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8114650475087435e-07,
+      "loss": 0.8389,
+      "step": 20980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.809179114587313e-07,
+      "loss": 0.8232,
+      "step": 20981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.806894084500639e-07,
+      "loss": 0.4595,
+      "step": 20982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.804609957292459e-07,
+      "loss": 0.8291,
+      "step": 20983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8023267330064553e-07,
+      "loss": 0.895,
+      "step": 20984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.800044411686315e-07,
+      "loss": 0.7905,
+      "step": 20985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.797762993375708e-07,
+      "loss": 0.8115,
+      "step": 20986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7954824781182664e-07,
+      "loss": 0.8501,
+      "step": 20987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.793202865957634e-07,
+      "loss": 0.8179,
+      "step": 20988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7909241569374245e-07,
+      "loss": 0.9058,
+      "step": 20989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.788646351101226e-07,
+      "loss": 0.874,
+      "step": 20990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7863694484926206e-07,
+      "loss": 0.7637,
+      "step": 20991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.784093449155162e-07,
+      "loss": 0.8608,
+      "step": 20992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7818183531323983e-07,
+      "loss": 0.8877,
+      "step": 20993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.779544160467862e-07,
+      "loss": 0.8301,
+      "step": 20994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.777270871205046e-07,
+      "loss": 0.832,
+      "step": 20995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7749984853874643e-07,
+      "loss": 0.813,
+      "step": 20996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.772727003058578e-07,
+      "loss": 0.8735,
+      "step": 20997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.770456424261836e-07,
+      "loss": 0.8511,
+      "step": 20998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.768186749040697e-07,
+      "loss": 0.7859,
+      "step": 20999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7659179774385647e-07,
+      "loss": 0.8628,
+      "step": 21000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.76365010949885e-07,
+      "loss": 0.8413,
+      "step": 21001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.761383145264956e-07,
+      "loss": 0.3215,
+      "step": 21002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7591170847802383e-07,
+      "loss": 0.8174,
+      "step": 21003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.756851928088056e-07,
+      "loss": 0.8208,
+      "step": 21004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.754587675231729e-07,
+      "loss": 0.8389,
+      "step": 21005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.752324326254596e-07,
+      "loss": 0.7856,
+      "step": 21006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7500618811999557e-07,
+      "loss": 0.8066,
+      "step": 21007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.747800340111079e-07,
+      "loss": 0.8789,
+      "step": 21008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7455397030312523e-07,
+      "loss": 0.7881,
+      "step": 21009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.743279970003715e-07,
+      "loss": 0.8062,
+      "step": 21010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7410211410716925e-07,
+      "loss": 0.8906,
+      "step": 21011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7387632162784116e-07,
+      "loss": 0.7554,
+      "step": 21012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73650619566706e-07,
+      "loss": 0.8584,
+      "step": 21013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73425007928082e-07,
+      "loss": 0.7725,
+      "step": 21014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731994867162868e-07,
+      "loss": 0.7788,
+      "step": 21015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7297405593563363e-07,
+      "loss": 0.8113,
+      "step": 21016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.727487155904351e-07,
+      "loss": 0.8291,
+      "step": 21017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.725234656850034e-07,
+      "loss": 0.8828,
+      "step": 21018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7229830622364673e-07,
+      "loss": 0.8472,
+      "step": 21019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7207323721067437e-07,
+      "loss": 0.7783,
+      "step": 21020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.718482586503901e-07,
+      "loss": 0.8218,
+      "step": 21021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7162337054710053e-07,
+      "loss": 0.8501,
+      "step": 21022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.713985729051066e-07,
+      "loss": 0.8486,
+      "step": 21023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7117386572870827e-07,
+      "loss": 0.8022,
+      "step": 21024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.70949249022206e-07,
+      "loss": 0.8521,
+      "step": 21025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7072472278989735e-07,
+      "loss": 0.8125,
+      "step": 21026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.705002870360762e-07,
+      "loss": 0.7449,
+      "step": 21027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7027594176503857e-07,
+      "loss": 0.8232,
+      "step": 21028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.700516869810749e-07,
+      "loss": 0.8135,
+      "step": 21029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.69827522688475e-07,
+      "loss": 0.8032,
+      "step": 21030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6960344889152943e-07,
+      "loss": 0.8354,
+      "step": 21031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6937946559452365e-07,
+      "loss": 0.8848,
+      "step": 21032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6915557280174256e-07,
+      "loss": 0.8564,
+      "step": 21033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6893177051747157e-07,
+      "loss": 0.8037,
+      "step": 21034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6870805874599057e-07,
+      "loss": 0.8765,
+      "step": 21035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6848443749158016e-07,
+      "loss": 0.8169,
+      "step": 21036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.682609067585179e-07,
+      "loss": 0.8003,
+      "step": 21037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.680374665510804e-07,
+      "loss": 0.8022,
+      "step": 21038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6781411687354377e-07,
+      "loss": 0.8369,
+      "step": 21039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6759085773017947e-07,
+      "loss": 0.3755,
+      "step": 21040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.673676891252597e-07,
+      "loss": 0.8311,
+      "step": 21041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6714461106305434e-07,
+      "loss": 0.8447,
+      "step": 21042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.669216235478295e-07,
+      "loss": 0.9053,
+      "step": 21043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.666987265838533e-07,
+      "loss": 0.8013,
+      "step": 21044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6647592017538803e-07,
+      "loss": 0.814,
+      "step": 21045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.66253204326698e-07,
+      "loss": 0.8403,
+      "step": 21046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.660305790420442e-07,
+      "loss": 0.8984,
+      "step": 21047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.658080443256855e-07,
+      "loss": 0.8091,
+      "step": 21048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.655856001818785e-07,
+      "loss": 0.8521,
+      "step": 21049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6536324661487923e-07,
+      "loss": 0.8394,
+      "step": 21050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6514098362894205e-07,
+      "loss": 0.8452,
+      "step": 21051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.649188112283191e-07,
+      "loss": 0.7583,
+      "step": 21052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.646967294172609e-07,
+      "loss": 0.8242,
+      "step": 21053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.644747382000162e-07,
+      "loss": 0.8022,
+      "step": 21054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.642528375808323e-07,
+      "loss": 0.8184,
+      "step": 21055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6403102756395404e-07,
+      "loss": 0.8018,
+      "step": 21056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6380930815362526e-07,
+      "loss": 0.8398,
+      "step": 21057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6358767935408703e-07,
+      "loss": 0.8838,
+      "step": 21058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.633661411695804e-07,
+      "loss": 0.813,
+      "step": 21059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6314469360434413e-07,
+      "loss": 0.8271,
+      "step": 21060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.629233366626144e-07,
+      "loss": 0.7969,
+      "step": 21061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.627020703486249e-07,
+      "loss": 0.8335,
+      "step": 21062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6248089466661077e-07,
+      "loss": 0.8809,
+      "step": 21063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.622598096208018e-07,
+      "loss": 0.7993,
+      "step": 21064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.620388152154291e-07,
+      "loss": 0.8579,
+      "step": 21065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6181791145471867e-07,
+      "loss": 0.8857,
+      "step": 21066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.615970983428989e-07,
+      "loss": 0.8169,
+      "step": 21067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.613763758841947e-07,
+      "loss": 0.8101,
+      "step": 21068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6115574408282607e-07,
+      "loss": 0.834,
+      "step": 21069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6093520294301567e-07,
+      "loss": 0.9087,
+      "step": 21070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.607147524689829e-07,
+      "loss": 0.8188,
+      "step": 21071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.604943926649445e-07,
+      "loss": 0.8203,
+      "step": 21072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6027412353511806e-07,
+      "loss": 0.8213,
+      "step": 21073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.600539450837153e-07,
+      "loss": 0.8623,
+      "step": 21074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.598338573149506e-07,
+      "loss": 0.8252,
+      "step": 21075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.596138602330339e-07,
+      "loss": 0.9561,
+      "step": 21076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5939395384217356e-07,
+      "loss": 0.8677,
+      "step": 21077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5917413814657734e-07,
+      "loss": 0.7788,
+      "step": 21078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.589544131504501e-07,
+      "loss": 0.8618,
+      "step": 21079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.587347788579958e-07,
+      "loss": 0.812,
+      "step": 21080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5851523527341714e-07,
+      "loss": 0.8135,
+      "step": 21081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.582957824009141e-07,
+      "loss": 0.8232,
+      "step": 21082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.580764202446834e-07,
+      "loss": 0.8071,
+      "step": 21083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5785714880892434e-07,
+      "loss": 0.8901,
+      "step": 21084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.576379680978297e-07,
+      "loss": 0.9033,
+      "step": 21085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.574188781155945e-07,
+      "loss": 0.824,
+      "step": 21086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5719987886640875e-07,
+      "loss": 0.9136,
+      "step": 21087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56980970354464e-07,
+      "loss": 0.835,
+      "step": 21088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56762152583947e-07,
+      "loss": 0.8242,
+      "step": 21089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.565434255590443e-07,
+      "loss": 0.8608,
+      "step": 21090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.563247892839399e-07,
+      "loss": 0.7759,
+      "step": 21091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5610624376281867e-07,
+      "loss": 0.8765,
+      "step": 21092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.558877889998596e-07,
+      "loss": 0.7988,
+      "step": 21093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5566942499924364e-07,
+      "loss": 0.8853,
+      "step": 21094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5545115176514754e-07,
+      "loss": 0.8008,
+      "step": 21095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.552329693017469e-07,
+      "loss": 0.8501,
+      "step": 21096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5501487761321666e-07,
+      "loss": 0.8408,
+      "step": 21097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5479687670372845e-07,
+      "loss": 0.8228,
+      "step": 21098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5457896657745397e-07,
+      "loss": 0.8457,
+      "step": 21099
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.543611472385621e-07,
+      "loss": 0.4229,
+      "step": 21100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.541434186912195e-07,
+      "loss": 0.7847,
+      "step": 21101
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.539257809395918e-07,
+      "loss": 0.8467,
+      "step": 21102
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.537082339878427e-07,
+      "loss": 0.9258,
+      "step": 21103
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5349077784013353e-07,
+      "loss": 0.8511,
+      "step": 21104
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.532734125006264e-07,
+      "loss": 0.8203,
+      "step": 21105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5305613797347806e-07,
+      "loss": 0.8013,
+      "step": 21106
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5283895426284676e-07,
+      "loss": 0.8242,
+      "step": 21107
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.52621861372887e-07,
+      "loss": 0.8169,
+      "step": 21108
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5240485930775106e-07,
+      "loss": 0.8271,
+      "step": 21109
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5218794807159276e-07,
+      "loss": 0.8193,
+      "step": 21110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5197112766855936e-07,
+      "loss": 0.8184,
+      "step": 21111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5175439810280035e-07,
+      "loss": 0.8862,
+      "step": 21112
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.515377593784624e-07,
+      "loss": 0.8555,
+      "step": 21113
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.513212114996899e-07,
+      "loss": 0.7758,
+      "step": 21114
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5110475447062575e-07,
+      "loss": 0.8013,
+      "step": 21115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.508883882954105e-07,
+      "loss": 0.8916,
+      "step": 21116
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.506721129781836e-07,
+      "loss": 0.9033,
+      "step": 21117
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5045592852308454e-07,
+      "loss": 0.793,
+      "step": 21118
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5023983493424667e-07,
+      "loss": 0.8389,
+      "step": 21119
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5002383221580604e-07,
+      "loss": 0.8374,
+      "step": 21120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4980792037189507e-07,
+      "loss": 0.7998,
+      "step": 21121
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.495920994066431e-07,
+      "loss": 0.9067,
+      "step": 21122
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.493763693241802e-07,
+      "loss": 0.8945,
+      "step": 21123
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4916073012863306e-07,
+      "loss": 0.7861,
+      "step": 21124
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4894518182412786e-07,
+      "loss": 0.8306,
+      "step": 21125
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4872972441478794e-07,
+      "loss": 0.8354,
+      "step": 21126
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.485143579047361e-07,
+      "loss": 0.8921,
+      "step": 21127
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.482990822980913e-07,
+      "loss": 0.8428,
+      "step": 21128
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.48083897598973e-07,
+      "loss": 0.8335,
+      "step": 21129
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.478688038114979e-07,
+      "loss": 0.8516,
+      "step": 21130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.476538009397811e-07,
+      "loss": 0.8472,
+      "step": 21131
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4743888898793533e-07,
+      "loss": 0.8271,
+      "step": 21132
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4722406796007346e-07,
+      "loss": 0.8447,
+      "step": 21133
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4700933786030443e-07,
+      "loss": 0.8252,
+      "step": 21134
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.467946986927361e-07,
+      "loss": 0.8296,
+      "step": 21135
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.465801504614762e-07,
+      "loss": 0.8359,
+      "step": 21136
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4636569317062764e-07,
+      "loss": 0.7793,
+      "step": 21137
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.461513268242938e-07,
+      "loss": 0.8643,
+      "step": 21138
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4593705142657684e-07,
+      "loss": 0.8301,
+      "step": 21139
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.457228669815759e-07,
+      "loss": 0.8359,
+      "step": 21140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.455087734933875e-07,
+      "loss": 0.8071,
+      "step": 21141
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4529477096610967e-07,
+      "loss": 0.8174,
+      "step": 21142
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.45080859403834e-07,
+      "loss": 0.8423,
+      "step": 21143
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.44867038810655e-07,
+      "loss": 0.877,
+      "step": 21144
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.446533091906622e-07,
+      "loss": 0.895,
+      "step": 21145
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.444396705479457e-07,
+      "loss": 0.8403,
+      "step": 21146
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.442261228865922e-07,
+      "loss": 0.855,
+      "step": 21147
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4401266621068675e-07,
+      "loss": 0.8062,
+      "step": 21148
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4379930052431334e-07,
+      "loss": 0.8926,
+      "step": 21149
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4358602583155475e-07,
+      "loss": 0.8203,
+      "step": 21150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.433728421364895e-07,
+      "loss": 0.8125,
+      "step": 21151
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4315974944319866e-07,
+      "loss": 0.856,
+      "step": 21152
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.429467477557573e-07,
+      "loss": 0.8999,
+      "step": 21153
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4273383707824006e-07,
+      "loss": 0.8296,
+      "step": 21154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4252101741472243e-07,
+      "loss": 0.8877,
+      "step": 21155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4230828876927293e-07,
+      "loss": 0.8057,
+      "step": 21156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4209565114596377e-07,
+      "loss": 0.8872,
+      "step": 21157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.418831045488629e-07,
+      "loss": 0.8657,
+      "step": 21158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4167064898203585e-07,
+      "loss": 0.8818,
+      "step": 21159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4145828444954777e-07,
+      "loss": 0.814,
+      "step": 21160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.412460109554604e-07,
+      "loss": 0.8022,
+      "step": 21161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.410338285038355e-07,
+      "loss": 0.8818,
+      "step": 21162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4082173709873366e-07,
+      "loss": 0.8545,
+      "step": 21163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4060973674421064e-07,
+      "loss": 0.4773,
+      "step": 21164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4039782744432417e-07,
+      "loss": 0.8608,
+      "step": 21165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.401860092031272e-07,
+      "loss": 0.8486,
+      "step": 21166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3997428202467144e-07,
+      "loss": 0.8008,
+      "step": 21167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.397626459130098e-07,
+      "loss": 0.9136,
+      "step": 21168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3955110087218905e-07,
+      "loss": 0.8428,
+      "step": 21169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3933964690625697e-07,
+      "loss": 0.7798,
+      "step": 21170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.391282840192605e-07,
+      "loss": 0.877,
+      "step": 21171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3891701221524175e-07,
+      "loss": 0.7559,
+      "step": 21172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.387058314982432e-07,
+      "loss": 0.8867,
+      "step": 21173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.384947418723044e-07,
+      "loss": 0.8237,
+      "step": 21174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3828374334146475e-07,
+      "loss": 0.855,
+      "step": 21175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.380728359097606e-07,
+      "loss": 0.8032,
+      "step": 21176
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3786201958122644e-07,
+      "loss": 0.8457,
+      "step": 21177
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.376512943598963e-07,
+      "loss": 0.7866,
+      "step": 21178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.374406602498025e-07,
+      "loss": 0.8462,
+      "step": 21179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3723011725497292e-07,
+      "loss": 0.752,
+      "step": 21180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3701966537943655e-07,
+      "loss": 0.8511,
+      "step": 21181
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3680930462721906e-07,
+      "loss": 0.8804,
+      "step": 21182
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3659903500234504e-07,
+      "loss": 0.8325,
+      "step": 21183
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3638885650883902e-07,
+      "loss": 0.8726,
+      "step": 21184
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3617876915071948e-07,
+      "loss": 0.8594,
+      "step": 21185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.359687729320076e-07,
+      "loss": 0.7974,
+      "step": 21186
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3575886785672076e-07,
+      "loss": 0.8574,
+      "step": 21187
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3554905392887295e-07,
+      "loss": 0.7876,
+      "step": 21188
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.35339331152481e-07,
+      "loss": 0.5264,
+      "step": 21189
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.351296995315544e-07,
+      "loss": 0.7935,
+      "step": 21190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3492015907010556e-07,
+      "loss": 0.8584,
+      "step": 21191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.347107097721435e-07,
+      "loss": 0.8457,
+      "step": 21192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.34501351641675e-07,
+      "loss": 0.8652,
+      "step": 21193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3429208468270403e-07,
+      "loss": 0.3701,
+      "step": 21194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3408290889923636e-07,
+      "loss": 0.8311,
+      "step": 21195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.338738242952715e-07,
+      "loss": 0.7864,
+      "step": 21196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3366483087481186e-07,
+      "loss": 0.8394,
+      "step": 21197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3345592864185417e-07,
+      "loss": 0.8755,
+      "step": 21198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3324711760039643e-07,
+      "loss": 0.8433,
+      "step": 21199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3303839775443314e-07,
+      "loss": 0.8286,
+      "step": 21200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3282976910795619e-07,
+      "loss": 0.7983,
+      "step": 21201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3262123166495842e-07,
+      "loss": 0.8652,
+      "step": 21202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3241278542942836e-07,
+      "loss": 0.8569,
+      "step": 21203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3220443040535445e-07,
+      "loss": 0.8218,
+      "step": 21204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3199616659672352e-07,
+      "loss": 0.7998,
+      "step": 21205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3178799400751905e-07,
+      "loss": 0.8799,
+      "step": 21206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3157991264172398e-07,
+      "loss": 0.8369,
+      "step": 21207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3137192250331952e-07,
+      "loss": 0.8315,
+      "step": 21208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.311640235962842e-07,
+      "loss": 0.8628,
+      "step": 21209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.309562159245965e-07,
+      "loss": 0.3452,
+      "step": 21210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3074849949223044e-07,
+      "loss": 0.8633,
+      "step": 21211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3054087430316174e-07,
+      "loss": 0.813,
+      "step": 21212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3033334036136225e-07,
+      "loss": 0.8477,
+      "step": 21213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3012589767080097e-07,
+      "loss": 0.8428,
+      "step": 21214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2991854623544806e-07,
+      "loss": 0.8911,
+      "step": 21215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.297112860592704e-07,
+      "loss": 0.853,
+      "step": 21216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2950411714623255e-07,
+      "loss": 0.8906,
+      "step": 21217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2929703950029858e-07,
+      "loss": 0.8584,
+      "step": 21218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.290900531254303e-07,
+      "loss": 0.8691,
+      "step": 21219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2888315802558626e-07,
+      "loss": 0.8408,
+      "step": 21220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2867635420472656e-07,
+      "loss": 0.7788,
+      "step": 21221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2846964166680641e-07,
+      "loss": 0.832,
+      "step": 21222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2826302041578098e-07,
+      "loss": 0.7866,
+      "step": 21223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2805649045560374e-07,
+      "loss": 0.8525,
+      "step": 21224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2785005179022546e-07,
+      "loss": 0.3297,
+      "step": 21225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2764370442359573e-07,
+      "loss": 0.8198,
+      "step": 21226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2743744835966198e-07,
+      "loss": 0.7944,
+      "step": 21227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.272312836023699e-07,
+      "loss": 0.8457,
+      "step": 21228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2702521015566526e-07,
+      "loss": 0.7671,
+      "step": 21229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2681922802348877e-07,
+      "loss": 0.8398,
+      "step": 21230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2661333720978285e-07,
+      "loss": 0.8608,
+      "step": 21231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2640753771848544e-07,
+      "loss": 0.834,
+      "step": 21232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2620182955353398e-07,
+      "loss": 0.8242,
+      "step": 21233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.259962127188642e-07,
+      "loss": 0.8965,
+      "step": 21234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2579068721840958e-07,
+      "loss": 0.8359,
+      "step": 21235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2558525305610202e-07,
+      "loss": 0.8013,
+      "step": 21236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.253799102358728e-07,
+      "loss": 0.8369,
+      "step": 21237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.251746587616499e-07,
+      "loss": 0.8628,
+      "step": 21238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2496949863735962e-07,
+      "loss": 0.8672,
+      "step": 21239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2476442986692714e-07,
+      "loss": 0.8398,
+      "step": 21240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2455945245427602e-07,
+      "loss": 0.8135,
+      "step": 21241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.243545664033281e-07,
+      "loss": 0.7632,
+      "step": 21242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.241497717180019e-07,
+      "loss": 0.8569,
+      "step": 21243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2394506840221764e-07,
+      "loss": 0.8569,
+      "step": 21244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2374045645988995e-07,
+      "loss": 0.8467,
+      "step": 21245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2353593589493295e-07,
+      "loss": 0.7803,
+      "step": 21246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2333150671126125e-07,
+      "loss": 0.8638,
+      "step": 21247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2312716891278397e-07,
+      "loss": 0.7979,
+      "step": 21248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2292292250341185e-07,
+      "loss": 0.8232,
+      "step": 21249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2271876748705233e-07,
+      "loss": 0.8154,
+      "step": 21250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2251470386761065e-07,
+      "loss": 0.8208,
+      "step": 21251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2231073164899086e-07,
+      "loss": 0.8193,
+      "step": 21252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2210685083509598e-07,
+      "loss": 0.7954,
+      "step": 21253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.219030614298251e-07,
+      "loss": 0.8047,
+      "step": 21254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2169936343707898e-07,
+      "loss": 0.8257,
+      "step": 21255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2149575686075287e-07,
+      "loss": 0.8008,
+      "step": 21256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.212922417047436e-07,
+      "loss": 0.8975,
+      "step": 21257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2108881797294424e-07,
+      "loss": 0.8232,
+      "step": 21258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.208854856692455e-07,
+      "loss": 0.7861,
+      "step": 21259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.206822447975382e-07,
+      "loss": 0.8145,
+      "step": 21260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2047909536171197e-07,
+      "loss": 0.8755,
+      "step": 21261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.202760373656515e-07,
+      "loss": 0.8472,
+      "step": 21262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2007307081324258e-07,
+      "loss": 0.8193,
+      "step": 21263
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1987019570836822e-07,
+      "loss": 0.8071,
+      "step": 21264
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1966741205490917e-07,
+      "loss": 0.834,
+      "step": 21265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1946471985674622e-07,
+      "loss": 0.8379,
+      "step": 21266
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1926211911775519e-07,
+      "loss": 0.7993,
+      "step": 21267
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1905960984181352e-07,
+      "loss": 0.349,
+      "step": 21268
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1885719203279587e-07,
+      "loss": 0.8096,
+      "step": 21269
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1865486569457473e-07,
+      "loss": 0.8359,
+      "step": 21270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1845263083101975e-07,
+      "loss": 0.8564,
+      "step": 21271
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1825048744600062e-07,
+      "loss": 0.8037,
+      "step": 21272
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1804843554338483e-07,
+      "loss": 0.8252,
+      "step": 21273
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1784647512703817e-07,
+      "loss": 0.8027,
+      "step": 21274
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.176446062008236e-07,
+      "loss": 0.8921,
+      "step": 21275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.174428287686048e-07,
+      "loss": 0.7788,
+      "step": 21276
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.172411428342408e-07,
+      "loss": 0.8018,
+      "step": 21277
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.170395484015897e-07,
+      "loss": 0.8364,
+      "step": 21278
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1683804547450948e-07,
+      "loss": 0.8286,
+      "step": 21279
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.166366340568543e-07,
+      "loss": 0.7871,
+      "step": 21280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.164353141524783e-07,
+      "loss": 0.8286,
+      "step": 21281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1623408576523285e-07,
+      "loss": 0.8237,
+      "step": 21282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.160329488989682e-07,
+      "loss": 0.8462,
+      "step": 21283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1583190355753126e-07,
+      "loss": 0.8335,
+      "step": 21284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.156309497447684e-07,
+      "loss": 0.8833,
+      "step": 21285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1543008746452487e-07,
+      "loss": 0.8198,
+      "step": 21286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1522931672064429e-07,
+      "loss": 0.8345,
+      "step": 21287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1502863751696524e-07,
+      "loss": 0.8828,
+      "step": 21288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1482804985732907e-07,
+      "loss": 0.8477,
+      "step": 21289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1462755374557386e-07,
+      "loss": 0.9038,
+      "step": 21290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1442714918553376e-07,
+      "loss": 0.8335,
+      "step": 21291
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1422683618104346e-07,
+      "loss": 0.8101,
+      "step": 21292
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1402661473593432e-07,
+      "loss": 0.9053,
+      "step": 21293
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1382648485403834e-07,
+      "loss": 0.793,
+      "step": 21294
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1362644653918408e-07,
+      "loss": 0.8687,
+      "step": 21295
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1342649979519792e-07,
+      "loss": 0.873,
+      "step": 21296
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.132266446259057e-07,
+      "loss": 0.8188,
+      "step": 21297
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1302688103513102e-07,
+      "loss": 0.7969,
+      "step": 21298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.128272090266953e-07,
+      "loss": 0.8325,
+      "step": 21299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1262762860441877e-07,
+      "loss": 0.8926,
+      "step": 21300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1242813977211895e-07,
+      "loss": 0.8555,
+      "step": 21301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1222874253361337e-07,
+      "loss": 0.8218,
+      "step": 21302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1202943689271726e-07,
+      "loss": 0.8711,
+      "step": 21303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1183022285324262e-07,
+      "loss": 0.8613,
+      "step": 21304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1163110041900135e-07,
+      "loss": 0.8774,
+      "step": 21305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1143206959380215e-07,
+      "loss": 0.8477,
+      "step": 21306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.11233130381453e-07,
+      "loss": 0.792,
+      "step": 21307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1103428278576089e-07,
+      "loss": 0.7856,
+      "step": 21308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.108355268105283e-07,
+      "loss": 0.8647,
+      "step": 21309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1063686245956007e-07,
+      "loss": 0.7783,
+      "step": 21310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.104382897366558e-07,
+      "loss": 0.7861,
+      "step": 21311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1023980864561367e-07,
+      "loss": 0.8018,
+      "step": 21312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1004141919023225e-07,
+      "loss": 0.8208,
+      "step": 21313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.098431213743063e-07,
+      "loss": 0.8862,
+      "step": 21314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0964491520162945e-07,
+      "loss": 0.8589,
+      "step": 21315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0944680067599476e-07,
+      "loss": 0.8457,
+      "step": 21316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0924877780119145e-07,
+      "loss": 0.8169,
+      "step": 21317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0905084658100815e-07,
+      "loss": 0.8477,
+      "step": 21318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0885300701923238e-07,
+      "loss": 0.3694,
+      "step": 21319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0865525911964724e-07,
+      "loss": 0.8594,
+      "step": 21320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0845760288603856e-07,
+      "loss": 0.8345,
+      "step": 21321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0826003832218554e-07,
+      "loss": 0.3051,
+      "step": 21322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0806256543186908e-07,
+      "loss": 0.791,
+      "step": 21323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.078651842188678e-07,
+      "loss": 0.8408,
+      "step": 21324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0766789468695536e-07,
+      "loss": 0.8145,
+      "step": 21325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.074706968399087e-07,
+      "loss": 0.8569,
+      "step": 21326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0727359068149987e-07,
+      "loss": 0.7585,
+      "step": 21327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0707657621549915e-07,
+      "loss": 0.8398,
+      "step": 21328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0687965344567684e-07,
+      "loss": 0.7993,
+      "step": 21329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0668282237579995e-07,
+      "loss": 0.8179,
+      "step": 21330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0648608300963325e-07,
+      "loss": 0.8096,
+      "step": 21331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0628943535094205e-07,
+      "loss": 0.8208,
+      "step": 21332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0609287940348722e-07,
+      "loss": 0.7881,
+      "step": 21333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0589641517102964e-07,
+      "loss": 0.8154,
+      "step": 21334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0570004265732913e-07,
+      "loss": 0.3162,
+      "step": 21335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0550376186614153e-07,
+      "loss": 0.813,
+      "step": 21336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0530757280122216e-07,
+      "loss": 0.8223,
+      "step": 21337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0511147546632416e-07,
+      "loss": 0.8638,
+      "step": 21338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0491546986519896e-07,
+      "loss": 0.8745,
+      "step": 21339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0471955600159798e-07,
+      "loss": 0.8662,
+      "step": 21340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0452373387926715e-07,
+      "loss": 0.8232,
+      "step": 21341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0432800350195515e-07,
+      "loss": 0.8721,
+      "step": 21342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0413236487340504e-07,
+      "loss": 0.8232,
+      "step": 21343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0393681799736e-07,
+      "loss": 0.8218,
+      "step": 21344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0374136287756196e-07,
+      "loss": 0.8184,
+      "step": 21345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0354599951774856e-07,
+      "loss": 0.8149,
+      "step": 21346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0335072792165845e-07,
+      "loss": 0.8394,
+      "step": 21347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0315554809302862e-07,
+      "loss": 0.8833,
+      "step": 21348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0296046003559168e-07,
+      "loss": 0.835,
+      "step": 21349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0276546375308015e-07,
+      "loss": 0.8628,
+      "step": 21350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0257055924922442e-07,
+      "loss": 0.8037,
+      "step": 21351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0237574652775371e-07,
+      "loss": 0.8169,
+      "step": 21352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0218102559239561e-07,
+      "loss": 0.8384,
+      "step": 21353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0198639644687433e-07,
+      "loss": 0.7737,
+      "step": 21354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0179185909491473e-07,
+      "loss": 0.8672,
+      "step": 21355
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0159741354023766e-07,
+      "loss": 0.8198,
+      "step": 21356
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0140305978656294e-07,
+      "loss": 0.873,
+      "step": 21357
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0120879783760983e-07,
+      "loss": 0.8306,
+      "step": 21358
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0101462769709425e-07,
+      "loss": 0.813,
+      "step": 21359
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.008205493687304e-07,
+      "loss": 0.8418,
+      "step": 21360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0062656285623317e-07,
+      "loss": 0.802,
+      "step": 21361
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.004326681633123e-07,
+      "loss": 0.8374,
+      "step": 21362
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0023886529367708e-07,
+      "loss": 0.8193,
+      "step": 21363
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0004515425103676e-07,
+      "loss": 0.7832,
+      "step": 21364
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.998515350390956e-07,
+      "loss": 0.8096,
+      "step": 21365
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9965800766155952e-07,
+      "loss": 0.8032,
+      "step": 21366
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9946457212212945e-07,
+      "loss": 0.8647,
+      "step": 21367
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9927122842450741e-07,
+      "loss": 0.7905,
+      "step": 21368
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.990779765723916e-07,
+      "loss": 0.8477,
+      "step": 21369
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.98884816569479e-07,
+      "loss": 0.7993,
+      "step": 21370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.986917484194656e-07,
+      "loss": 0.8047,
+      "step": 21371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9849877212604507e-07,
+      "loss": 0.8086,
+      "step": 21372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9830588769290894e-07,
+      "loss": 0.8228,
+      "step": 21373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.981130951237481e-07,
+      "loss": 0.8589,
+      "step": 21374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9792039442225076e-07,
+      "loss": 0.8887,
+      "step": 21375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9772778559210281e-07,
+      "loss": 0.3435,
+      "step": 21376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9753526863699024e-07,
+      "loss": 0.8374,
+      "step": 21377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9734284356059564e-07,
+      "loss": 0.8667,
+      "step": 21378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9715051036660048e-07,
+      "loss": 0.8745,
+      "step": 21379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.969582690586841e-07,
+      "loss": 0.7603,
+      "step": 21380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9676611964052516e-07,
+      "loss": 0.8018,
+      "step": 21381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9657406211579966e-07,
+      "loss": 0.8232,
+      "step": 21382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9638209648818074e-07,
+      "loss": 0.8643,
+      "step": 21383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9619022276134159e-07,
+      "loss": 0.8613,
+      "step": 21384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9599844093895425e-07,
+      "loss": 0.8242,
+      "step": 21385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9580675102468638e-07,
+      "loss": 0.8423,
+      "step": 21386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9561515302220613e-07,
+      "loss": 0.8887,
+      "step": 21387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.954236469351789e-07,
+      "loss": 0.8364,
+      "step": 21388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9523223276726789e-07,
+      "loss": 0.7905,
+      "step": 21389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9504091052213625e-07,
+      "loss": 0.8716,
+      "step": 21390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9484968020344332e-07,
+      "loss": 0.791,
+      "step": 21391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9465854181484723e-07,
+      "loss": 0.8125,
+      "step": 21392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9446749536000675e-07,
+      "loss": 0.8325,
+      "step": 21393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9427654084257563e-07,
+      "loss": 0.8413,
+      "step": 21394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9408567826620705e-07,
+      "loss": 0.873,
+      "step": 21395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.93894907634552e-07,
+      "loss": 0.7891,
+      "step": 21396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9370422895126028e-07,
+      "loss": 0.834,
+      "step": 21397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.935136422199818e-07,
+      "loss": 0.7944,
+      "step": 21398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9332314744436086e-07,
+      "loss": 0.8486,
+      "step": 21399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9313274462804177e-07,
+      "loss": 0.8403,
+      "step": 21400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9294243377466993e-07,
+      "loss": 0.7778,
+      "step": 21401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9275221488788298e-07,
+      "loss": 0.8105,
+      "step": 21402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9256208797132137e-07,
+      "loss": 0.7944,
+      "step": 21403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9237205302862273e-07,
+      "loss": 0.8315,
+      "step": 21404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9218211006342246e-07,
+      "loss": 0.7568,
+      "step": 21405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9199225907935492e-07,
+      "loss": 0.8779,
+      "step": 21406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9180250008005163e-07,
+      "loss": 0.7852,
+      "step": 21407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9161283306914413e-07,
+      "loss": 0.7961,
+      "step": 21408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.914232580502595e-07,
+      "loss": 0.835,
+      "step": 21409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9123377502702489e-07,
+      "loss": 0.8345,
+      "step": 21410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.910443840030668e-07,
+      "loss": 0.8311,
+      "step": 21411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9085508498200678e-07,
+      "loss": 0.856,
+      "step": 21412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.906658779674675e-07,
+      "loss": 0.8398,
+      "step": 21413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9047676296306884e-07,
+      "loss": 0.8081,
+      "step": 21414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9028773997242844e-07,
+      "loss": 0.8169,
+      "step": 21415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9009880899916288e-07,
+      "loss": 0.8325,
+      "step": 21416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8990997004688648e-07,
+      "loss": 0.877,
+      "step": 21417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8972122311921136e-07,
+      "loss": 0.8618,
+      "step": 21418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8953256821975018e-07,
+      "loss": 0.7773,
+      "step": 21419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8934400535211117e-07,
+      "loss": 0.8809,
+      "step": 21420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8915553451990197e-07,
+      "loss": 0.8604,
+      "step": 21421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8896715572672863e-07,
+      "loss": 0.8174,
+      "step": 21422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8877886897619435e-07,
+      "loss": 0.7705,
+      "step": 21423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8859067427190237e-07,
+      "loss": 0.8809,
+      "step": 21424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.88402571617452e-07,
+      "loss": 0.8262,
+      "step": 21425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.882145610164421e-07,
+      "loss": 0.8804,
+      "step": 21426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8802664247247137e-07,
+      "loss": 0.8081,
+      "step": 21427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.878388159891331e-07,
+      "loss": 0.8696,
+      "step": 21428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8765108157002109e-07,
+      "loss": 0.8213,
+      "step": 21429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8746343921872801e-07,
+      "loss": 0.8286,
+      "step": 21430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.872758889388421e-07,
+      "loss": 0.8457,
+      "step": 21431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8708843073395266e-07,
+      "loss": 0.792,
+      "step": 21432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.869010646076458e-07,
+      "loss": 0.8579,
+      "step": 21433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8671379056350637e-07,
+      "loss": 0.897,
+      "step": 21434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8652660860511652e-07,
+      "loss": 0.8501,
+      "step": 21435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8633951873605727e-07,
+      "loss": 0.8257,
+      "step": 21436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8615252095990966e-07,
+      "loss": 0.8086,
+      "step": 21437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.859656152802486e-07,
+      "loss": 0.8706,
+      "step": 21438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8577880170065176e-07,
+      "loss": 0.7959,
+      "step": 21439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.85592080224693e-07,
+      "loss": 0.8501,
+      "step": 21440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8540545085594498e-07,
+      "loss": 0.8188,
+      "step": 21441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8521891359797595e-07,
+      "loss": 0.8413,
+      "step": 21442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8503246845435751e-07,
+      "loss": 0.8096,
+      "step": 21443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.848461154286546e-07,
+      "loss": 0.7314,
+      "step": 21444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8465985452443435e-07,
+      "loss": 0.7817,
+      "step": 21445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8447368574525782e-07,
+      "loss": 0.813,
+      "step": 21446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8428760909468878e-07,
+      "loss": 0.8633,
+      "step": 21447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8410162457628668e-07,
+      "loss": 0.8257,
+      "step": 21448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.839157321936086e-07,
+      "loss": 0.8652,
+      "step": 21449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.837299319502117e-07,
+      "loss": 0.8516,
+      "step": 21450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8354422384965153e-07,
+      "loss": 0.8003,
+      "step": 21451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.833586078954791e-07,
+      "loss": 0.835,
+      "step": 21452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8317308409124767e-07,
+      "loss": 0.7861,
+      "step": 21453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8298765244050497e-07,
+      "loss": 0.7759,
+      "step": 21454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8280231294679929e-07,
+      "loss": 0.7891,
+      "step": 21455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8261706561367608e-07,
+      "loss": 0.8408,
+      "step": 21456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8243191044467977e-07,
+      "loss": 0.7935,
+      "step": 21457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8224684744335253e-07,
+      "loss": 0.8828,
+      "step": 21458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8206187661323481e-07,
+      "loss": 0.8223,
+      "step": 21459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8187699795786607e-07,
+      "loss": 0.8672,
+      "step": 21460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8169221148078287e-07,
+      "loss": 0.8315,
+      "step": 21461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8150751718551963e-07,
+      "loss": 0.8892,
+      "step": 21462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8132291507561074e-07,
+      "loss": 0.8301,
+      "step": 21463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8113840515458836e-07,
+      "loss": 0.8057,
+      "step": 21464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.809539874259808e-07,
+      "loss": 0.8286,
+      "step": 21465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8076966189331801e-07,
+      "loss": 0.8911,
+      "step": 21466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8058542856012607e-07,
+      "loss": 0.8125,
+      "step": 21467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.804012874299288e-07,
+      "loss": 0.8408,
+      "step": 21468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8021723850625007e-07,
+      "loss": 0.8267,
+      "step": 21469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8003328179261036e-07,
+      "loss": 0.8374,
+      "step": 21470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.798494172925286e-07,
+      "loss": 0.7351,
+      "step": 21471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7966564500952412e-07,
+      "loss": 0.2946,
+      "step": 21472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.794819649471119e-07,
+      "loss": 0.8867,
+      "step": 21473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7929837710880582e-07,
+      "loss": 0.8066,
+      "step": 21474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7911488149811752e-07,
+      "loss": 0.8242,
+      "step": 21475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7893147811855805e-07,
+      "loss": 0.8447,
+      "step": 21476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.787481669736374e-07,
+      "loss": 0.8354,
+      "step": 21477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7856494806686108e-07,
+      "loss": 0.7954,
+      "step": 21478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.783818214017352e-07,
+      "loss": 0.8643,
+      "step": 21479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7819878698176308e-07,
+      "loss": 0.7432,
+      "step": 21480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7801584481044633e-07,
+      "loss": 0.8354,
+      "step": 21481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7783299489128492e-07,
+      "loss": 0.8945,
+      "step": 21482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7765023722777718e-07,
+      "loss": 0.9072,
+      "step": 21483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7746757182341866e-07,
+      "loss": 0.8218,
+      "step": 21484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7728499868170602e-07,
+      "loss": 0.8442,
+      "step": 21485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7710251780613085e-07,
+      "loss": 0.7979,
+      "step": 21486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7692012920018376e-07,
+      "loss": 0.8398,
+      "step": 21487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7673783286735579e-07,
+      "loss": 0.8218,
+      "step": 21488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7655562881113252e-07,
+      "loss": 0.8271,
+      "step": 21489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7637351703500171e-07,
+      "loss": 0.7939,
+      "step": 21490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7619149754244614e-07,
+      "loss": 0.8447,
+      "step": 21491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.760095703369491e-07,
+      "loss": 0.8916,
+      "step": 21492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7582773542199062e-07,
+      "loss": 0.7803,
+      "step": 21493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.75645992801049e-07,
+      "loss": 0.7881,
+      "step": 21494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7546434247760147e-07,
+      "loss": 0.7939,
+      "step": 21495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.752827844551247e-07,
+      "loss": 0.8267,
+      "step": 21496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.751013187370898e-07,
+      "loss": 0.8501,
+      "step": 21497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7491994532697065e-07,
+      "loss": 0.8257,
+      "step": 21498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7473866422823672e-07,
+      "loss": 0.8394,
+      "step": 21499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7455747544435464e-07,
+      "loss": 0.8247,
+      "step": 21500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7437637897879278e-07,
+      "loss": 0.7949,
+      "step": 21501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7419537483501446e-07,
+      "loss": 0.8193,
+      "step": 21502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.740144630164836e-07,
+      "loss": 0.8218,
+      "step": 21503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7383364352666076e-07,
+      "loss": 0.8442,
+      "step": 21504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7365291636900595e-07,
+      "loss": 0.8579,
+      "step": 21505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.734722815469758e-07,
+      "loss": 0.8398,
+      "step": 21506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7329173906402653e-07,
+      "loss": 0.8247,
+      "step": 21507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.731112889236114e-07,
+      "loss": 0.8853,
+      "step": 21508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.729309311291849e-07,
+      "loss": 0.8945,
+      "step": 21509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.727506656841954e-07,
+      "loss": 0.8726,
+      "step": 21510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7257049259209292e-07,
+      "loss": 0.8311,
+      "step": 21511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.723904118563241e-07,
+      "loss": 0.8022,
+      "step": 21512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7221042348033346e-07,
+      "loss": 0.8525,
+      "step": 21513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7203052746756543e-07,
+      "loss": 0.8579,
+      "step": 21514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7185072382146118e-07,
+      "loss": 0.8809,
+      "step": 21515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7167101254546015e-07,
+      "loss": 0.8105,
+      "step": 21516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7149139364300238e-07,
+      "loss": 0.8921,
+      "step": 21517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7131186711752234e-07,
+      "loss": 0.8706,
+      "step": 21518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.711324329724562e-07,
+      "loss": 0.8267,
+      "step": 21519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7095309121123506e-07,
+      "loss": 0.8672,
+      "step": 21520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.707738418372906e-07,
+      "loss": 0.8008,
+      "step": 21521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.705946848540535e-07,
+      "loss": 0.8022,
+      "step": 21522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7041562026494985e-07,
+      "loss": 0.8379,
+      "step": 21523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7023664807340524e-07,
+      "loss": 0.7705,
+      "step": 21524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7005776828284527e-07,
+      "loss": 0.8711,
+      "step": 21525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.698789808966911e-07,
+      "loss": 0.8511,
+      "step": 21526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.697002859183633e-07,
+      "loss": 0.8403,
+      "step": 21527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6952168335128028e-07,
+      "loss": 0.8472,
+      "step": 21528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6934317319885928e-07,
+      "loss": 0.8101,
+      "step": 21529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6916475546451594e-07,
+      "loss": 0.7632,
+      "step": 21530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.689864301516625e-07,
+      "loss": 0.8359,
+      "step": 21531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6880819726371234e-07,
+      "loss": 0.8311,
+      "step": 21532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6863005680407441e-07,
+      "loss": 0.8018,
+      "step": 21533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6845200877615596e-07,
+      "loss": 0.8555,
+      "step": 21534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6827405318336487e-07,
+      "loss": 0.3806,
+      "step": 21535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.680961900291045e-07,
+      "loss": 0.7944,
+      "step": 21536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6791841931677822e-07,
+      "loss": 0.8521,
+      "step": 21537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6774074104978723e-07,
+      "loss": 0.8989,
+      "step": 21538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6756315523153099e-07,
+      "loss": 0.8955,
+      "step": 21539
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6738566186540628e-07,
+      "loss": 0.7915,
+      "step": 21540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.672082609548087e-07,
+      "loss": 0.875,
+      "step": 21541
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6703095250313216e-07,
+      "loss": 0.8066,
+      "step": 21542
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6685373651377013e-07,
+      "loss": 0.8135,
+      "step": 21543
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6667661299011206e-07,
+      "loss": 0.7871,
+      "step": 21544
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6649958193554694e-07,
+      "loss": 0.8208,
+      "step": 21545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6632264335346148e-07,
+      "loss": 0.811,
+      "step": 21546
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6614579724724022e-07,
+      "loss": 0.8442,
+      "step": 21547
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6596904362026766e-07,
+      "loss": 0.7681,
+      "step": 21548
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6579238247592443e-07,
+      "loss": 0.8135,
+      "step": 21549
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6561581381759006e-07,
+      "loss": 0.8125,
+      "step": 21550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.654393376486446e-07,
+      "loss": 0.855,
+      "step": 21551
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6526295397246262e-07,
+      "loss": 0.8262,
+      "step": 21552
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6508666279241803e-07,
+      "loss": 0.7817,
+      "step": 21553
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6491046411188538e-07,
+      "loss": 0.8237,
+      "step": 21554
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.647343579342342e-07,
+      "loss": 0.8931,
+      "step": 21555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6455834426283458e-07,
+      "loss": 0.8511,
+      "step": 21556
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6438242310105267e-07,
+      "loss": 0.8413,
+      "step": 21557
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6420659445225584e-07,
+      "loss": 0.7842,
+      "step": 21558
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6403085831980746e-07,
+      "loss": 0.8687,
+      "step": 21559
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6385521470706823e-07,
+      "loss": 0.7881,
+      "step": 21560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6367966361739984e-07,
+      "loss": 0.8413,
+      "step": 21561
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6350420505416075e-07,
+      "loss": 0.8599,
+      "step": 21562
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6332883902070772e-07,
+      "loss": 0.8232,
+      "step": 21563
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.631535655203953e-07,
+      "loss": 0.8232,
+      "step": 21564
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.62978384556578e-07,
+      "loss": 0.8853,
+      "step": 21565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6280329613260536e-07,
+      "loss": 0.8672,
+      "step": 21566
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6262830025182863e-07,
+      "loss": 0.856,
+      "step": 21567
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.624533969175951e-07,
+      "loss": 0.8423,
+      "step": 21568
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.62278586133251e-07,
+      "loss": 0.877,
+      "step": 21569
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6210386790214083e-07,
+      "loss": 0.8384,
+      "step": 21570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6192924222760807e-07,
+      "loss": 0.8262,
+      "step": 21571
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.617547091129923e-07,
+      "loss": 0.771,
+      "step": 21572
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6158026856163244e-07,
+      "loss": 0.7881,
+      "step": 21573
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6140592057686643e-07,
+      "loss": 0.8184,
+      "step": 21574
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6123166516203105e-07,
+      "loss": 0.8389,
+      "step": 21575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6105750232045748e-07,
+      "loss": 0.9004,
+      "step": 21576
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.608834320554803e-07,
+      "loss": 0.8633,
+      "step": 21577
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6070945437042796e-07,
+      "loss": 0.8618,
+      "step": 21578
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6053556926862946e-07,
+      "loss": 0.8193,
+      "step": 21579
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6036177675341214e-07,
+      "loss": 0.8169,
+      "step": 21580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6018807682809945e-07,
+      "loss": 0.8047,
+      "step": 21581
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.600144694960154e-07,
+      "loss": 0.2867,
+      "step": 21582
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5984095476048233e-07,
+      "loss": 0.8809,
+      "step": 21583
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5966753262481872e-07,
+      "loss": 0.8413,
+      "step": 21584
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5949420309234298e-07,
+      "loss": 0.8223,
+      "step": 21585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5932096616636972e-07,
+      "loss": 0.7998,
+      "step": 21586
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.591478218502146e-07,
+      "loss": 0.8169,
+      "step": 21587
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5897477014719055e-07,
+      "loss": 0.8369,
+      "step": 21588
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.588018110606071e-07,
+      "loss": 0.8081,
+      "step": 21589
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5862894459377387e-07,
+      "loss": 0.8735,
+      "step": 21590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5845617074999874e-07,
+      "loss": 0.8613,
+      "step": 21591
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5828348953258522e-07,
+      "loss": 0.8296,
+      "step": 21592
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5811090094483893e-07,
+      "loss": 0.8799,
+      "step": 21593
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5793840499006063e-07,
+      "loss": 0.8774,
+      "step": 21594
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5776600167155042e-07,
+      "loss": 0.812,
+      "step": 21595
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.575936909926079e-07,
+      "loss": 0.8716,
+      "step": 21596
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5742147295652822e-07,
+      "loss": 0.8071,
+      "step": 21597
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5724934756660649e-07,
+      "loss": 0.8584,
+      "step": 21598
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5707731482613675e-07,
+      "loss": 0.7729,
+      "step": 21599
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5690537473840862e-07,
+      "loss": 0.7998,
+      "step": 21600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5673352730671331e-07,
+      "loss": 0.8511,
+      "step": 21601
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5656177253433656e-07,
+      "loss": 0.8936,
+      "step": 21602
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.563901104245663e-07,
+      "loss": 0.7773,
+      "step": 21603
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5621854098068601e-07,
+      "loss": 0.916,
+      "step": 21604
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5604706420597693e-07,
+      "loss": 0.8271,
+      "step": 21605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5587568010372145e-07,
+      "loss": 0.8052,
+      "step": 21606
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5570438867719695e-07,
+      "loss": 0.8306,
+      "step": 21607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.555331899296808e-07,
+      "loss": 0.8921,
+      "step": 21608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5536208386444985e-07,
+      "loss": 0.8047,
+      "step": 21609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5519107048477587e-07,
+      "loss": 0.3733,
+      "step": 21610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5502014979393132e-07,
+      "loss": 0.7871,
+      "step": 21611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.548493217951863e-07,
+      "loss": 0.8823,
+      "step": 21612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5467858649180767e-07,
+      "loss": 0.8374,
+      "step": 21613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.545079438870639e-07,
+      "loss": 0.8091,
+      "step": 21614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5433739398421855e-07,
+      "loss": 0.8472,
+      "step": 21615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.541669367865345e-07,
+      "loss": 0.8081,
+      "step": 21616
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.539965722972736e-07,
+      "loss": 0.8296,
+      "step": 21617
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5382630051969383e-07,
+      "loss": 0.856,
+      "step": 21618
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5365612145705366e-07,
+      "loss": 0.8682,
+      "step": 21619
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.534860351126094e-07,
+      "loss": 0.8052,
+      "step": 21620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5331604148961343e-07,
+      "loss": 0.894,
+      "step": 21621
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.531461405913204e-07,
+      "loss": 0.8628,
+      "step": 21622
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5297633242097876e-07,
+      "loss": 0.8408,
+      "step": 21623
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5280661698183709e-07,
+      "loss": 0.6733,
+      "step": 21624
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5263699427714386e-07,
+      "loss": 0.8271,
+      "step": 21625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5246746431014315e-07,
+      "loss": 0.8091,
+      "step": 21626
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5229802708407847e-07,
+      "loss": 0.835,
+      "step": 21627
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5212868260219226e-07,
+      "loss": 0.8389,
+      "step": 21628
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5195943086772246e-07,
+      "loss": 0.8418,
+      "step": 21629
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.517902718839098e-07,
+      "loss": 0.8105,
+      "step": 21630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.516212056539884e-07,
+      "loss": 0.7935,
+      "step": 21631
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5145223218119286e-07,
+      "loss": 0.7974,
+      "step": 21632
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5128335146875727e-07,
+      "loss": 0.8438,
+      "step": 21633
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5111456351991127e-07,
+      "loss": 0.835,
+      "step": 21634
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.509458683378845e-07,
+      "loss": 0.9092,
+      "step": 21635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.507772659259049e-07,
+      "loss": 0.8066,
+      "step": 21636
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5060875628719773e-07,
+      "loss": 0.7852,
+      "step": 21637
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5044033942498647e-07,
+      "loss": 0.8384,
+      "step": 21638
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5027201534249302e-07,
+      "loss": 0.8242,
+      "step": 21639
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5010378404293813e-07,
+      "loss": 0.8286,
+      "step": 21640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4993564552954142e-07,
+      "loss": 0.8237,
+      "step": 21641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4976759980551758e-07,
+      "loss": 0.256,
+      "step": 21642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.495996468740829e-07,
+      "loss": 0.8271,
+      "step": 21643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4943178673845037e-07,
+      "loss": 0.8091,
+      "step": 21644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4926401940183134e-07,
+      "loss": 0.8423,
+      "step": 21645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.490963448674354e-07,
+      "loss": 0.8389,
+      "step": 21646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4892876313847004e-07,
+      "loss": 0.8018,
+      "step": 21647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4876127421814156e-07,
+      "loss": 0.8569,
+      "step": 21648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4859387810965519e-07,
+      "loss": 0.8589,
+      "step": 21649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.484265748162128e-07,
+      "loss": 0.7485,
+      "step": 21650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4825936434101573e-07,
+      "loss": 0.8486,
+      "step": 21651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.480922466872614e-07,
+      "loss": 0.8462,
+      "step": 21652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4792522185814785e-07,
+      "loss": 0.8306,
+      "step": 21653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4775828985687135e-07,
+      "loss": 0.8325,
+      "step": 21654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.475914506866244e-07,
+      "loss": 0.8472,
+      "step": 21655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4742470435059996e-07,
+      "loss": 0.8633,
+      "step": 21656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4725805085198775e-07,
+      "loss": 0.7881,
+      "step": 21657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4709149019397574e-07,
+      "loss": 0.7959,
+      "step": 21658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.469250223797508e-07,
+      "loss": 0.7886,
+      "step": 21659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4675864741249767e-07,
+      "loss": 0.8271,
+      "step": 21660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4659236529539876e-07,
+      "loss": 0.8018,
+      "step": 21661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4642617603163712e-07,
+      "loss": 0.8359,
+      "step": 21662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4626007962439072e-07,
+      "loss": 0.8291,
+      "step": 21663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4609407607683645e-07,
+      "loss": 0.8867,
+      "step": 21664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4592816539215238e-07,
+      "loss": 0.874,
+      "step": 21665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4576234757351094e-07,
+      "loss": 0.8438,
+      "step": 21666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.455966226240857e-07,
+      "loss": 0.8628,
+      "step": 21667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4543099054704635e-07,
+      "loss": 0.8296,
+      "step": 21668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.45265451345562e-07,
+      "loss": 0.7871,
+      "step": 21669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4510000502280018e-07,
+      "loss": 0.814,
+      "step": 21670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.44934651581925e-07,
+      "loss": 0.7915,
+      "step": 21671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4476939102610111e-07,
+      "loss": 0.9033,
+      "step": 21672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4460422335848878e-07,
+      "loss": 0.7944,
+      "step": 21673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4443914858224938e-07,
+      "loss": 0.8374,
+      "step": 21674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.442741667005404e-07,
+      "loss": 0.7622,
+      "step": 21675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4410927771651873e-07,
+      "loss": 0.8228,
+      "step": 21676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4394448163333795e-07,
+      "loss": 0.8481,
+      "step": 21677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.437797784541517e-07,
+      "loss": 0.7866,
+      "step": 21678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4361516818211018e-07,
+      "loss": 0.8301,
+      "step": 21679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.434506508203637e-07,
+      "loss": 0.8125,
+      "step": 21680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4328622637205915e-07,
+      "loss": 0.811,
+      "step": 21681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4312189484034233e-07,
+      "loss": 0.8267,
+      "step": 21682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4295765622835743e-07,
+      "loss": 0.9067,
+      "step": 21683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4279351053924584e-07,
+      "loss": 0.8311,
+      "step": 21684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4262945777614835e-07,
+      "loss": 0.8237,
+      "step": 21685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.424654979422041e-07,
+      "loss": 0.7563,
+      "step": 21686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4230163104054896e-07,
+      "loss": 0.8379,
+      "step": 21687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.421378570743187e-07,
+      "loss": 0.8799,
+      "step": 21688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4197417604664642e-07,
+      "loss": 0.8599,
+      "step": 21689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4181058796066293e-07,
+      "loss": 0.853,
+      "step": 21690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4164709281949907e-07,
+      "loss": 0.7773,
+      "step": 21691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4148369062628175e-07,
+      "loss": 0.8047,
+      "step": 21692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4132038138413683e-07,
+      "loss": 0.8345,
+      "step": 21693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4115716509619072e-07,
+      "loss": 0.9028,
+      "step": 21694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.409940417655642e-07,
+      "loss": 0.4252,
+      "step": 21695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4083101139537813e-07,
+      "loss": 0.7983,
+      "step": 21696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.406680739887517e-07,
+      "loss": 0.8809,
+      "step": 21697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4050522954880242e-07,
+      "loss": 0.7572,
+      "step": 21698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4034247807864665e-07,
+      "loss": 0.7632,
+      "step": 21699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4017981958139583e-07,
+      "loss": 0.834,
+      "step": 21700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4001725406016409e-07,
+      "loss": 0.8228,
+      "step": 21701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3985478151806065e-07,
+      "loss": 0.8081,
+      "step": 21702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3969240195819356e-07,
+      "loss": 0.7852,
+      "step": 21703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.395301153836698e-07,
+      "loss": 0.8184,
+      "step": 21704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3936792179759406e-07,
+      "loss": 0.8188,
+      "step": 21705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3920582120306892e-07,
+      "loss": 0.8315,
+      "step": 21706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3904381360319685e-07,
+      "loss": 0.7642,
+      "step": 21707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3888189900107652e-07,
+      "loss": 0.8379,
+      "step": 21708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3872007739980543e-07,
+      "loss": 0.4325,
+      "step": 21709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3855834880247942e-07,
+      "loss": 0.8555,
+      "step": 21710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3839671321219272e-07,
+      "loss": 0.8569,
+      "step": 21711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823517063203838e-07,
+      "loss": 0.8223,
+      "step": 21712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380737210651062e-07,
+      "loss": 0.8516,
+      "step": 21713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791236451448532e-07,
+      "loss": 0.8091,
+      "step": 21714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3775110098326273e-07,
+      "loss": 0.7896,
+      "step": 21715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3758993047452317e-07,
+      "loss": 0.8604,
+      "step": 21716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.374288529913509e-07,
+      "loss": 0.8213,
+      "step": 21717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3726786853682617e-07,
+      "loss": 0.8384,
+      "step": 21718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3710697711403041e-07,
+      "loss": 0.8872,
+      "step": 21719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3694617872604122e-07,
+      "loss": 0.8579,
+      "step": 21720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3678547337593494e-07,
+      "loss": 0.8535,
+      "step": 21721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3662486106678584e-07,
+      "loss": 0.3499,
+      "step": 21722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3646434180166702e-07,
+      "loss": 0.7681,
+      "step": 21723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3630391558364875e-07,
+      "loss": 0.7759,
+      "step": 21724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3614358241580193e-07,
+      "loss": 0.7905,
+      "step": 21725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3598334230119192e-07,
+      "loss": 0.8643,
+      "step": 21726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3582319524288567e-07,
+      "loss": 0.8257,
+      "step": 21727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3566314124394686e-07,
+      "loss": 0.8071,
+      "step": 21728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3550318030743692e-07,
+      "loss": 0.8364,
+      "step": 21729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3534331243641675e-07,
+      "loss": 0.8223,
+      "step": 21730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.35183537633945e-07,
+      "loss": 0.8438,
+      "step": 21731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3502385590307753e-07,
+      "loss": 0.8091,
+      "step": 21732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3486426724687085e-07,
+      "loss": 0.8516,
+      "step": 21733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3470477166837692e-07,
+      "loss": 0.8687,
+      "step": 21734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.345453691706472e-07,
+      "loss": 0.8196,
+      "step": 21735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3438605975673203e-07,
+      "loss": 0.7856,
+      "step": 21736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.342268434296784e-07,
+      "loss": 0.7979,
+      "step": 21737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3406772019253278e-07,
+      "loss": 0.7866,
+      "step": 21738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.339086900483394e-07,
+      "loss": 0.8354,
+      "step": 21739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.337497530001408e-07,
+      "loss": 0.7837,
+      "step": 21740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.335909090509785e-07,
+      "loss": 0.8135,
+      "step": 21741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3343215820389e-07,
+      "loss": 0.8384,
+      "step": 21742
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3327350046191234e-07,
+      "loss": 0.8921,
+      "step": 21743
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3311493582808256e-07,
+      "loss": 0.8101,
+      "step": 21744
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3295646430543262e-07,
+      "loss": 0.8032,
+      "step": 21745
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3279808589699573e-07,
+      "loss": 0.8481,
+      "step": 21746
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3263980060580105e-07,
+      "loss": 0.7781,
+      "step": 21747
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.324816084348768e-07,
+      "loss": 0.7988,
+      "step": 21748
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3232350938724992e-07,
+      "loss": 0.7837,
+      "step": 21749
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3216550346594413e-07,
+      "loss": 0.8501,
+      "step": 21750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3200759067398371e-07,
+      "loss": 0.7896,
+      "step": 21751
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.31849771014389e-07,
+      "loss": 0.8706,
+      "step": 21752
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.316920444901787e-07,
+      "loss": 0.8203,
+      "step": 21753
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3153441110437204e-07,
+      "loss": 0.8091,
+      "step": 21754
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3137687085998385e-07,
+      "loss": 0.8467,
+      "step": 21755
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3121942376002726e-07,
+      "loss": 0.8506,
+      "step": 21756
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.31062069807516e-07,
+      "loss": 0.7583,
+      "step": 21757
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3090480900545933e-07,
+      "loss": 0.8242,
+      "step": 21758
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3074764135686646e-07,
+      "loss": 0.8545,
+      "step": 21759
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3059056686474504e-07,
+      "loss": 0.8794,
+      "step": 21760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3043358553209873e-07,
+      "loss": 0.8872,
+      "step": 21761
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3027669736193181e-07,
+      "loss": 0.8408,
+      "step": 21762
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.301199023572447e-07,
+      "loss": 0.8015,
+      "step": 21763
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2996320052103773e-07,
+      "loss": 0.3345,
+      "step": 21764
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2980659185630962e-07,
+      "loss": 0.8662,
+      "step": 21765
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.296500763660552e-07,
+      "loss": 0.8501,
+      "step": 21766
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.294936540532704e-07,
+      "loss": 0.7969,
+      "step": 21767
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2933732492094676e-07,
+      "loss": 0.8228,
+      "step": 21768
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.291810889720746e-07,
+      "loss": 0.8618,
+      "step": 21769
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2902494620964433e-07,
+      "loss": 0.7974,
+      "step": 21770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2886889663664193e-07,
+      "loss": 0.8005,
+      "step": 21771
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2871294025605384e-07,
+      "loss": 0.7954,
+      "step": 21772
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2855707707086385e-07,
+      "loss": 0.3955,
+      "step": 21773
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2840130708405284e-07,
+      "loss": 0.8164,
+      "step": 21774
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2824563029860183e-07,
+      "loss": 0.8301,
+      "step": 21775
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2809004671748838e-07,
+      "loss": 0.8276,
+      "step": 21776
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2793455634368957e-07,
+      "loss": 0.7729,
+      "step": 21777
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.277791591801808e-07,
+      "loss": 0.7563,
+      "step": 21778
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.276238552299336e-07,
+      "loss": 0.814,
+      "step": 21779
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2746864449592055e-07,
+      "loss": 0.856,
+      "step": 21780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2731352698111043e-07,
+      "loss": 0.8594,
+      "step": 21781
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2715850268847084e-07,
+      "loss": 0.8267,
+      "step": 21782
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2700357162096776e-07,
+      "loss": 0.7964,
+      "step": 21783
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2684873378156493e-07,
+      "loss": 0.8535,
+      "step": 21784
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2669398917322496e-07,
+      "loss": 0.8398,
+      "step": 21785
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2653933779890882e-07,
+      "loss": 0.8384,
+      "step": 21786
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.263847796615747e-07,
+      "loss": 0.8467,
+      "step": 21787
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2623031476417914e-07,
+      "loss": 0.8188,
+      "step": 21788
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.260759431096781e-07,
+      "loss": 0.8774,
+      "step": 21789
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.259216647010242e-07,
+      "loss": 0.811,
+      "step": 21790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2576747954117007e-07,
+      "loss": 0.8267,
+      "step": 21791
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.256133876330645e-07,
+      "loss": 0.9136,
+      "step": 21792
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2545938897965625e-07,
+      "loss": 0.7891,
+      "step": 21793
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2530548358389128e-07,
+      "loss": 0.8682,
+      "step": 21794
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2515167144871333e-07,
+      "loss": 0.8145,
+      "step": 21795
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.249979525770656e-07,
+      "loss": 0.8447,
+      "step": 21796
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2484432697188965e-07,
+      "loss": 0.8462,
+      "step": 21797
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2469079463612366e-07,
+      "loss": 0.7939,
+      "step": 21798
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2453735557270586e-07,
+      "loss": 0.8384,
+      "step": 21799
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2438400978457111e-07,
+      "loss": 0.8428,
+      "step": 21800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2423075727465262e-07,
+      "loss": 0.8335,
+      "step": 21801
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2407759804588415e-07,
+      "loss": 0.8267,
+      "step": 21802
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2392453210119337e-07,
+      "loss": 0.8643,
+      "step": 21803
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.237715594435107e-07,
+      "loss": 0.8535,
+      "step": 21804
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2361868007576217e-07,
+      "loss": 0.7896,
+      "step": 21805
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2346589400087262e-07,
+      "loss": 0.7822,
+      "step": 21806
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2331320122176472e-07,
+      "loss": 0.8511,
+      "step": 21807
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2316060174136e-07,
+      "loss": 0.8296,
+      "step": 21808
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2300809556257786e-07,
+      "loss": 0.8647,
+      "step": 21809
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.228556826883359e-07,
+      "loss": 0.8369,
+      "step": 21810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2270336312155018e-07,
+      "loss": 0.791,
+      "step": 21811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2255113686513554e-07,
+      "loss": 0.8408,
+      "step": 21812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2239900392200298e-07,
+      "loss": 0.8071,
+      "step": 21813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2224696429506355e-07,
+      "loss": 0.8223,
+      "step": 21814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.22095017987226e-07,
+      "loss": 0.8921,
+      "step": 21815
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2194316500139746e-07,
+      "loss": 0.8428,
+      "step": 21816
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2179140534048283e-07,
+      "loss": 0.8354,
+      "step": 21817
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2163973900738645e-07,
+      "loss": 0.8804,
+      "step": 21818
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.214881660050088e-07,
+      "loss": 0.8545,
+      "step": 21819
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.213366863362503e-07,
+      "loss": 0.8604,
+      "step": 21820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2118530000400807e-07,
+      "loss": 0.8887,
+      "step": 21821
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2103400701117985e-07,
+      "loss": 0.8564,
+      "step": 21822
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2088280736065938e-07,
+      "loss": 0.8486,
+      "step": 21823
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2073170105533883e-07,
+      "loss": 0.8149,
+      "step": 21824
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2058068809810975e-07,
+      "loss": 0.8208,
+      "step": 21825
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.204297684918615e-07,
+      "loss": 0.8159,
+      "step": 21826
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2027894223948066e-07,
+      "loss": 0.894,
+      "step": 21827
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2012820934385328e-07,
+      "loss": 0.8379,
+      "step": 21828
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1997756980786256e-07,
+      "loss": 0.8359,
+      "step": 21829
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1982702363439124e-07,
+      "loss": 0.8262,
+      "step": 21830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1967657082631923e-07,
+      "loss": 0.8569,
+      "step": 21831
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.195262113865253e-07,
+      "loss": 0.8301,
+      "step": 21832
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1937594531788443e-07,
+      "loss": 0.8423,
+      "step": 21833
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1922577262327374e-07,
+      "loss": 0.7603,
+      "step": 21834
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1907569330556424e-07,
+      "loss": 0.811,
+      "step": 21835
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1892570736762921e-07,
+      "loss": 0.8125,
+      "step": 21836
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.187758148123358e-07,
+      "loss": 0.8193,
+      "step": 21837
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1862601564255393e-07,
+      "loss": 0.854,
+      "step": 21838
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1847630986114856e-07,
+      "loss": 0.8516,
+      "step": 21839
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1832669747098291e-07,
+      "loss": 0.8174,
+      "step": 21840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1817717847492028e-07,
+      "loss": 0.8442,
+      "step": 21841
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1802775287582114e-07,
+      "loss": 0.8223,
+      "step": 21842
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1787842067654376e-07,
+      "loss": 0.8574,
+      "step": 21843
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1772918187994586e-07,
+      "loss": 0.8511,
+      "step": 21844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1758003648888239e-07,
+      "loss": 0.8677,
+      "step": 21845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1743098450620605e-07,
+      "loss": 0.8394,
+      "step": 21846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1728202593476956e-07,
+      "loss": 0.835,
+      "step": 21847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1713316077742176e-07,
+      "loss": 0.8555,
+      "step": 21848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1698438903701092e-07,
+      "loss": 0.8042,
+      "step": 21849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1683571071638367e-07,
+      "loss": 0.9121,
+      "step": 21850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1668712581838381e-07,
+      "loss": 0.8193,
+      "step": 21851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1653863434585522e-07,
+      "loss": 0.8555,
+      "step": 21852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1639023630163725e-07,
+      "loss": 0.9092,
+      "step": 21853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1624193168856934e-07,
+      "loss": 0.8335,
+      "step": 21854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1609372050948975e-07,
+      "loss": 0.8467,
+      "step": 21855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1594560276723288e-07,
+      "loss": 0.79,
+      "step": 21856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1579757846463368e-07,
+      "loss": 0.9072,
+      "step": 21857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1564964760452213e-07,
+      "loss": 0.7925,
+      "step": 21858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1550181018973094e-07,
+      "loss": 0.7944,
+      "step": 21859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1535406622308676e-07,
+      "loss": 0.7544,
+      "step": 21860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1520641570741619e-07,
+      "loss": 0.8369,
+      "step": 21861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1505885864554423e-07,
+      "loss": 0.8125,
+      "step": 21862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1491139504029414e-07,
+      "loss": 0.8584,
+      "step": 21863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1476402489448702e-07,
+      "loss": 0.7939,
+      "step": 21864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1461674821094282e-07,
+      "loss": 0.814,
+      "step": 21865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1446956499247819e-07,
+      "loss": 0.7368,
+      "step": 21866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1432247524190921e-07,
+      "loss": 0.8052,
+      "step": 21867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1417547896205084e-07,
+      "loss": 0.8291,
+      "step": 21868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1402857615571416e-07,
+      "loss": 0.834,
+      "step": 21869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1388176682570973e-07,
+      "loss": 0.7998,
+      "step": 21870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.137350509748475e-07,
+      "loss": 0.8789,
+      "step": 21871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.13588428605933e-07,
+      "loss": 0.8174,
+      "step": 21872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1344189972177233e-07,
+      "loss": 0.8584,
+      "step": 21873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1329546432516824e-07,
+      "loss": 0.7578,
+      "step": 21874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1314912241892184e-07,
+      "loss": 0.8179,
+      "step": 21875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.130028740058342e-07,
+      "loss": 0.8398,
+      "step": 21876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.12856719088702e-07,
+      "loss": 0.8853,
+      "step": 21877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1271065767032241e-07,
+      "loss": 0.8096,
+      "step": 21878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1256468975348933e-07,
+      "loss": 0.8457,
+      "step": 21879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1241881534099496e-07,
+      "loss": 0.8745,
+      "step": 21880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1227303443563098e-07,
+      "loss": 0.2877,
+      "step": 21881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1212734704018513e-07,
+      "loss": 0.7993,
+      "step": 21882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1198175315744575e-07,
+      "loss": 0.8091,
+      "step": 21883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1183625279019838e-07,
+      "loss": 0.8218,
+      "step": 21884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1169084594122637e-07,
+      "loss": 0.7959,
+      "step": 21885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1154553261331191e-07,
+      "loss": 0.7896,
+      "step": 21886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1140031280923336e-07,
+      "loss": 0.8604,
+      "step": 21887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1125518653177125e-07,
+      "loss": 0.7573,
+      "step": 21888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.111101537837006e-07,
+      "loss": 0.8364,
+      "step": 21889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1096521456779697e-07,
+      "loss": 0.8818,
+      "step": 21890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1082036888683311e-07,
+      "loss": 0.291,
+      "step": 21891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1067561674357963e-07,
+      "loss": 0.7939,
+      "step": 21892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.105309581408065e-07,
+      "loss": 0.8262,
+      "step": 21893
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1038639308128097e-07,
+      "loss": 0.8755,
+      "step": 21894
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1024192156776803e-07,
+      "loss": 0.8198,
+      "step": 21895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1009754360303327e-07,
+      "loss": 0.8271,
+      "step": 21896
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.099532591898378e-07,
+      "loss": 0.7944,
+      "step": 21897
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0980906833094273e-07,
+      "loss": 0.8081,
+      "step": 21898
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0966497102910533e-07,
+      "loss": 0.8184,
+      "step": 21899
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0952096728708394e-07,
+      "loss": 0.8213,
+      "step": 21900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0937705710763192e-07,
+      "loss": 0.8579,
+      "step": 21901
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0923324049350426e-07,
+      "loss": 0.8044,
+      "step": 21902
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0908951744745155e-07,
+      "loss": 0.8726,
+      "step": 21903
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0894588797222328e-07,
+      "loss": 0.8779,
+      "step": 21904
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0880235207056777e-07,
+      "loss": 0.8032,
+      "step": 21905
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0865890974523063e-07,
+      "loss": 0.8403,
+      "step": 21906
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0851556099895632e-07,
+      "loss": 0.79,
+      "step": 21907
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.083723058344871e-07,
+      "loss": 0.7593,
+      "step": 21908
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.082291442545641e-07,
+      "loss": 0.8472,
+      "step": 21909
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.080860762619268e-07,
+      "loss": 0.7964,
+      "step": 21910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0794310185931078e-07,
+      "loss": 0.8159,
+      "step": 21911
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.078002210494522e-07,
+      "loss": 0.8247,
+      "step": 21912
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0765743383508498e-07,
+      "loss": 0.7559,
+      "step": 21913
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.075147402189397e-07,
+      "loss": 0.8438,
+      "step": 21914
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0737214020374754e-07,
+      "loss": 0.814,
+      "step": 21915
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0722963379223572e-07,
+      "loss": 0.8452,
+      "step": 21916
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0708722098713154e-07,
+      "loss": 0.7939,
+      "step": 21917
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0694490179115946e-07,
+      "loss": 0.8345,
+      "step": 21918
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0680267620704066e-07,
+      "loss": 0.8374,
+      "step": 21919
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0666054423749795e-07,
+      "loss": 0.8643,
+      "step": 21920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0651850588525026e-07,
+      "loss": 0.3384,
+      "step": 21921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0637656115301431e-07,
+      "loss": 0.8169,
+      "step": 21922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0623471004350683e-07,
+      "loss": 0.7593,
+      "step": 21923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0609295255944063e-07,
+      "loss": 0.8657,
+      "step": 21924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.05951288703528e-07,
+      "loss": 0.8735,
+      "step": 21925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0580971847847954e-07,
+      "loss": 0.8628,
+      "step": 21926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0566824188700253e-07,
+      "loss": 0.8701,
+      "step": 21927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0552685893180537e-07,
+      "loss": 0.813,
+      "step": 21928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0538556961559198e-07,
+      "loss": 0.8345,
+      "step": 21929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.052443739410658e-07,
+      "loss": 0.7832,
+      "step": 21930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0510327191092795e-07,
+      "loss": 0.8101,
+      "step": 21931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0496226352787742e-07,
+      "loss": 0.8198,
+      "step": 21932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.048213487946126e-07,
+      "loss": 0.8149,
+      "step": 21933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0468052771382909e-07,
+      "loss": 0.855,
+      "step": 21934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0453980028822142e-07,
+      "loss": 0.8477,
+      "step": 21935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0439916652048188e-07,
+      "loss": 0.8691,
+      "step": 21936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0425862641330053e-07,
+      "loss": 0.8208,
+      "step": 21937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.041181799693658e-07,
+      "loss": 0.8628,
+      "step": 21938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0397782719136607e-07,
+      "loss": 0.8003,
+      "step": 21939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0383756808198531e-07,
+      "loss": 0.8311,
+      "step": 21940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0369740264390694e-07,
+      "loss": 0.3291,
+      "step": 21941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.035573308798138e-07,
+      "loss": 0.8643,
+      "step": 21942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.034173527923843e-07,
+      "loss": 0.8374,
+      "step": 21943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0327746838429686e-07,
+      "loss": 0.7866,
+      "step": 21944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0313767765822769e-07,
+      "loss": 0.8901,
+      "step": 21945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0299798061685074e-07,
+      "loss": 0.8643,
+      "step": 21946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0285837726283999e-07,
+      "loss": 0.7974,
+      "step": 21947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0271886759886496e-07,
+      "loss": 0.3771,
+      "step": 21948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.025794516275952e-07,
+      "loss": 0.429,
+      "step": 21949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0244012935169801e-07,
+      "loss": 0.7993,
+      "step": 21950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0230090077383848e-07,
+      "loss": 0.8306,
+      "step": 21951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0216176589668114e-07,
+      "loss": 0.8516,
+      "step": 21952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0202272472288665e-07,
+      "loss": 0.8643,
+      "step": 21953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0188377725511567e-07,
+      "loss": 0.8149,
+      "step": 21954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.017449234960266e-07,
+      "loss": 0.8091,
+      "step": 21955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0160616344827623e-07,
+      "loss": 0.8164,
+      "step": 21956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0146749711451853e-07,
+      "loss": 0.8921,
+      "step": 21957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0132892449740694e-07,
+      "loss": 0.8081,
+      "step": 21958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0119044559959212e-07,
+      "loss": 0.8115,
+      "step": 21959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0105206042372417e-07,
+      "loss": 0.7837,
+      "step": 21960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0091376897244931e-07,
+      "loss": 0.8618,
+      "step": 21961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0077557124841375e-07,
+      "loss": 0.8394,
+      "step": 21962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0063746725426316e-07,
+      "loss": 0.7886,
+      "step": 21963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0049945699263764e-07,
+      "loss": 0.9106,
+      "step": 21964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.003615404661773e-07,
+      "loss": 0.8486,
+      "step": 21965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0022371767752281e-07,
+      "loss": 0.7817,
+      "step": 21966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0008598862930874e-07,
+      "loss": 0.832,
+      "step": 21967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.994835332417186e-08,
+      "loss": 0.7891,
+      "step": 21968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.981081176474339e-08,
+      "loss": 0.9185,
+      "step": 21969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.967336395365678e-08,
+      "loss": 0.8623,
+      "step": 21970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.953600989354051e-08,
+      "loss": 0.7681,
+      "step": 21971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.939874958702245e-08,
+      "loss": 0.7832,
+      "step": 21972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.926158303672828e-08,
+      "loss": 0.8364,
+      "step": 21973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.912451024528313e-08,
+      "loss": 0.8032,
+      "step": 21974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.898753121530879e-08,
+      "loss": 0.7969,
+      "step": 21975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.885064594942595e-08,
+      "loss": 0.7925,
+      "step": 21976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.871385445025416e-08,
+      "loss": 0.813,
+      "step": 21977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.857715672040857e-08,
+      "loss": 0.8306,
+      "step": 21978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.844055276250653e-08,
+      "loss": 0.4648,
+      "step": 21979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.830404257916038e-08,
+      "loss": 0.8652,
+      "step": 21980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.816762617298192e-08,
+      "loss": 0.7876,
+      "step": 21981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.803130354658131e-08,
+      "loss": 0.8608,
+      "step": 21982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.789507470256643e-08,
+      "loss": 0.8374,
+      "step": 21983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.775893964354355e-08,
+      "loss": 0.8193,
+      "step": 21984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.762289837211669e-08,
+      "loss": 0.8706,
+      "step": 21985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.748695089088877e-08,
+      "loss": 0.873,
+      "step": 21986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.73510972024616e-08,
+      "loss": 0.9019,
+      "step": 21987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.72153373094331e-08,
+      "loss": 0.9023,
+      "step": 21988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.707967121440176e-08,
+      "loss": 0.8179,
+      "step": 21989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694409891996215e-08,
+      "loss": 0.7905,
+      "step": 21990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.68086204287072e-08,
+      "loss": 0.7949,
+      "step": 21991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.667323574323095e-08,
+      "loss": 0.7817,
+      "step": 21992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.653794486612133e-08,
+      "loss": 0.7939,
+      "step": 21993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.64027477999685e-08,
+      "loss": 0.8579,
+      "step": 21994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.626764454735815e-08,
+      "loss": 0.8848,
+      "step": 21995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.61326351108749e-08,
+      "loss": 0.8262,
+      "step": 21996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.599771949310221e-08,
+      "loss": 0.7837,
+      "step": 21997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.586289769662028e-08,
+      "loss": 0.7803,
+      "step": 21998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.572816972400922e-08,
+      "loss": 0.7896,
+      "step": 21999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.559353557784701e-08,
+      "loss": 0.3405,
+      "step": 22000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.545899526070823e-08,
+      "loss": 0.7871,
+      "step": 22001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.53245487751675e-08,
+      "loss": 0.855,
+      "step": 22002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.519019612379776e-08,
+      "loss": 0.9209,
+      "step": 22003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.505593730916751e-08,
+      "loss": 0.8174,
+      "step": 22004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.492177233384692e-08,
+      "loss": 0.8203,
+      "step": 22005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.478770120040225e-08,
+      "loss": 0.7993,
+      "step": 22006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.465372391139816e-08,
+      "loss": 0.7871,
+      "step": 22007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.451984046939866e-08,
+      "loss": 0.8784,
+      "step": 22008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.43860508769645e-08,
+      "loss": 0.8867,
+      "step": 22009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.425235513665587e-08,
+      "loss": 0.8682,
+      "step": 22010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.41187532510296e-08,
+      "loss": 0.8623,
+      "step": 22011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.398524522264252e-08,
+      "loss": 0.3723,
+      "step": 22012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.385183105404872e-08,
+      "loss": 0.8306,
+      "step": 22013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.371851074780002e-08,
+      "loss": 0.875,
+      "step": 22014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.358528430644831e-08,
+      "loss": 0.8081,
+      "step": 22015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.345215173254152e-08,
+      "loss": 0.8501,
+      "step": 22016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.33191130286265e-08,
+      "loss": 0.8066,
+      "step": 22017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.318616819724901e-08,
+      "loss": 0.8076,
+      "step": 22018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.305331724095201e-08,
+      "loss": 0.8496,
+      "step": 22019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.292056016227735e-08,
+      "loss": 0.8135,
+      "step": 22020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.278789696376522e-08,
+      "loss": 0.8242,
+      "step": 22021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.26553276479536e-08,
+      "loss": 0.915,
+      "step": 22022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.252285221737823e-08,
+      "loss": 0.7434,
+      "step": 22023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.239047067457429e-08,
+      "loss": 0.8164,
+      "step": 22024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.225818302207367e-08,
+      "loss": 0.8262,
+      "step": 22025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.212598926240823e-08,
+      "loss": 0.8979,
+      "step": 22026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.199388939810594e-08,
+      "loss": 0.835,
+      "step": 22027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.186188343169478e-08,
+      "loss": 0.8398,
+      "step": 22028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.17299713657005e-08,
+      "loss": 0.8662,
+      "step": 22029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.159815320264554e-08,
+      "loss": 0.8037,
+      "step": 22030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.146642894505287e-08,
+      "loss": 0.832,
+      "step": 22031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.13347985954427e-08,
+      "loss": 0.8364,
+      "step": 22032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.12032621563319e-08,
+      "loss": 0.8413,
+      "step": 22033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.107181963023904e-08,
+      "loss": 0.792,
+      "step": 22034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.094047101967763e-08,
+      "loss": 0.7773,
+      "step": 22035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.080921632716067e-08,
+      "loss": 0.7363,
+      "step": 22036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.067805555519893e-08,
+      "loss": 0.8984,
+      "step": 22037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.054698870630262e-08,
+      "loss": 0.8843,
+      "step": 22038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.04160157829781e-08,
+      "loss": 0.8223,
+      "step": 22039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.02851367877322e-08,
+      "loss": 0.833,
+      "step": 22040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.015435172306853e-08,
+      "loss": 0.413,
+      "step": 22041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.002366059148893e-08,
+      "loss": 0.8398,
+      "step": 22042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.989306339549365e-08,
+      "loss": 0.8198,
+      "step": 22043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.976256013758122e-08,
+      "loss": 0.7471,
+      "step": 22044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.963215082024912e-08,
+      "loss": 0.8594,
+      "step": 22045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.950183544599145e-08,
+      "loss": 0.7759,
+      "step": 22046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.937161401730232e-08,
+      "loss": 0.8525,
+      "step": 22047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.924148653667198e-08,
+      "loss": 0.8032,
+      "step": 22048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.911145300659007e-08,
+      "loss": 0.7734,
+      "step": 22049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.89815134295452e-08,
+      "loss": 0.7983,
+      "step": 22050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.885166780802201e-08,
+      "loss": 0.7783,
+      "step": 22051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.872191614450576e-08,
+      "loss": 0.8135,
+      "step": 22052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.859225844147889e-08,
+      "loss": 0.8726,
+      "step": 22053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.84626947014211e-08,
+      "loss": 0.8535,
+      "step": 22054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.833322492681206e-08,
+      "loss": 0.8428,
+      "step": 22055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.820384912012758e-08,
+      "loss": 0.7847,
+      "step": 22056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.807456728384344e-08,
+      "loss": 0.833,
+      "step": 22057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.79453794204338e-08,
+      "loss": 0.8862,
+      "step": 22058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.781628553236887e-08,
+      "loss": 0.854,
+      "step": 22059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.768728562211948e-08,
+      "loss": 0.8594,
+      "step": 22060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.755837969215309e-08,
+      "loss": 0.835,
+      "step": 22061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.74295677449355e-08,
+      "loss": 0.7935,
+      "step": 22062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.730084978293196e-08,
+      "loss": 0.8413,
+      "step": 22063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.71722258086044e-08,
+      "loss": 0.8418,
+      "step": 22064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.704369582441363e-08,
+      "loss": 0.876,
+      "step": 22065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.691525983281934e-08,
+      "loss": 0.7993,
+      "step": 22066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.67869178362779e-08,
+      "loss": 0.8174,
+      "step": 22067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.665866983724514e-08,
+      "loss": 0.7949,
+      "step": 22068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.653051583817462e-08,
+      "loss": 0.3092,
+      "step": 22069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.640245584151718e-08,
+      "loss": 0.8813,
+      "step": 22070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.627448984972475e-08,
+      "loss": 0.855,
+      "step": 22071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.61466178652437e-08,
+      "loss": 0.8096,
+      "step": 22072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.601883989052095e-08,
+      "loss": 0.8545,
+      "step": 22073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.589115592800178e-08,
+      "loss": 0.8633,
+      "step": 22074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.576356598012814e-08,
+      "loss": 0.8467,
+      "step": 22075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.563607004934193e-08,
+      "loss": 0.8389,
+      "step": 22076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.550866813808123e-08,
+      "loss": 0.7935,
+      "step": 22077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.538136024878351e-08,
+      "loss": 0.8882,
+      "step": 22078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.525414638388574e-08,
+      "loss": 0.9019,
+      "step": 22079
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.512702654582039e-08,
+      "loss": 0.832,
+      "step": 22080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.500000073701997e-08,
+      "loss": 0.7964,
+      "step": 22081
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.487306895991476e-08,
+      "loss": 0.8408,
+      "step": 22082
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.474623121693281e-08,
+      "loss": 0.8438,
+      "step": 22083
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.461948751050108e-08,
+      "loss": 0.7925,
+      "step": 22084
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.44928378430443e-08,
+      "loss": 0.7778,
+      "step": 22085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.436628221698495e-08,
+      "loss": 0.7769,
+      "step": 22086
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.423982063474501e-08,
+      "loss": 0.8198,
+      "step": 22087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.411345309874419e-08,
+      "loss": 0.8462,
+      "step": 22088
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.398717961139834e-08,
+      "loss": 0.8237,
+      "step": 22089
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.386100017512556e-08,
+      "loss": 0.8765,
+      "step": 22090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.373491479233775e-08,
+      "loss": 0.8628,
+      "step": 22091
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.360892346544858e-08,
+      "loss": 0.792,
+      "step": 22092
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.348302619686776e-08,
+      "loss": 0.8247,
+      "step": 22093
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.33572229890045e-08,
+      "loss": 0.7827,
+      "step": 22094
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.32315138442652e-08,
+      "loss": 0.7915,
+      "step": 22095
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.310589876505403e-08,
+      "loss": 0.856,
+      "step": 22096
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.298037775377577e-08,
+      "loss": 0.3217,
+      "step": 22097
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.285495081283124e-08,
+      "loss": 0.8057,
+      "step": 22098
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.272961794461964e-08,
+      "loss": 0.8442,
+      "step": 22099
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.260437915153907e-08,
+      "loss": 0.7498,
+      "step": 22100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.247923443598649e-08,
+      "loss": 0.8403,
+      "step": 22101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.235418380035387e-08,
+      "loss": 0.8223,
+      "step": 22102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.222922724703597e-08,
+      "loss": 0.8589,
+      "step": 22103
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.210436477842198e-08,
+      "loss": 0.8218,
+      "step": 22104
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.197959639690057e-08,
+      "loss": 0.853,
+      "step": 22105
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.185492210486035e-08,
+      "loss": 0.8394,
+      "step": 22106
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.173034190468498e-08,
+      "loss": 0.8145,
+      "step": 22107
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.160585579875869e-08,
+      "loss": 0.853,
+      "step": 22108
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.148146378946231e-08,
+      "loss": 0.8013,
+      "step": 22109
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.135716587917619e-08,
+      "loss": 0.8467,
+      "step": 22110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.123296207027898e-08,
+      "loss": 0.8828,
+      "step": 22111
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.110885236514543e-08,
+      "loss": 0.7876,
+      "step": 22112
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.098483676615143e-08,
+      "loss": 0.8154,
+      "step": 22113
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.086091527566842e-08,
+      "loss": 0.8076,
+      "step": 22114
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.073708789606782e-08,
+      "loss": 0.8047,
+      "step": 22115
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.061335462971887e-08,
+      "loss": 0.8149,
+      "step": 22116
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.048971547898799e-08,
+      "loss": 0.8267,
+      "step": 22117
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.036617044624107e-08,
+      "loss": 0.8564,
+      "step": 22118
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.024271953384177e-08,
+      "loss": 0.876,
+      "step": 22119
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.011936274415211e-08,
+      "loss": 0.7744,
+      "step": 22120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.999610007953184e-08,
+      "loss": 0.8247,
+      "step": 22121
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.987293154233855e-08,
+      "loss": 0.8691,
+      "step": 22122
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.974985713492922e-08,
+      "loss": 0.8228,
+      "step": 22123
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.962687685965864e-08,
+      "loss": 0.3573,
+      "step": 22124
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.950399071887938e-08,
+      "loss": 0.8159,
+      "step": 22125
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.938119871494288e-08,
+      "loss": 0.8521,
+      "step": 22126
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.925850085019837e-08,
+      "loss": 0.8623,
+      "step": 22127
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.913589712699177e-08,
+      "loss": 0.8418,
+      "step": 22128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.901338754767063e-08,
+      "loss": 0.7773,
+      "step": 22129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.889097211457753e-08,
+      "loss": 0.3518,
+      "step": 22130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.876865083005447e-08,
+      "loss": 0.8887,
+      "step": 22131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.864642369644293e-08,
+      "loss": 0.4109,
+      "step": 22132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.852429071607992e-08,
+      "loss": 0.7539,
+      "step": 22133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.840225189130246e-08,
+      "loss": 0.8315,
+      "step": 22134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.828030722444535e-08,
+      "loss": 0.8057,
+      "step": 22135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.815845671784173e-08,
+      "loss": 0.7905,
+      "step": 22136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.803670037382249e-08,
+      "loss": 0.8452,
+      "step": 22137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.791503819471747e-08,
+      "loss": 0.8477,
+      "step": 22138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.779347018285366e-08,
+      "loss": 0.8525,
+      "step": 22139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.767199634055811e-08,
+      "loss": 0.8149,
+      "step": 22140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.755061667015284e-08,
+      "loss": 0.7695,
+      "step": 22141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.742933117396157e-08,
+      "loss": 0.8423,
+      "step": 22142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.730813985430407e-08,
+      "loss": 0.8193,
+      "step": 22143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.718704271349908e-08,
+      "loss": 0.8101,
+      "step": 22144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.706603975386362e-08,
+      "loss": 0.8086,
+      "step": 22145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.694513097771195e-08,
+      "loss": 0.8223,
+      "step": 22146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.682431638735777e-08,
+      "loss": 0.9019,
+      "step": 22147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.670359598511256e-08,
+      "loss": 0.8003,
+      "step": 22148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.658296977328561e-08,
+      "loss": 0.8091,
+      "step": 22149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.646243775418504e-08,
+      "loss": 0.8398,
+      "step": 22150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.634199993011626e-08,
+      "loss": 0.7944,
+      "step": 22151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.622165630338407e-08,
+      "loss": 0.8726,
+      "step": 22152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.610140687629053e-08,
+      "loss": 0.8408,
+      "step": 22153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.598125165113546e-08,
+      "loss": 0.8716,
+      "step": 22154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.586119063021869e-08,
+      "loss": 0.8779,
+      "step": 22155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.574122381583726e-08,
+      "loss": 0.8154,
+      "step": 22156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.562135121028546e-08,
+      "loss": 0.8662,
+      "step": 22157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.550157281585702e-08,
+      "loss": 0.8403,
+      "step": 22158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.538188863484396e-08,
+      "loss": 0.8335,
+      "step": 22159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.526229866953505e-08,
+      "loss": 0.855,
+      "step": 22160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.514280292221953e-08,
+      "loss": 0.895,
+      "step": 22161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.502340139518228e-08,
+      "loss": 0.7822,
+      "step": 22162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.490409409070809e-08,
+      "loss": 0.8696,
+      "step": 22163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.478488101107961e-08,
+      "loss": 0.8242,
+      "step": 22164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.466576215857779e-08,
+      "loss": 0.8555,
+      "step": 22165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.454673753548136e-08,
+      "loss": 0.7988,
+      "step": 22166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.442780714406683e-08,
+      "loss": 0.8262,
+      "step": 22167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.430897098661016e-08,
+      "loss": 0.8481,
+      "step": 22168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.419022906538564e-08,
+      "loss": 0.7974,
+      "step": 22169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.407158138266313e-08,
+      "loss": 0.8232,
+      "step": 22170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.395302794071412e-08,
+      "loss": 0.7983,
+      "step": 22171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.383456874180628e-08,
+      "loss": 0.8062,
+      "step": 22172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.371620378820555e-08,
+      "loss": 0.9043,
+      "step": 22173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.359793308217677e-08,
+      "loss": 0.897,
+      "step": 22174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.347975662598205e-08,
+      "loss": 0.8018,
+      "step": 22175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.336167442188346e-08,
+      "loss": 0.8159,
+      "step": 22176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.32436864721392e-08,
+      "loss": 0.814,
+      "step": 22177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.312579277900744e-08,
+      "loss": 0.8286,
+      "step": 22178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.300799334474252e-08,
+      "loss": 0.812,
+      "step": 22179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.289028817159871e-08,
+      "loss": 0.8154,
+      "step": 22180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.277267726182758e-08,
+      "loss": 0.8584,
+      "step": 22181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.265516061767952e-08,
+      "loss": 0.8521,
+      "step": 22182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.253773824140275e-08,
+      "loss": 0.7539,
+      "step": 22183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.24204101352438e-08,
+      "loss": 0.8091,
+      "step": 22184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.230317630144812e-08,
+      "loss": 0.8701,
+      "step": 22185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.218603674225722e-08,
+      "loss": 0.832,
+      "step": 22186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.206899145991263e-08,
+      "loss": 0.8364,
+      "step": 22187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.195204045665316e-08,
+      "loss": 0.8672,
+      "step": 22188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.183518373471754e-08,
+      "loss": 0.8442,
+      "step": 22189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.171842129634066e-08,
+      "loss": 0.8247,
+      "step": 22190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.160175314375573e-08,
+      "loss": 0.8135,
+      "step": 22191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.148517927919596e-08,
+      "loss": 0.8022,
+      "step": 22192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.136869970489124e-08,
+      "loss": 0.8706,
+      "step": 22193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.125231442306924e-08,
+      "loss": 0.7808,
+      "step": 22194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.113602343595816e-08,
+      "loss": 0.4043,
+      "step": 22195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.101982674578067e-08,
+      "loss": 0.7583,
+      "step": 22196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.090372435476167e-08,
+      "loss": 0.8286,
+      "step": 22197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.07877162651216e-08,
+      "loss": 0.9004,
+      "step": 22198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.067180247908035e-08,
+      "loss": 0.8071,
+      "step": 22199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.055598299885503e-08,
+      "loss": 0.7261,
+      "step": 22200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.044025782666164e-08,
+      "loss": 0.7856,
+      "step": 22201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.032462696471343e-08,
+      "loss": 0.8374,
+      "step": 22202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.020909041522417e-08,
+      "loss": 0.8589,
+      "step": 22203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.009364818040265e-08,
+      "loss": 0.8818,
+      "step": 22204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.997830026245878e-08,
+      "loss": 0.8403,
+      "step": 22205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.986304666359911e-08,
+      "loss": 0.8242,
+      "step": 22206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.974788738602744e-08,
+      "loss": 0.7954,
+      "step": 22207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.963282243194758e-08,
+      "loss": 0.8296,
+      "step": 22208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.951785180356219e-08,
+      "loss": 0.8032,
+      "step": 22209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.940297550306895e-08,
+      "loss": 0.8652,
+      "step": 22210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.928819353266669e-08,
+      "loss": 0.8501,
+      "step": 22211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.91735058945514e-08,
+      "loss": 0.7769,
+      "step": 22212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.90589125909169e-08,
+      "loss": 0.73,
+      "step": 22213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.894441362395532e-08,
+      "loss": 0.8276,
+      "step": 22214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.883000899585767e-08,
+      "loss": 0.8242,
+      "step": 22215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.871569870881223e-08,
+      "loss": 0.8149,
+      "step": 22216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.860148276500667e-08,
+      "loss": 0.8501,
+      "step": 22217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.848736116662535e-08,
+      "loss": 0.7959,
+      "step": 22218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.83733339158521e-08,
+      "loss": 0.8809,
+      "step": 22219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.825940101486794e-08,
+      "loss": 0.8657,
+      "step": 22220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.81455624658528e-08,
+      "loss": 0.8208,
+      "step": 22221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.803181827098548e-08,
+      "loss": 0.8994,
+      "step": 22222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.791816843244036e-08,
+      "loss": 0.855,
+      "step": 22223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.78046129523935e-08,
+      "loss": 0.8657,
+      "step": 22224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.769115183301589e-08,
+      "loss": 0.8604,
+      "step": 22225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.757778507647916e-08,
+      "loss": 0.8506,
+      "step": 22226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.746451268495213e-08,
+      "loss": 0.8081,
+      "step": 22227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.735133466060139e-08,
+      "loss": 0.8013,
+      "step": 22228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.723825100559245e-08,
+      "loss": 0.811,
+      "step": 22229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.712526172208911e-08,
+      "loss": 0.8149,
+      "step": 22230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.701236681225298e-08,
+      "loss": 0.8257,
+      "step": 22231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.689956627824345e-08,
+      "loss": 0.792,
+      "step": 22232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.678686012221825e-08,
+      "loss": 0.4351,
+      "step": 22233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.667424834633452e-08,
+      "loss": 0.9224,
+      "step": 22234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.656173095274665e-08,
+      "loss": 0.2833,
+      "step": 22235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.644930794360682e-08,
+      "loss": 0.8164,
+      "step": 22236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.633697932106609e-08,
+      "loss": 0.8154,
+      "step": 22237
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.622474508727383e-08,
+      "loss": 0.7637,
+      "step": 22238
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.611260524437613e-08,
+      "loss": 0.7812,
+      "step": 22239
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.600055979451958e-08,
+      "loss": 0.8462,
+      "step": 22240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.588860873984693e-08,
+      "loss": 0.8496,
+      "step": 22241
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.577675208250034e-08,
+      "loss": 0.8076,
+      "step": 22242
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.566498982462088e-08,
+      "loss": 0.8882,
+      "step": 22243
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.555332196834463e-08,
+      "loss": 0.8428,
+      "step": 22244
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.544174851580987e-08,
+      "loss": 0.7642,
+      "step": 22245
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.533026946914988e-08,
+      "loss": 0.834,
+      "step": 22246
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.521888483049743e-08,
+      "loss": 0.8159,
+      "step": 22247
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.510759460198469e-08,
+      "loss": 0.7876,
+      "step": 22248
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.499639878573993e-08,
+      "loss": 0.835,
+      "step": 22249
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.488529738389093e-08,
+      "loss": 0.8638,
+      "step": 22250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.477429039856265e-08,
+      "loss": 0.8662,
+      "step": 22251
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.466337783187892e-08,
+      "loss": 0.8169,
+      "step": 22252
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.455255968596197e-08,
+      "loss": 0.8525,
+      "step": 22253
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.444183596293229e-08,
+      "loss": 0.853,
+      "step": 22254
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.433120666490711e-08,
+      "loss": 0.8057,
+      "step": 22255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.422067179400415e-08,
+      "loss": 0.8936,
+      "step": 22256
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.411023135233729e-08,
+      "loss": 0.8325,
+      "step": 22257
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.399988534201984e-08,
+      "loss": 0.8652,
+      "step": 22258
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.388963376516289e-08,
+      "loss": 0.8335,
+      "step": 22259
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.377947662387473e-08,
+      "loss": 0.8042,
+      "step": 22260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.366941392026482e-08,
+      "loss": 0.8584,
+      "step": 22261
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.355944565643701e-08,
+      "loss": 0.8853,
+      "step": 22262
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.34495718344963e-08,
+      "loss": 0.8716,
+      "step": 22263
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.333979245654431e-08,
+      "loss": 0.7871,
+      "step": 22264
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.323010752468106e-08,
+      "loss": 0.8247,
+      "step": 22265
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.312051704100486e-08,
+      "loss": 0.8296,
+      "step": 22266
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.301102100761347e-08,
+      "loss": 0.8013,
+      "step": 22267
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.290161942660023e-08,
+      "loss": 0.8257,
+      "step": 22268
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.279231230005956e-08,
+      "loss": 0.8374,
+      "step": 22269
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.268309963008201e-08,
+      "loss": 0.875,
+      "step": 22270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.257398141875648e-08,
+      "loss": 0.8281,
+      "step": 22271
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.246495766817184e-08,
+      "loss": 0.7969,
+      "step": 22272
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.235602838041254e-08,
+      "loss": 0.8281,
+      "step": 22273
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.224719355756304e-08,
+      "loss": 0.8281,
+      "step": 22274
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.213845320170609e-08,
+      "loss": 0.8057,
+      "step": 22275
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.20298073149217e-08,
+      "loss": 0.8486,
+      "step": 22276
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.192125589928821e-08,
+      "loss": 0.8203,
+      "step": 22277
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.18127989568823e-08,
+      "loss": 0.8105,
+      "step": 22278
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.170443648977953e-08,
+      "loss": 0.8066,
+      "step": 22279
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.159616850005268e-08,
+      "loss": 0.3376,
+      "step": 22280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.148799498977231e-08,
+      "loss": 0.8218,
+      "step": 22281
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.137991596100956e-08,
+      "loss": 0.8491,
+      "step": 22282
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.127193141583166e-08,
+      "loss": 0.8892,
+      "step": 22283
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.116404135630305e-08,
+      "loss": 0.8853,
+      "step": 22284
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.105624578448988e-08,
+      "loss": 0.8472,
+      "step": 22285
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.094854470245326e-08,
+      "loss": 0.3561,
+      "step": 22286
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.084093811225322e-08,
+      "loss": 0.7852,
+      "step": 22287
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.073342601595033e-08,
+      "loss": 0.8071,
+      "step": 22288
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.062600841560018e-08,
+      "loss": 0.8706,
+      "step": 22289
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.051868531325778e-08,
+      "loss": 0.79,
+      "step": 22290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.041145671097592e-08,
+      "loss": 0.7915,
+      "step": 22291
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.030432261080688e-08,
+      "loss": 0.8755,
+      "step": 22292
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.019728301480066e-08,
+      "loss": 0.3276,
+      "step": 22293
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.009033792500451e-08,
+      "loss": 0.8452,
+      "step": 22294
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.998348734346405e-08,
+      "loss": 0.8545,
+      "step": 22295
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.987673127222481e-08,
+      "loss": 0.9048,
+      "step": 22296
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9770069713328e-08,
+      "loss": 0.7988,
+      "step": 22297
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.96635026688147e-08,
+      "loss": 0.8101,
+      "step": 22298
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.955703014072334e-08,
+      "loss": 0.8174,
+      "step": 22299
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.945065213109058e-08,
+      "loss": 0.8335,
+      "step": 22300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9344368641953145e-08,
+      "loss": 0.8271,
+      "step": 22301
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9238179675342735e-08,
+      "loss": 0.7832,
+      "step": 22302
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.913208523329217e-08,
+      "loss": 0.8267,
+      "step": 22303
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.902608531783094e-08,
+      "loss": 0.8535,
+      "step": 22304
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.892017993098575e-08,
+      "loss": 0.793,
+      "step": 22305
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.881436907478444e-08,
+      "loss": 0.8169,
+      "step": 22306
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.870865275125037e-08,
+      "loss": 0.8359,
+      "step": 22307
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8603030962405824e-08,
+      "loss": 0.8374,
+      "step": 22308
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.849750371027252e-08,
+      "loss": 0.7969,
+      "step": 22309
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8392070996868834e-08,
+      "loss": 0.7622,
+      "step": 22310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8286732824212045e-08,
+      "loss": 0.8711,
+      "step": 22311
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8181489194316654e-08,
+      "loss": 0.8252,
+      "step": 22312
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.807634010919661e-08,
+      "loss": 0.8735,
+      "step": 22313
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.797128557086418e-08,
+      "loss": 0.3611,
+      "step": 22314
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7866325581328874e-08,
+      "loss": 0.8291,
+      "step": 22315
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.776146014259798e-08,
+      "loss": 0.8005,
+      "step": 22316
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.765668925667878e-08,
+      "loss": 0.8179,
+      "step": 22317
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.755201292557522e-08,
+      "loss": 0.8735,
+      "step": 22318
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.744743115129014e-08,
+      "loss": 0.8447,
+      "step": 22319
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7342943935823624e-08,
+      "loss": 0.8628,
+      "step": 22320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.723855128117572e-08,
+      "loss": 0.8184,
+      "step": 22321
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.713425318934318e-08,
+      "loss": 0.854,
+      "step": 22322
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.703004966232106e-08,
+      "loss": 0.8657,
+      "step": 22323
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.692594070210333e-08,
+      "loss": 0.8306,
+      "step": 22324
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.682192631068173e-08,
+      "loss": 0.8218,
+      "step": 22325
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.671800649004578e-08,
+      "loss": 0.8193,
+      "step": 22326
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6614181242184985e-08,
+      "loss": 0.9023,
+      "step": 22327
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.651045056908389e-08,
+      "loss": 0.8311,
+      "step": 22328
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.640681447272811e-08,
+      "loss": 0.8525,
+      "step": 22329
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6303272955099964e-08,
+      "loss": 0.7954,
+      "step": 22330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6199826018180634e-08,
+      "loss": 0.916,
+      "step": 22331
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.60964736639491e-08,
+      "loss": 0.8633,
+      "step": 22332
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.599321589438267e-08,
+      "loss": 0.853,
+      "step": 22333
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.589005271145642e-08,
+      "loss": 0.9058,
+      "step": 22334
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.578698411714489e-08,
+      "loss": 0.7426,
+      "step": 22335
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.568401011341929e-08,
+      "loss": 0.8896,
+      "step": 22336
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5581130702249687e-08,
+      "loss": 0.7527,
+      "step": 22337
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.547834588560508e-08,
+      "loss": 0.8315,
+      "step": 22338
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5375655665450555e-08,
+      "loss": 0.894,
+      "step": 22339
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.527306004375177e-08,
+      "loss": 0.8672,
+      "step": 22340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.517055902247159e-08,
+      "loss": 0.814,
+      "step": 22341
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.506815260357068e-08,
+      "loss": 0.8159,
+      "step": 22342
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.496584078900802e-08,
+      "loss": 0.8398,
+      "step": 22343
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.486362358074093e-08,
+      "loss": 0.7974,
+      "step": 22344
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4761500980725634e-08,
+      "loss": 0.8618,
+      "step": 22345
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.465947299091556e-08,
+      "loss": 0.8018,
+      "step": 22346
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.455753961326249e-08,
+      "loss": 0.8242,
+      "step": 22347
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4455700849717074e-08,
+      "loss": 0.7996,
+      "step": 22348
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.435395670222721e-08,
+      "loss": 0.7764,
+      "step": 22349
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4252307172739105e-08,
+      "loss": 0.8369,
+      "step": 22350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4150752263198435e-08,
+      "loss": 0.8184,
+      "step": 22351
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.404929197554698e-08,
+      "loss": 0.8501,
+      "step": 22352
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3947926311727074e-08,
+      "loss": 0.8311,
+      "step": 22353
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3846655273677164e-08,
+      "loss": 0.7881,
+      "step": 22354
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.374547886333459e-08,
+      "loss": 0.8804,
+      "step": 22355
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.364439708263558e-08,
+      "loss": 0.8447,
+      "step": 22356
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.354340993351359e-08,
+      "loss": 0.8652,
+      "step": 22357
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.344251741790096e-08,
+      "loss": 0.8276,
+      "step": 22358
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3341719537727824e-08,
+      "loss": 0.8276,
+      "step": 22359
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.324101629492207e-08,
+      "loss": 0.8618,
+      "step": 22360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.31404076914116e-08,
+      "loss": 0.7571,
+      "step": 22361
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3039893729119886e-08,
+      "loss": 0.8096,
+      "step": 22362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.293947440997038e-08,
+      "loss": 0.7881,
+      "step": 22363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2839149735884886e-08,
+      "loss": 0.769,
+      "step": 22364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.273891970878131e-08,
+      "loss": 0.8257,
+      "step": 22365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.263878433057867e-08,
+      "loss": 0.8345,
+      "step": 22366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.253874360319211e-08,
+      "loss": 0.8716,
+      "step": 22367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.24387975285362e-08,
+      "loss": 0.3743,
+      "step": 22368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2338946108521635e-08,
+      "loss": 0.8271,
+      "step": 22369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.223918934505967e-08,
+      "loss": 0.8315,
+      "step": 22370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.213952724005877e-08,
+      "loss": 0.7687,
+      "step": 22371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2039959795425756e-08,
+      "loss": 0.8306,
+      "step": 22372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.19404870130652e-08,
+      "loss": 0.8501,
+      "step": 22373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.184110889488059e-08,
+      "loss": 0.8584,
+      "step": 22374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.174182544277262e-08,
+      "loss": 0.8193,
+      "step": 22375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1642636658640886e-08,
+      "loss": 0.8242,
+      "step": 22376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1543542544383874e-08,
+      "loss": 0.856,
+      "step": 22377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1444543101895613e-08,
+      "loss": 0.8501,
+      "step": 22378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.134563833307182e-08,
+      "loss": 0.2734,
+      "step": 22379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.124682823980431e-08,
+      "loss": 0.8184,
+      "step": 22380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1148112823982684e-08,
+      "loss": 0.7881,
+      "step": 22381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1049492087496545e-08,
+      "loss": 0.8608,
+      "step": 22382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0950966032231595e-08,
+      "loss": 0.8491,
+      "step": 22383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.085253466007412e-08,
+      "loss": 0.8477,
+      "step": 22384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0754197972905927e-08,
+      "loss": 0.7622,
+      "step": 22385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0655955972609416e-08,
+      "loss": 0.8477,
+      "step": 22386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0557808661063635e-08,
+      "loss": 0.8354,
+      "step": 22387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.045975604014597e-08,
+      "loss": 0.8984,
+      "step": 22388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.03617981117327e-08,
+      "loss": 0.8652,
+      "step": 22389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0263934877697874e-08,
+      "loss": 0.8413,
+      "step": 22390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.016616633991445e-08,
+      "loss": 0.8638,
+      "step": 22391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.006849250025148e-08,
+      "loss": 0.8213,
+      "step": 22392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.997091336057858e-08,
+      "loss": 0.8145,
+      "step": 22393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.987342892276259e-08,
+      "loss": 0.8638,
+      "step": 22394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9776039188668136e-08,
+      "loss": 0.7842,
+      "step": 22395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.967874416015872e-08,
+      "loss": 0.791,
+      "step": 22396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9581543839095637e-08,
+      "loss": 0.8159,
+      "step": 22397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9484438227338504e-08,
+      "loss": 0.8833,
+      "step": 22398
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9387427326745287e-08,
+      "loss": 0.8379,
+      "step": 22399
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9290511139171716e-08,
+      "loss": 0.8057,
+      "step": 22400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.919368966647187e-08,
+      "loss": 0.8242,
+      "step": 22401
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.909696291049815e-08,
+      "loss": 0.8301,
+      "step": 22402
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.90003308731013e-08,
+      "loss": 0.7788,
+      "step": 22403
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.890379355612984e-08,
+      "loss": 0.8315,
+      "step": 22404
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8807350961431186e-08,
+      "loss": 0.9082,
+      "step": 22405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8711003090849415e-08,
+      "loss": 0.8818,
+      "step": 22406
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8614749946229166e-08,
+      "loss": 0.8652,
+      "step": 22407
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.851859152941063e-08,
+      "loss": 0.8398,
+      "step": 22408
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8422527842234004e-08,
+      "loss": 0.8037,
+      "step": 22409
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.832655888653726e-08,
+      "loss": 0.7749,
+      "step": 22410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.823068466415615e-08,
+      "loss": 0.8643,
+      "step": 22411
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.813490517692532e-08,
+      "loss": 0.8555,
+      "step": 22412
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8039220426676636e-08,
+      "loss": 0.876,
+      "step": 22413
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7943630415241415e-08,
+      "loss": 0.8433,
+      "step": 22414
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.784813514444764e-08,
+      "loss": 0.7671,
+      "step": 22415
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.775273461612273e-08,
+      "loss": 0.7949,
+      "step": 22416
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.76574288320919e-08,
+      "loss": 0.917,
+      "step": 22417
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7562217794178137e-08,
+      "loss": 0.7725,
+      "step": 22418
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.746710150420275e-08,
+      "loss": 0.3336,
+      "step": 22419
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.737207996398707e-08,
+      "loss": 0.8403,
+      "step": 22420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.727715317534742e-08,
+      "loss": 0.7866,
+      "step": 22421
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.718232114010013e-08,
+      "loss": 0.8999,
+      "step": 22422
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.708758386005985e-08,
+      "loss": 0.8247,
+      "step": 22423
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6992941337038467e-08,
+      "loss": 0.8643,
+      "step": 22424
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.689839357284787e-08,
+      "loss": 0.8291,
+      "step": 22425
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.680394056929549e-08,
+      "loss": 0.8052,
+      "step": 22426
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.670958232818934e-08,
+      "loss": 0.8359,
+      "step": 22427
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.661531885133464e-08,
+      "loss": 0.7944,
+      "step": 22428
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.652115014053382e-08,
+      "loss": 0.8301,
+      "step": 22429
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.642707619758935e-08,
+      "loss": 0.8135,
+      "step": 22430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.633309702430089e-08,
+      "loss": 0.8359,
+      "step": 22431
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.623921262246589e-08,
+      "loss": 0.8657,
+      "step": 22432
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.614542299388125e-08,
+      "loss": 0.7993,
+      "step": 22433
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.605172814034109e-08,
+      "loss": 0.833,
+      "step": 22434
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.595812806363731e-08,
+      "loss": 0.8608,
+      "step": 22435
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5864622765561804e-08,
+      "loss": 0.855,
+      "step": 22436
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5771212247902597e-08,
+      "loss": 0.8799,
+      "step": 22437
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.567789651244658e-08,
+      "loss": 0.8589,
+      "step": 22438
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5584675560979565e-08,
+      "loss": 0.8784,
+      "step": 22439
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.549154939528566e-08,
+      "loss": 0.8608,
+      "step": 22440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.539851801714512e-08,
+      "loss": 0.8643,
+      "step": 22441
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5305581428338186e-08,
+      "loss": 0.8105,
+      "step": 22442
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5212739630643434e-08,
+      "loss": 0.7974,
+      "step": 22443
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.511999262583666e-08,
+      "loss": 0.8467,
+      "step": 22444
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.502734041569201e-08,
+      "loss": 0.3201,
+      "step": 22445
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.493478300198306e-08,
+      "loss": 0.8467,
+      "step": 22446
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4842320386480065e-08,
+      "loss": 0.812,
+      "step": 22447
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4749952570951047e-08,
+      "loss": 0.7969,
+      "step": 22448
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.46576795571646e-08,
+      "loss": 0.834,
+      "step": 22449
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.456550134688487e-08,
+      "loss": 0.7725,
+      "step": 22450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.447341794187598e-08,
+      "loss": 0.8354,
+      "step": 22451
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.438142934389933e-08,
+      "loss": 0.8286,
+      "step": 22452
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.428953555471571e-08,
+      "loss": 0.8672,
+      "step": 22453
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.419773657608206e-08,
+      "loss": 0.8784,
+      "step": 22454
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4106032409755307e-08,
+      "loss": 0.8057,
+      "step": 22455
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4014423057489594e-08,
+      "loss": 0.7549,
+      "step": 22456
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.392290852103742e-08,
+      "loss": 0.853,
+      "step": 22457
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3831488802150154e-08,
+      "loss": 0.8403,
+      "step": 22458
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.374016390257641e-08,
+      "loss": 0.7881,
+      "step": 22459
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.36489338240631e-08,
+      "loss": 0.8423,
+      "step": 22460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.355779856835607e-08,
+      "loss": 0.8467,
+      "step": 22461
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.346675813719892e-08,
+      "loss": 0.8367,
+      "step": 22462
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3375812532333027e-08,
+      "loss": 0.3656,
+      "step": 22463
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3284961755498124e-08,
+      "loss": 0.8315,
+      "step": 22464
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3194205808433363e-08,
+      "loss": 0.854,
+      "step": 22465
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3103544692874586e-08,
+      "loss": 0.8018,
+      "step": 22466
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.30129784105554e-08,
+      "loss": 0.8247,
+      "step": 22467
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.292250696320943e-08,
+      "loss": 0.8262,
+      "step": 22468
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.283213035256695e-08,
+      "loss": 0.8276,
+      "step": 22469
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2741848580358234e-08,
+      "loss": 0.8574,
+      "step": 22470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.265166164830859e-08,
+      "loss": 0.8311,
+      "step": 22471
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.25615695581455e-08,
+      "loss": 0.8062,
+      "step": 22472
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.247157231159094e-08,
+      "loss": 0.7832,
+      "step": 22473
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2381669910367406e-08,
+      "loss": 0.8076,
+      "step": 22474
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.229186235619465e-08,
+      "loss": 0.8643,
+      "step": 22475
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.220214965079073e-08,
+      "loss": 0.8345,
+      "step": 22476
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.21125317958726e-08,
+      "loss": 0.771,
+      "step": 22477
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.202300879315446e-08,
+      "loss": 0.8496,
+      "step": 22478
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1933580644348824e-08,
+      "loss": 0.855,
+      "step": 22479
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.18442473511671e-08,
+      "loss": 0.8042,
+      "step": 22480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1755008915317364e-08,
+      "loss": 0.8359,
+      "step": 22481
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1665865338507695e-08,
+      "loss": 0.8115,
+      "step": 22482
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1576816622443395e-08,
+      "loss": 0.8501,
+      "step": 22483
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.14878627688281e-08,
+      "loss": 0.8501,
+      "step": 22484
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1399003779363786e-08,
+      "loss": 0.8052,
+      "step": 22485
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.131023965575021e-08,
+      "loss": 0.8496,
+      "step": 22486
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.122157039968544e-08,
+      "loss": 0.8809,
+      "step": 22487
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1132996012866464e-08,
+      "loss": 0.8784,
+      "step": 22488
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.104451649698749e-08,
+      "loss": 0.8945,
+      "step": 22489
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.095613185374103e-08,
+      "loss": 0.8145,
+      "step": 22490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.086784208481853e-08,
+      "loss": 0.8408,
+      "step": 22491
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.077964719190863e-08,
+      "loss": 0.7812,
+      "step": 22492
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.069154717669888e-08,
+      "loss": 0.7896,
+      "step": 22493
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.060354204087458e-08,
+      "loss": 0.8721,
+      "step": 22494
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.051563178611939e-08,
+      "loss": 0.8374,
+      "step": 22495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.042781641411586e-08,
+      "loss": 0.9238,
+      "step": 22496
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.03400959265432e-08,
+      "loss": 0.8276,
+      "step": 22497
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.025247032508006e-08,
+      "loss": 0.8418,
+      "step": 22498
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0164939611402887e-08,
+      "loss": 0.8369,
+      "step": 22499
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.007750378718589e-08,
+      "loss": 0.8726,
+      "step": 22500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.999016285410218e-08,
+      "loss": 0.8394,
+      "step": 22501
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9902916813822635e-08,
+      "loss": 0.7473,
+      "step": 22502
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.981576566801648e-08,
+      "loss": 0.3722,
+      "step": 22503
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9728709418351274e-08,
+      "loss": 0.8198,
+      "step": 22504
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.964174806649179e-08,
+      "loss": 0.292,
+      "step": 22505
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.955488161410226e-08,
+      "loss": 0.814,
+      "step": 22506
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9468110062845234e-08,
+      "loss": 0.8203,
+      "step": 22507
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9381433414379386e-08,
+      "loss": 0.8213,
+      "step": 22508
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.92948516703634e-08,
+      "loss": 0.8232,
+      "step": 22509
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.920836483245483e-08,
+      "loss": 0.7776,
+      "step": 22510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.912197290230735e-08,
+      "loss": 0.8047,
+      "step": 22511
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.903567588157353e-08,
+      "loss": 0.8823,
+      "step": 22512
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.894947377190483e-08,
+      "loss": 0.8218,
+      "step": 22513
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.886336657495049e-08,
+      "loss": 0.8066,
+      "step": 22514
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8777354292357515e-08,
+      "loss": 0.7865,
+      "step": 22515
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.869143692577182e-08,
+      "loss": 0.7988,
+      "step": 22516
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.860561447683708e-08,
+      "loss": 0.7974,
+      "step": 22517
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.851988694719477e-08,
+      "loss": 0.8433,
+      "step": 22518
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.84342543384858e-08,
+      "loss": 0.8223,
+      "step": 22519
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.834871665234774e-08,
+      "loss": 0.7812,
+      "step": 22520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.826327389041706e-08,
+      "loss": 0.873,
+      "step": 22521
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8177926054329126e-08,
+      "loss": 0.8447,
+      "step": 22522
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8092673145715965e-08,
+      "loss": 0.8916,
+      "step": 22523
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.800751516620904e-08,
+      "loss": 0.8315,
+      "step": 22524
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7922452117438166e-08,
+      "loss": 0.7637,
+      "step": 22525
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.783748400102927e-08,
+      "loss": 0.812,
+      "step": 22526
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.775261081860881e-08,
+      "loss": 0.7759,
+      "step": 22527
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.766783257180051e-08,
+      "loss": 0.833,
+      "step": 22528
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.758314926222584e-08,
+      "loss": 0.8135,
+      "step": 22529
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.749856089150572e-08,
+      "loss": 0.8491,
+      "step": 22530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.741406746125831e-08,
+      "loss": 0.8188,
+      "step": 22531
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.732966897309953e-08,
+      "loss": 0.7939,
+      "step": 22532
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.724536542864476e-08,
+      "loss": 0.8257,
+      "step": 22533
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.716115682950605e-08,
+      "loss": 0.854,
+      "step": 22534
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.707704317729488e-08,
+      "loss": 0.8188,
+      "step": 22535
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.699302447362052e-08,
+      "loss": 0.8184,
+      "step": 22536
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6909100720090576e-08,
+      "loss": 0.8423,
+      "step": 22537
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.682527191831042e-08,
+      "loss": 0.854,
+      "step": 22538
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.674153806988323e-08,
+      "loss": 0.7905,
+      "step": 22539
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.665789917641216e-08,
+      "loss": 0.3323,
+      "step": 22540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6574355239496486e-08,
+      "loss": 0.835,
+      "step": 22541
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.649090626073437e-08,
+      "loss": 0.8896,
+      "step": 22542
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.640755224172343e-08,
+      "loss": 0.8628,
+      "step": 22543
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.632429318405739e-08,
+      "loss": 0.8511,
+      "step": 22544
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6241129089329416e-08,
+      "loss": 0.9062,
+      "step": 22545
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6158059959130466e-08,
+      "loss": 0.8774,
+      "step": 22546
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6075085795049815e-08,
+      "loss": 0.834,
+      "step": 22547
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.599220659867508e-08,
+      "loss": 0.853,
+      "step": 22548
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5909422371591676e-08,
+      "loss": 0.8604,
+      "step": 22549
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5826733115383316e-08,
+      "loss": 0.8359,
+      "step": 22550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5744138831632634e-08,
+      "loss": 0.3455,
+      "step": 22551
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5661639521918924e-08,
+      "loss": 0.3372,
+      "step": 22552
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5579235187820916e-08,
+      "loss": 0.9072,
+      "step": 22553
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.549692583091513e-08,
+      "loss": 0.8115,
+      "step": 22554
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5414711452775864e-08,
+      "loss": 0.8623,
+      "step": 22555
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.533259205497686e-08,
+      "loss": 0.8652,
+      "step": 22556
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5250567639088516e-08,
+      "loss": 0.8384,
+      "step": 22557
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.516863820668015e-08,
+      "loss": 0.7861,
+      "step": 22558
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5086803759319385e-08,
+      "loss": 0.7761,
+      "step": 22559
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.50050642985722e-08,
+      "loss": 0.3611,
+      "step": 22560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.492341982600178e-08,
+      "loss": 0.8115,
+      "step": 22561
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.484187034317022e-08,
+      "loss": 0.8242,
+      "step": 22562
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.476041585163792e-08,
+      "loss": 0.813,
+      "step": 22563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.467905635296309e-08,
+      "loss": 0.8452,
+      "step": 22564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.459779184870227e-08,
+      "loss": 0.7773,
+      "step": 22565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.451662234040975e-08,
+      "loss": 0.8247,
+      "step": 22566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4435547829639314e-08,
+      "loss": 0.8262,
+      "step": 22567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4354568317941374e-08,
+      "loss": 0.8091,
+      "step": 22568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.42736838068658e-08,
+      "loss": 0.7798,
+      "step": 22569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.419289429795969e-08,
+      "loss": 0.8237,
+      "step": 22570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.411219979276847e-08,
+      "loss": 0.8257,
+      "step": 22571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.403160029283592e-08,
+      "loss": 0.8657,
+      "step": 22572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3951095799704126e-08,
+      "loss": 0.834,
+      "step": 22573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.387068631491353e-08,
+      "loss": 0.7891,
+      "step": 22574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.379037184000289e-08,
+      "loss": 0.8345,
+      "step": 22575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.371015237650765e-08,
+      "loss": 0.8433,
+      "step": 22576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.363002792596326e-08,
+      "loss": 0.8013,
+      "step": 22577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3549998489902924e-08,
+      "loss": 0.8076,
+      "step": 22578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3470064069856536e-08,
+      "loss": 0.8799,
+      "step": 22579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3390224667353974e-08,
+      "loss": 0.7935,
+      "step": 22580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.331048028392347e-08,
+      "loss": 0.8081,
+      "step": 22581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.323083092108992e-08,
+      "loss": 0.8389,
+      "step": 22582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.315127658037709e-08,
+      "loss": 0.8765,
+      "step": 22583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3071817263307105e-08,
+      "loss": 0.8574,
+      "step": 22584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.299245297139986e-08,
+      "loss": 0.8589,
+      "step": 22585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.291318370617469e-08,
+      "loss": 0.8149,
+      "step": 22586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.283400946914705e-08,
+      "loss": 0.8071,
+      "step": 22587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.275493026183185e-08,
+      "loss": 0.855,
+      "step": 22588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.267594608574287e-08,
+      "loss": 0.8477,
+      "step": 22589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.259705694239001e-08,
+      "loss": 0.8428,
+      "step": 22590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.251826283328374e-08,
+      "loss": 0.9233,
+      "step": 22591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.243956375993007e-08,
+      "loss": 0.8101,
+      "step": 22592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.236095972383557e-08,
+      "loss": 0.856,
+      "step": 22593
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.22824507265046e-08,
+      "loss": 0.752,
+      "step": 22594
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.220403676943762e-08,
+      "loss": 0.8091,
+      "step": 22595
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.212571785413676e-08,
+      "loss": 0.8105,
+      "step": 22596
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.204749398209861e-08,
+      "loss": 0.8145,
+      "step": 22597
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.19693651548203e-08,
+      "loss": 0.833,
+      "step": 22598
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1891331373797294e-08,
+      "loss": 0.8516,
+      "step": 22599
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.181339264052119e-08,
+      "loss": 0.8213,
+      "step": 22600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.173554895648412e-08,
+      "loss": 0.7812,
+      "step": 22601
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.165780032317489e-08,
+      "loss": 0.8154,
+      "step": 22602
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1580146742081207e-08,
+      "loss": 0.8828,
+      "step": 22603
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.150258821468854e-08,
+      "loss": 0.8447,
+      "step": 22604
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.142512474248072e-08,
+      "loss": 0.8179,
+      "step": 22605
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1347756326939874e-08,
+      "loss": 0.7693,
+      "step": 22606
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1270482969545936e-08,
+      "loss": 0.8687,
+      "step": 22607
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.119330467177717e-08,
+      "loss": 0.7925,
+      "step": 22608
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1116221435110727e-08,
+      "loss": 0.8823,
+      "step": 22609
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.103923326102099e-08,
+      "loss": 0.8037,
+      "step": 22610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.096234015098065e-08,
+      "loss": 0.8081,
+      "step": 22611
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.088554210646133e-08,
+      "loss": 0.8203,
+      "step": 22612
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.080883912893129e-08,
+      "loss": 0.8013,
+      "step": 22613
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.07322312198588e-08,
+      "loss": 0.8398,
+      "step": 22614
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.065571838070936e-08,
+      "loss": 0.8247,
+      "step": 22615
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.057930061294734e-08,
+      "loss": 0.9038,
+      "step": 22616
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.050297791803325e-08,
+      "loss": 0.8037,
+      "step": 22617
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0426750297428696e-08,
+      "loss": 0.811,
+      "step": 22618
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0350617752591406e-08,
+      "loss": 0.7856,
+      "step": 22619
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0274580284977985e-08,
+      "loss": 0.8262,
+      "step": 22620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.019863789604283e-08,
+      "loss": 0.832,
+      "step": 22621
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.012279058723977e-08,
+      "loss": 0.8105,
+      "step": 22622
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0047038360018764e-08,
+      "loss": 0.8706,
+      "step": 22623
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.997138121582976e-08,
+      "loss": 0.8179,
+      "step": 22624
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.989581915611994e-08,
+      "loss": 0.7822,
+      "step": 22625
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.982035218233537e-08,
+      "loss": 0.4023,
+      "step": 22626
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9744980295918768e-08,
+      "loss": 0.8008,
+      "step": 22627
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9669703498313442e-08,
+      "loss": 0.8579,
+      "step": 22628
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9594521790958785e-08,
+      "loss": 0.7896,
+      "step": 22629
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.95194351752931e-08,
+      "loss": 0.8711,
+      "step": 22630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9444443652753564e-08,
+      "loss": 0.8354,
+      "step": 22631
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9369547224774032e-08,
+      "loss": 0.7959,
+      "step": 22632
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9294745892788357e-08,
+      "loss": 0.8535,
+      "step": 22633
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9220039658226506e-08,
+      "loss": 0.8179,
+      "step": 22634
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.914542852251845e-08,
+      "loss": 0.8652,
+      "step": 22635
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.907091248709193e-08,
+      "loss": 0.8467,
+      "step": 22636
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8996491553371364e-08,
+      "loss": 0.8462,
+      "step": 22637
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.892216572278117e-08,
+      "loss": 0.8506,
+      "step": 22638
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8847934996743543e-08,
+      "loss": 0.8535,
+      "step": 22639
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.877379937667901e-08,
+      "loss": 0.8633,
+      "step": 22640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.869975886400478e-08,
+      "loss": 0.7974,
+      "step": 22641
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.862581346013804e-08,
+      "loss": 0.8301,
+      "step": 22642
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.855196316649378e-08,
+      "loss": 0.8828,
+      "step": 22643
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8478207984484195e-08,
+      "loss": 0.8306,
+      "step": 22644
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8404547915520385e-08,
+      "loss": 0.8145,
+      "step": 22645
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8330982961012333e-08,
+      "loss": 0.7754,
+      "step": 22646
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8257513122366685e-08,
+      "loss": 0.8345,
+      "step": 22647
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.818413840098899e-08,
+      "loss": 0.8955,
+      "step": 22648
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8110858798284236e-08,
+      "loss": 0.8296,
+      "step": 22649
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8037674315652963e-08,
+      "loss": 0.8115,
+      "step": 22650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7964584954495722e-08,
+      "loss": 0.8359,
+      "step": 22651
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7891590716210837e-08,
+      "loss": 0.8652,
+      "step": 22652
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.781869160219497e-08,
+      "loss": 0.8179,
+      "step": 22653
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.774588761384256e-08,
+      "loss": 0.8252,
+      "step": 22654
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7673178752546938e-08,
+      "loss": 0.8271,
+      "step": 22655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.760056501969921e-08,
+      "loss": 0.8823,
+      "step": 22656
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7528046416687715e-08,
+      "loss": 0.8511,
+      "step": 22657
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.745562294490023e-08,
+      "loss": 0.7981,
+      "step": 22658
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.738329460572231e-08,
+      "loss": 0.8193,
+      "step": 22659
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7311061400538407e-08,
+      "loss": 0.835,
+      "step": 22660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7238923330729085e-08,
+      "loss": 0.8232,
+      "step": 22661
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.716688039767601e-08,
+      "loss": 0.8364,
+      "step": 22662
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7094932602756418e-08,
+      "loss": 0.8345,
+      "step": 22663
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7023079947346985e-08,
+      "loss": 0.8687,
+      "step": 22664
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6951322432822723e-08,
+      "loss": 0.8398,
+      "step": 22665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6879660060555868e-08,
+      "loss": 0.7544,
+      "step": 22666
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6808092831918098e-08,
+      "loss": 0.8237,
+      "step": 22667
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.673662074827832e-08,
+      "loss": 0.8433,
+      "step": 22668
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.666524381100377e-08,
+      "loss": 0.8452,
+      "step": 22669
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.659396202146003e-08,
+      "loss": 0.7469,
+      "step": 22670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6522775381011e-08,
+      "loss": 0.792,
+      "step": 22671
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6451683891018375e-08,
+      "loss": 0.8389,
+      "step": 22672
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.638068755284273e-08,
+      "loss": 0.8462,
+      "step": 22673
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6309786367841316e-08,
+      "loss": 0.7915,
+      "step": 22674
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6238980337371934e-08,
+      "loss": 0.8154,
+      "step": 22675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6168269462788497e-08,
+      "loss": 0.8306,
+      "step": 22676
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6097653745443262e-08,
+      "loss": 0.832,
+      "step": 22677
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6027133186688482e-08,
+      "loss": 0.8608,
+      "step": 22678
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.595670778787196e-08,
+      "loss": 0.8384,
+      "step": 22679
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5886377550342067e-08,
+      "loss": 0.8682,
+      "step": 22680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5816142475443838e-08,
+      "loss": 0.854,
+      "step": 22681
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5746002564521756e-08,
+      "loss": 0.8872,
+      "step": 22682
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5675957818916407e-08,
+      "loss": 0.8657,
+      "step": 22683
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.560600823996895e-08,
+      "loss": 0.853,
+      "step": 22684
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.553615382901664e-08,
+      "loss": 0.8364,
+      "step": 22685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5466394587397302e-08,
+      "loss": 0.8242,
+      "step": 22686
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5396730516444312e-08,
+      "loss": 0.8569,
+      "step": 22687
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5327161617491046e-08,
+      "loss": 0.8682,
+      "step": 22688
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.525768789186811e-08,
+      "loss": 0.8633,
+      "step": 22689
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5188309340904437e-08,
+      "loss": 0.7896,
+      "step": 22690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.511902596592841e-08,
+      "loss": 0.8516,
+      "step": 22691
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5049837768264528e-08,
+      "loss": 0.8911,
+      "step": 22692
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.498074474923673e-08,
+      "loss": 0.8345,
+      "step": 22693
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.491174691016729e-08,
+      "loss": 0.7646,
+      "step": 22694
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.484284425237571e-08,
+      "loss": 0.8184,
+      "step": 22695
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4774036777180376e-08,
+      "loss": 0.3063,
+      "step": 22696
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4705324485897465e-08,
+      "loss": 0.8252,
+      "step": 22697
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.463670737984203e-08,
+      "loss": 0.8442,
+      "step": 22698
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4568185460326353e-08,
+      "loss": 0.2948,
+      "step": 22699
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.449975872866217e-08,
+      "loss": 0.7822,
+      "step": 22700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4431427186157873e-08,
+      "loss": 0.8462,
+      "step": 22701
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4363190834120753e-08,
+      "loss": 0.7896,
+      "step": 22702
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4295049673855876e-08,
+      "loss": 0.8018,
+      "step": 22703
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.422700370666775e-08,
+      "loss": 0.8115,
+      "step": 22704
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4159052933858117e-08,
+      "loss": 0.8403,
+      "step": 22705
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4091197356726493e-08,
+      "loss": 0.9478,
+      "step": 22706
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4023436976571836e-08,
+      "loss": 0.8604,
+      "step": 22707
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3955771794689775e-08,
+      "loss": 0.729,
+      "step": 22708
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3888201812374834e-08,
+      "loss": 0.8242,
+      "step": 22709
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.382072703092042e-08,
+      "loss": 0.4016,
+      "step": 22710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.375334745161606e-08,
+      "loss": 0.8521,
+      "step": 22711
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3686063075752385e-08,
+      "loss": 0.8066,
+      "step": 22712
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3618873904616147e-08,
+      "loss": 0.8232,
+      "step": 22713
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3551779939492424e-08,
+      "loss": 0.8013,
+      "step": 22714
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3484781181665196e-08,
+      "loss": 0.7974,
+      "step": 22715
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3417877632416208e-08,
+      "loss": 0.8037,
+      "step": 22716
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3351069293024997e-08,
+      "loss": 0.8022,
+      "step": 22717
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3284356164769984e-08,
+      "loss": 0.8726,
+      "step": 22718
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3217738248927924e-08,
+      "loss": 0.8271,
+      "step": 22719
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.31512155467728e-08,
+      "loss": 0.8643,
+      "step": 22720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3084788059578034e-08,
+      "loss": 0.8237,
+      "step": 22721
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.301845578861317e-08,
+      "loss": 0.8569,
+      "step": 22722
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2952218735148302e-08,
+      "loss": 0.8467,
+      "step": 22723
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2886076900450195e-08,
+      "loss": 0.8696,
+      "step": 22724
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2820030285783945e-08,
+      "loss": 0.813,
+      "step": 22725
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.27540788924141e-08,
+      "loss": 0.8545,
+      "step": 22726
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2688222721602425e-08,
+      "loss": 0.8237,
+      "step": 22727
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2622461774607362e-08,
+      "loss": 0.7673,
+      "step": 22728
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2556796052688456e-08,
+      "loss": 0.8247,
+      "step": 22729
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.249122555710137e-08,
+      "loss": 0.7783,
+      "step": 22730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.242575028910121e-08,
+      "loss": 0.8208,
+      "step": 22731
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2360370249939202e-08,
+      "loss": 0.77,
+      "step": 22732
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.229508544086767e-08,
+      "loss": 0.8379,
+      "step": 22733
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2229895863135064e-08,
+      "loss": 0.813,
+      "step": 22734
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.216480151798872e-08,
+      "loss": 0.8525,
+      "step": 22735
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.20998024066732e-08,
+      "loss": 0.8237,
+      "step": 22736
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2034898530433058e-08,
+      "loss": 0.8315,
+      "step": 22737
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1970089890509527e-08,
+      "loss": 0.7993,
+      "step": 22738
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1905376488142726e-08,
+      "loss": 0.769,
+      "step": 22739
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1840758324570554e-08,
+      "loss": 0.8315,
+      "step": 22740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.177623540102869e-08,
+      "loss": 0.8228,
+      "step": 22741
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.171180771875281e-08,
+      "loss": 0.8081,
+      "step": 22742
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1647475278974706e-08,
+      "loss": 0.8428,
+      "step": 22743
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1583238082925063e-08,
+      "loss": 0.8193,
+      "step": 22744
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.151909613183345e-08,
+      "loss": 0.8931,
+      "step": 22745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1455049426926666e-08,
+      "loss": 0.7539,
+      "step": 22746
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.139109796942984e-08,
+      "loss": 0.8579,
+      "step": 22747
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.132724176056644e-08,
+      "loss": 0.8311,
+      "step": 22748
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1263480801558267e-08,
+      "loss": 0.8179,
+      "step": 22749
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1199815093626008e-08,
+      "loss": 0.8086,
+      "step": 22750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1136244637985914e-08,
+      "loss": 0.8311,
+      "step": 22751
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.10727694358559e-08,
+      "loss": 0.876,
+      "step": 22752
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1009389488448882e-08,
+      "loss": 0.8315,
+      "step": 22753
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0946104796978893e-08,
+      "loss": 0.8906,
+      "step": 22754
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.088291536265552e-08,
+      "loss": 0.8359,
+      "step": 22755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.081982118668835e-08,
+      "loss": 0.874,
+      "step": 22756
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.075682227028364e-08,
+      "loss": 0.8916,
+      "step": 22757
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0693918614647645e-08,
+      "loss": 0.8311,
+      "step": 22758
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0631110220983296e-08,
+      "loss": 0.8311,
+      "step": 22759
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0568397090492408e-08,
+      "loss": 0.7979,
+      "step": 22760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0505779224375134e-08,
+      "loss": 0.8335,
+      "step": 22761
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.044325662382829e-08,
+      "loss": 0.7358,
+      "step": 22762
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0380829290049255e-08,
+      "loss": 0.7856,
+      "step": 22763
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0318497224231515e-08,
+      "loss": 0.329,
+      "step": 22764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0256260427568007e-08,
+      "loss": 0.8335,
+      "step": 22765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0194118901249448e-08,
+      "loss": 0.7861,
+      "step": 22766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0132072646464328e-08,
+      "loss": 0.8154,
+      "step": 22767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.007012166440059e-08,
+      "loss": 0.8447,
+      "step": 22768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.000826595624228e-08,
+      "loss": 0.8359,
+      "step": 22769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9946505523173454e-08,
+      "loss": 0.8057,
+      "step": 22770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.988484036637539e-08,
+      "loss": 0.772,
+      "step": 22771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9823270487028256e-08,
+      "loss": 0.8311,
+      "step": 22772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9761795886309444e-08,
+      "loss": 0.8154,
+      "step": 22773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9700416565395787e-08,
+      "loss": 0.8477,
+      "step": 22774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.963913252546079e-08,
+      "loss": 0.832,
+      "step": 22775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9577943767677408e-08,
+      "loss": 0.8569,
+      "step": 22776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.951685029321637e-08,
+      "loss": 0.7603,
+      "step": 22777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.945585210324563e-08,
+      "loss": 0.8291,
+      "step": 22778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9394949198933145e-08,
+      "loss": 0.791,
+      "step": 22779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9334141581444087e-08,
+      "loss": 0.8657,
+      "step": 22780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.927342925194087e-08,
+      "loss": 0.8008,
+      "step": 22781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9212812211585886e-08,
+      "loss": 0.9141,
+      "step": 22782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9152290461538768e-08,
+      "loss": 0.8789,
+      "step": 22783
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.909186400295693e-08,
+      "loss": 0.7715,
+      "step": 22784
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9031532836996657e-08,
+      "loss": 0.8604,
+      "step": 22785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8971296964812035e-08,
+      "loss": 0.7852,
+      "step": 22786
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.891115638755603e-08,
+      "loss": 0.8008,
+      "step": 22787
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8851111106378272e-08,
+      "loss": 0.8481,
+      "step": 22788
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8791161122428403e-08,
+      "loss": 0.8291,
+      "step": 22789
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.873130643685328e-08,
+      "loss": 0.8872,
+      "step": 22790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.867154705079699e-08,
+      "loss": 0.7671,
+      "step": 22791
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.861188296540417e-08,
+      "loss": 0.8745,
+      "step": 22792
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8552314181816132e-08,
+      "loss": 0.8774,
+      "step": 22793
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.849284070117141e-08,
+      "loss": 0.8257,
+      "step": 22794
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.843346252460909e-08,
+      "loss": 0.8049,
+      "step": 22795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8374179653263825e-08,
+      "loss": 0.8218,
+      "step": 22796
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8314992088271368e-08,
+      "loss": 0.7891,
+      "step": 22797
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.825589983076248e-08,
+      "loss": 0.8076,
+      "step": 22798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.819690288186904e-08,
+      "loss": 0.8237,
+      "step": 22799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.813800124271903e-08,
+      "loss": 0.7439,
+      "step": 22800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.807919491443988e-08,
+      "loss": 0.7612,
+      "step": 22801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8020483898155694e-08,
+      "loss": 0.8403,
+      "step": 22802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.796186819499057e-08,
+      "loss": 0.8306,
+      "step": 22803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7903347806065285e-08,
+      "loss": 0.8555,
+      "step": 22804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7844922732500048e-08,
+      "loss": 0.8164,
+      "step": 22805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7786592975412852e-08,
+      "loss": 0.8301,
+      "step": 22806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7728358535918368e-08,
+      "loss": 0.8276,
+      "step": 22807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7670219415131806e-08,
+      "loss": 0.7847,
+      "step": 22808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.761217561416506e-08,
+      "loss": 0.7856,
+      "step": 22809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7554227134128355e-08,
+      "loss": 0.8457,
+      "step": 22810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7496373976130797e-08,
+      "loss": 0.7715,
+      "step": 22811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7438616141279287e-08,
+      "loss": 0.8398,
+      "step": 22812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7380953630678488e-08,
+      "loss": 0.8198,
+      "step": 22813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7323386445431965e-08,
+      "loss": 0.8931,
+      "step": 22814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7265914586639954e-08,
+      "loss": 0.8462,
+      "step": 22815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.720853805540379e-08,
+      "loss": 0.8784,
+      "step": 22816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7151256852819266e-08,
+      "loss": 0.8823,
+      "step": 22817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7094070979983834e-08,
+      "loss": 0.835,
+      "step": 22818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7036980437991068e-08,
+      "loss": 0.8423,
+      "step": 22819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6979985227932317e-08,
+      "loss": 0.8267,
+      "step": 22820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.692308535089948e-08,
+      "loss": 0.8408,
+      "step": 22821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6866280807979474e-08,
+      "loss": 0.875,
+      "step": 22822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6809571600260865e-08,
+      "loss": 0.8125,
+      "step": 22823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6752957728827234e-08,
+      "loss": 0.8379,
+      "step": 22824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.669643919476216e-08,
+      "loss": 0.8516,
+      "step": 22825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6640015999146995e-08,
+      "loss": 0.8101,
+      "step": 22826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.658368814306088e-08,
+      "loss": 0.7998,
+      "step": 22827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6527455627581844e-08,
+      "loss": 0.8228,
+      "step": 22828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.647131845378569e-08,
+      "loss": 0.8223,
+      "step": 22829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6415276622746558e-08,
+      "loss": 0.8721,
+      "step": 22830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.635933013553581e-08,
+      "loss": 0.8765,
+      "step": 22831
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.630347899322482e-08,
+      "loss": 0.7847,
+      "step": 22832
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.624772319688106e-08,
+      "loss": 0.8682,
+      "step": 22833
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6192062747572013e-08,
+      "loss": 0.7993,
+      "step": 22834
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.613649764636238e-08,
+      "loss": 0.8369,
+      "step": 22835
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.608102789431465e-08,
+      "loss": 0.8462,
+      "step": 22836
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.60256534924913e-08,
+      "loss": 0.7583,
+      "step": 22837
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5970374441950376e-08,
+      "loss": 0.8608,
+      "step": 22838
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5915190743750476e-08,
+      "loss": 0.8633,
+      "step": 22839
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5860102398946864e-08,
+      "loss": 0.8926,
+      "step": 22840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5805109408593146e-08,
+      "loss": 0.8608,
+      "step": 22841
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5750211773741807e-08,
+      "loss": 0.835,
+      "step": 22842
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5695409495443127e-08,
+      "loss": 0.8101,
+      "step": 22843
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5640702574746257e-08,
+      "loss": 0.8105,
+      "step": 22844
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5586091012696482e-08,
+      "loss": 0.7739,
+      "step": 22845
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5531574810339622e-08,
+      "loss": 0.8555,
+      "step": 22846
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5477153968717627e-08,
+      "loss": 0.7817,
+      "step": 22847
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5422828488873e-08,
+      "loss": 0.8457,
+      "step": 22848
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.536859837184379e-08,
+      "loss": 0.7871,
+      "step": 22849
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.531446361866862e-08,
+      "loss": 0.8193,
+      "step": 22850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5260424230382763e-08,
+      "loss": 0.8315,
+      "step": 22851
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5206480208019293e-08,
+      "loss": 0.8608,
+      "step": 22852
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5152631552611265e-08,
+      "loss": 0.8452,
+      "step": 22853
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.509887826518841e-08,
+      "loss": 0.8403,
+      "step": 22854
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5045220346779356e-08,
+      "loss": 0.8599,
+      "step": 22855
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4991657798410496e-08,
+      "loss": 0.8174,
+      "step": 22856
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4938190621106574e-08,
+      "loss": 0.8164,
+      "step": 22857
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.488481881589121e-08,
+      "loss": 0.8457,
+      "step": 22858
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4831542383784148e-08,
+      "loss": 0.7983,
+      "step": 22859
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4778361325805125e-08,
+      "loss": 0.8062,
+      "step": 22860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4725275642972214e-08,
+      "loss": 0.8687,
+      "step": 22861
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4672285336300718e-08,
+      "loss": 0.8008,
+      "step": 22862
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4619390406803713e-08,
+      "loss": 0.8608,
+      "step": 22863
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4566590855494278e-08,
+      "loss": 0.8486,
+      "step": 22864
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4513886683382162e-08,
+      "loss": 0.8706,
+      "step": 22865
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.446127789147489e-08,
+      "loss": 0.7847,
+      "step": 22866
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4408764480779991e-08,
+      "loss": 0.7339,
+      "step": 22867
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4356346452301662e-08,
+      "loss": 0.8257,
+      "step": 22868
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4304023807042988e-08,
+      "loss": 0.3148,
+      "step": 22869
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4251796546004836e-08,
+      "loss": 0.8345,
+      "step": 22870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4199664670185853e-08,
+      "loss": 0.9053,
+      "step": 22871
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4147628180584682e-08,
+      "loss": 0.8506,
+      "step": 22872
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.409568707819553e-08,
+      "loss": 0.8091,
+      "step": 22873
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4043841364013156e-08,
+      "loss": 0.8647,
+      "step": 22874
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3992091039028432e-08,
+      "loss": 0.8545,
+      "step": 22875
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3940436104232236e-08,
+      "loss": 0.8291,
+      "step": 22876
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3888876560612663e-08,
+      "loss": 0.8779,
+      "step": 22877
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3837412409156147e-08,
+      "loss": 0.8667,
+      "step": 22878
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3786043650846904e-08,
+      "loss": 0.8457,
+      "step": 22879
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.373477028666803e-08,
+      "loss": 0.79,
+      "step": 22880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3683592317600414e-08,
+      "loss": 0.3123,
+      "step": 22881
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3632509744623269e-08,
+      "loss": 0.8442,
+      "step": 22882
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3581522568713035e-08,
+      "loss": 0.8735,
+      "step": 22883
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.353063079084671e-08,
+      "loss": 0.7832,
+      "step": 22884
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3479834411996851e-08,
+      "loss": 0.8828,
+      "step": 22885
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3429133433135454e-08,
+      "loss": 0.8398,
+      "step": 22886
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.337852785523286e-08,
+      "loss": 0.8076,
+      "step": 22887
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3328017679256622e-08,
+      "loss": 0.8398,
+      "step": 22888
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3277602906173193e-08,
+      "loss": 0.8105,
+      "step": 22889
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.322728353694791e-08,
+      "loss": 0.7744,
+      "step": 22890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3177059572542784e-08,
+      "loss": 0.8677,
+      "step": 22891
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3126931013918709e-08,
+      "loss": 0.812,
+      "step": 22892
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3076897862034366e-08,
+      "loss": 0.8071,
+      "step": 22893
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3026960117847876e-08,
+      "loss": 0.8691,
+      "step": 22894
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.297711778231403e-08,
+      "loss": 0.8525,
+      "step": 22895
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2927370856385957e-08,
+      "loss": 0.8125,
+      "step": 22896
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.287771934101678e-08,
+      "loss": 0.853,
+      "step": 22897
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.282816323715519e-08,
+      "loss": 0.8394,
+      "step": 22898
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2778702545749311e-08,
+      "loss": 0.8364,
+      "step": 22899
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2729337267746168e-08,
+      "loss": 0.8701,
+      "step": 22900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2680067404089447e-08,
+      "loss": 0.8232,
+      "step": 22901
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2630892955722285e-08,
+      "loss": 0.8159,
+      "step": 22902
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.258181392358504e-08,
+      "loss": 0.832,
+      "step": 22903
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2532830308617516e-08,
+      "loss": 0.8506,
+      "step": 22904
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2483942111755632e-08,
+      "loss": 0.9019,
+      "step": 22905
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2435149333935303e-08,
+      "loss": 0.8711,
+      "step": 22906
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.238645197609023e-08,
+      "loss": 0.8364,
+      "step": 22907
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2337850039151889e-08,
+      "loss": 0.7622,
+      "step": 22908
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2289343524049536e-08,
+      "loss": 0.8774,
+      "step": 22909
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2240932431712427e-08,
+      "loss": 0.8013,
+      "step": 22910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2192616763065379e-08,
+      "loss": 0.772,
+      "step": 22911
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2144396519033763e-08,
+      "loss": 0.8472,
+      "step": 22912
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.209627170053962e-08,
+      "loss": 0.8652,
+      "step": 22913
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2048242308503878e-08,
+      "loss": 0.8145,
+      "step": 22914
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.200030834384469e-08,
+      "loss": 0.8169,
+      "step": 22915
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1952469807480771e-08,
+      "loss": 0.7827,
+      "step": 22916
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1904726700325276e-08,
+      "loss": 0.8428,
+      "step": 22917
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1857079023293027e-08,
+      "loss": 0.896,
+      "step": 22918
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1809526777295521e-08,
+      "loss": 0.813,
+      "step": 22919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1762069963241474e-08,
+      "loss": 0.8579,
+      "step": 22920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1714708582040158e-08,
+      "loss": 0.7866,
+      "step": 22921
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.166744263459696e-08,
+      "loss": 0.8628,
+      "step": 22922
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1620272121816156e-08,
+      "loss": 0.8433,
+      "step": 22923
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1573197044599804e-08,
+      "loss": 0.7217,
+      "step": 22924
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1526217403849404e-08,
+      "loss": 0.7871,
+      "step": 22925
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1479333200463127e-08,
+      "loss": 0.8574,
+      "step": 22926
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1432544435338588e-08,
+      "loss": 0.8022,
+      "step": 22927
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1385851109369517e-08,
+      "loss": 0.8428,
+      "step": 22928
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1339253223451308e-08,
+      "loss": 0.3669,
+      "step": 22929
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1292750778473805e-08,
+      "loss": 0.8599,
+      "step": 22930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.124634377532685e-08,
+      "loss": 0.8057,
+      "step": 22931
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1200032214899182e-08,
+      "loss": 0.8784,
+      "step": 22932
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1153816098076197e-08,
+      "loss": 0.7769,
+      "step": 22933
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1107695425742193e-08,
+      "loss": 0.7798,
+      "step": 22934
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.106167019877924e-08,
+      "loss": 0.8423,
+      "step": 22935
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1015740418068299e-08,
+      "loss": 0.8379,
+      "step": 22936
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0969906084488113e-08,
+      "loss": 0.8521,
+      "step": 22937
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0924167198914647e-08,
+      "loss": 0.7993,
+      "step": 22938
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0878523762224425e-08,
+      "loss": 0.834,
+      "step": 22939
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0832975775289523e-08,
+      "loss": 0.7983,
+      "step": 22940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0787523238981468e-08,
+      "loss": 0.8931,
+      "step": 22941
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0742166154170675e-08,
+      "loss": 0.8398,
+      "step": 22942
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0696904521724227e-08,
+      "loss": 0.7749,
+      "step": 22943
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.06517383425081e-08,
+      "loss": 0.8364,
+      "step": 22944
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.06066676173866e-08,
+      "loss": 0.8262,
+      "step": 22945
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0561692347221265e-08,
+      "loss": 0.8633,
+      "step": 22946
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0516812532873622e-08,
+      "loss": 0.7812,
+      "step": 22947
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0472028175201876e-08,
+      "loss": 0.8047,
+      "step": 22948
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0427339275062565e-08,
+      "loss": 0.8179,
+      "step": 22949
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0382745833311114e-08,
+      "loss": 0.8594,
+      "step": 22950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0338247850800177e-08,
+      "loss": 0.8115,
+      "step": 22951
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0293845328381291e-08,
+      "loss": 0.8296,
+      "step": 22952
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.024953826690378e-08,
+      "loss": 0.8682,
+      "step": 22953
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0205326667215298e-08,
+      "loss": 0.8477,
+      "step": 22954
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0161210530162392e-08,
+      "loss": 0.7783,
+      "step": 22955
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0117189856588273e-08,
+      "loss": 0.7856,
+      "step": 22956
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0073264647335601e-08,
+      "loss": 0.8359,
+      "step": 22957
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0029434903244262e-08,
+      "loss": 0.8047,
+      "step": 22958
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.985700625153027e-09,
+      "loss": 0.8716,
+      "step": 22959
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.942061813899007e-09,
+      "loss": 0.8345,
+      "step": 22960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.898518470316531e-09,
+      "loss": 0.8589,
+      "step": 22961
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.855070595238825e-09,
+      "loss": 0.8311,
+      "step": 22962
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.811718189496889e-09,
+      "loss": 0.8901,
+      "step": 22963
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.768461253920614e-09,
+      "loss": 0.8257,
+      "step": 22964
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.725299789337116e-09,
+      "loss": 0.8164,
+      "step": 22965
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.682233796572404e-09,
+      "loss": 0.36,
+      "step": 22966
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.639263276449707e-09,
+      "loss": 0.8462,
+      "step": 22967
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.596388229792253e-09,
+      "loss": 0.8208,
+      "step": 22968
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.5536086574205e-09,
+      "loss": 0.8311,
+      "step": 22969
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.510924560151013e-09,
+      "loss": 0.8545,
+      "step": 22970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.468335938802032e-09,
+      "loss": 0.8237,
+      "step": 22971
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.425842794188456e-09,
+      "loss": 0.8408,
+      "step": 22972
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.38344512712186e-09,
+      "loss": 0.9033,
+      "step": 22973
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.341142938414372e-09,
+      "loss": 0.8481,
+      "step": 22974
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.298936228875344e-09,
+      "loss": 0.8101,
+      "step": 22975
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.256824999311908e-09,
+      "loss": 0.8633,
+      "step": 22976
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.214809250530088e-09,
+      "loss": 0.8022,
+      "step": 22977
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.172888983333128e-09,
+      "loss": 0.8198,
+      "step": 22978
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.131064198523165e-09,
+      "loss": 0.8081,
+      "step": 22979
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.089334896901225e-09,
+      "loss": 0.8623,
+      "step": 22980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.047701079264448e-09,
+      "loss": 0.8218,
+      "step": 22981
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.00616274641053e-09,
+      "loss": 0.8672,
+      "step": 22982
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.964719899132724e-09,
+      "loss": 0.8057,
+      "step": 22983
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.923372538225394e-09,
+      "loss": 0.3394,
+      "step": 22984
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.882120664478466e-09,
+      "loss": 0.8281,
+      "step": 22985
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.840964278681863e-09,
+      "loss": 0.8101,
+      "step": 22986
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.799903381622732e-09,
+      "loss": 0.8823,
+      "step": 22987
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.758937974086557e-09,
+      "loss": 0.873,
+      "step": 22988
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.718068056857154e-09,
+      "loss": 0.8872,
+      "step": 22989
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.677293630716121e-09,
+      "loss": 0.8574,
+      "step": 22990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.636614696443945e-09,
+      "loss": 0.7808,
+      "step": 22991
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.596031254818893e-09,
+      "loss": 0.7959,
+      "step": 22992
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.555543306617009e-09,
+      "loss": 0.8335,
+      "step": 22993
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.51515085261323e-09,
+      "loss": 0.8413,
+      "step": 22994
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.47485389358027e-09,
+      "loss": 0.8916,
+      "step": 22995
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.434652430289181e-09,
+      "loss": 0.8032,
+      "step": 22996
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.394546463508235e-09,
+      "loss": 0.8154,
+      "step": 22997
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.354535994006263e-09,
+      "loss": 0.8208,
+      "step": 22998
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.314621022547098e-09,
+      "loss": 0.7944,
+      "step": 22999
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.274801549895684e-09,
+      "loss": 0.814,
+      "step": 23000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.235077576813077e-09,
+      "loss": 0.8101,
+      "step": 23001
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.195449104059783e-09,
+      "loss": 0.8188,
+      "step": 23002
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.155916132393526e-09,
+      "loss": 0.8286,
+      "step": 23003
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.11647866257037e-09,
+      "loss": 0.7324,
+      "step": 23004
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.077136695345823e-09,
+      "loss": 0.812,
+      "step": 23005
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.037890231471501e-09,
+      "loss": 0.834,
+      "step": 23006
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.998739271698474e-09,
+      "loss": 0.8325,
+      "step": 23007
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.959683816776698e-09,
+      "loss": 0.7896,
+      "step": 23008
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.92072386745224e-09,
+      "loss": 0.7915,
+      "step": 23009
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.881859424470617e-09,
+      "loss": 0.877,
+      "step": 23010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.843090488576232e-09,
+      "loss": 0.8711,
+      "step": 23011
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.804417060509606e-09,
+      "loss": 0.8569,
+      "step": 23012
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.76583914101181e-09,
+      "loss": 0.8521,
+      "step": 23013
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.727356730820035e-09,
+      "loss": 0.8647,
+      "step": 23014
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.688969830670357e-09,
+      "loss": 0.8486,
+      "step": 23015
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.650678441298298e-09,
+      "loss": 0.8477,
+      "step": 23016
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.612482563434942e-09,
+      "loss": 0.7949,
+      "step": 23017
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.57438219781248e-09,
+      "loss": 0.835,
+      "step": 23018
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.536377345158663e-09,
+      "loss": 0.8486,
+      "step": 23019
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.498468006201243e-09,
+      "loss": 0.8125,
+      "step": 23020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.460654181665749e-09,
+      "loss": 0.8101,
+      "step": 23021
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.422935872274939e-09,
+      "loss": 0.3608,
+      "step": 23022
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.385313078750456e-09,
+      "loss": 0.772,
+      "step": 23023
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.347785801812834e-09,
+      "loss": 0.7593,
+      "step": 23024
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.3103540421787246e-09,
+      "loss": 0.8633,
+      "step": 23025
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.27301780056533e-09,
+      "loss": 0.8154,
+      "step": 23026
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.235777077687079e-09,
+      "loss": 0.8164,
+      "step": 23027
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.198631874255624e-09,
+      "loss": 0.8374,
+      "step": 23028
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.161582190982619e-09,
+      "loss": 0.8394,
+      "step": 23029
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.124628028575831e-09,
+      "loss": 0.7554,
+      "step": 23030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.087769387742472e-09,
+      "loss": 0.8447,
+      "step": 23031
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.051006269188087e-09,
+      "loss": 0.8135,
+      "step": 23032
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.014338673616006e-09,
+      "loss": 0.8218,
+      "step": 23033
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.9777666017273315e-09,
+      "loss": 0.8833,
+      "step": 23034
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.941290054222616e-09,
+      "loss": 0.7708,
+      "step": 23035
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.904909031798524e-09,
+      "loss": 0.7725,
+      "step": 23036
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.868623535151719e-09,
+      "loss": 0.7998,
+      "step": 23037
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.832433564976648e-09,
+      "loss": 0.8384,
+      "step": 23038
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.796339121964979e-09,
+      "loss": 0.8755,
+      "step": 23039
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.760340206807825e-09,
+      "loss": 0.7676,
+      "step": 23040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.72443682019408e-09,
+      "loss": 0.9048,
+      "step": 23041
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.6886289628098624e-09,
+      "loss": 0.8032,
+      "step": 23042
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.652916635340734e-09,
+      "loss": 0.7876,
+      "step": 23043
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.617299838470037e-09,
+      "loss": 0.8057,
+      "step": 23044
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.581778572878339e-09,
+      "loss": 0.3252,
+      "step": 23045
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.5463528392467616e-09,
+      "loss": 0.8032,
+      "step": 23046
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.51102263825143e-09,
+      "loss": 0.8208,
+      "step": 23047
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.47578797056958e-09,
+      "loss": 0.8208,
+      "step": 23048
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.440648836874564e-09,
+      "loss": 0.8813,
+      "step": 23049
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.405605237838619e-09,
+      "loss": 0.7705,
+      "step": 23050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.370657174132322e-09,
+      "loss": 0.7739,
+      "step": 23051
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.3358046464245815e-09,
+      "loss": 0.8774,
+      "step": 23052
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.301047655381531e-09,
+      "loss": 0.8599,
+      "step": 23053
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.2663862016693054e-09,
+      "loss": 0.8394,
+      "step": 23054
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.231820285949597e-09,
+      "loss": 0.8262,
+      "step": 23055
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.197349908884653e-09,
+      "loss": 0.873,
+      "step": 23056
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.1629750711333924e-09,
+      "loss": 0.8921,
+      "step": 23057
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.128695773353621e-09,
+      "loss": 0.8257,
+      "step": 23058
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.094512016201481e-09,
+      "loss": 0.8247,
+      "step": 23059
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.060423800330895e-09,
+      "loss": 0.7773,
+      "step": 23060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.026431126394116e-09,
+      "loss": 0.9014,
+      "step": 23061
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.992533995040628e-09,
+      "loss": 0.9106,
+      "step": 23062
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.958732406919909e-09,
+      "loss": 0.8462,
+      "step": 23063
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.92502636267811e-09,
+      "loss": 0.8496,
+      "step": 23064
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.891415862960271e-09,
+      "loss": 0.8154,
+      "step": 23065
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.857900908409208e-09,
+      "loss": 0.8154,
+      "step": 23066
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.824481499666634e-09,
+      "loss": 0.8257,
+      "step": 23067
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.791157637371481e-09,
+      "loss": 0.8281,
+      "step": 23068
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.757929322161016e-09,
+      "loss": 0.8091,
+      "step": 23069
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.724796554671952e-09,
+      "loss": 0.8066,
+      "step": 23070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.691759335537117e-09,
+      "loss": 0.8066,
+      "step": 23071
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.658817665388783e-09,
+      "loss": 0.7861,
+      "step": 23072
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.625971544857556e-09,
+      "loss": 0.7905,
+      "step": 23073
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.593220974571822e-09,
+      "loss": 0.8003,
+      "step": 23074
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.560565955157748e-09,
+      "loss": 0.8452,
+      "step": 23075
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.5280064872403895e-09,
+      "loss": 0.8687,
+      "step": 23076
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.495542571443135e-09,
+      "loss": 0.8379,
+      "step": 23077
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.463174208385491e-09,
+      "loss": 0.8521,
+      "step": 23078
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.430901398688626e-09,
+      "loss": 0.8643,
+      "step": 23079
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.398724142968714e-09,
+      "loss": 0.873,
+      "step": 23080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.366642441841374e-09,
+      "loss": 0.8203,
+      "step": 23081
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3346562959211144e-09,
+      "loss": 0.8418,
+      "step": 23082
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.302765705819668e-09,
+      "loss": 0.8896,
+      "step": 23083
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.270970672146547e-09,
+      "loss": 0.7881,
+      "step": 23084
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.239271195510709e-09,
+      "loss": 0.8105,
+      "step": 23085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.207667276518336e-09,
+      "loss": 0.8457,
+      "step": 23086
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1761589157739435e-09,
+      "loss": 0.8203,
+      "step": 23087
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1447461138809386e-09,
+      "loss": 0.8525,
+      "step": 23088
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.113428871439397e-09,
+      "loss": 0.8174,
+      "step": 23089
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.082207189049393e-09,
+      "loss": 0.835,
+      "step": 23090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0510810673071176e-09,
+      "loss": 0.8149,
+      "step": 23091
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.020050506809315e-09,
+      "loss": 0.8711,
+      "step": 23092
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.989115508148845e-09,
+      "loss": 0.8418,
+      "step": 23093
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.958276071917456e-09,
+      "loss": 0.8174,
+      "step": 23094
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.927532198705787e-09,
+      "loss": 0.8291,
+      "step": 23095
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8968838891017e-09,
+      "loss": 0.8457,
+      "step": 23096
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.866331143691394e-09,
+      "loss": 0.8906,
+      "step": 23097
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.835873963059401e-09,
+      "loss": 0.7659,
+      "step": 23098
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.805512347789143e-09,
+      "loss": 0.835,
+      "step": 23099
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.775246298460157e-09,
+      "loss": 0.853,
+      "step": 23100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.745075815652533e-09,
+      "loss": 0.8384,
+      "step": 23101
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.715000899943034e-09,
+      "loss": 0.8403,
+      "step": 23102
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6850215519073094e-09,
+      "loss": 0.8506,
+      "step": 23103
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.65513777211879e-09,
+      "loss": 0.8247,
+      "step": 23104
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.62534956114924e-09,
+      "loss": 0.8291,
+      "step": 23105
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.595656919568203e-09,
+      "loss": 0.7808,
+      "step": 23106
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.566059847944116e-09,
+      "loss": 0.8643,
+      "step": 23107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5365583468431895e-09,
+      "loss": 0.918,
+      "step": 23108
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.507152416829974e-09,
+      "loss": 0.8213,
+      "step": 23109
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.477842058466242e-09,
+      "loss": 0.3444,
+      "step": 23110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4486272723137654e-09,
+      "loss": 0.8003,
+      "step": 23111
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.419508058930988e-09,
+      "loss": 0.8491,
+      "step": 23112
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3904844188752405e-09,
+      "loss": 0.8091,
+      "step": 23113
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.361556352701634e-09,
+      "loss": 0.8433,
+      "step": 23114
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.332723860963617e-09,
+      "loss": 0.8071,
+      "step": 23115
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3039869442124125e-09,
+      "loss": 0.8447,
+      "step": 23116
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.275345602998693e-09,
+      "loss": 0.8169,
+      "step": 23117
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.246799837869242e-09,
+      "loss": 0.8291,
+      "step": 23118
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2183496493714005e-09,
+      "loss": 0.814,
+      "step": 23119
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.189995038048622e-09,
+      "loss": 0.8457,
+      "step": 23120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.161736004443806e-09,
+      "loss": 0.8267,
+      "step": 23121
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1335725490976305e-09,
+      "loss": 0.8398,
+      "step": 23122
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1055046725479994e-09,
+      "loss": 0.8408,
+      "step": 23123
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.077532375333371e-09,
+      "loss": 0.8457,
+      "step": 23124
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.049655657987761e-09,
+      "loss": 0.8599,
+      "step": 23125
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.021874521045188e-09,
+      "loss": 0.8774,
+      "step": 23126
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9941889650363385e-09,
+      "loss": 0.7627,
+      "step": 23127
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.966598990491899e-09,
+      "loss": 0.8691,
+      "step": 23128
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.93910459793867e-09,
+      "loss": 0.8062,
+      "step": 23129
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.911705787903452e-09,
+      "loss": 0.7725,
+      "step": 23130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.884402560910272e-09,
+      "loss": 0.8047,
+      "step": 23131
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.857194917481488e-09,
+      "loss": 0.834,
+      "step": 23132
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8300828581372405e-09,
+      "loss": 0.8462,
+      "step": 23133
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8030663833965585e-09,
+      "loss": 0.8208,
+      "step": 23134
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.776145493776251e-09,
+      "loss": 0.8672,
+      "step": 23135
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.749320189791461e-09,
+      "loss": 0.8071,
+      "step": 23136
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.722590471955112e-09,
+      "loss": 0.8174,
+      "step": 23137
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6959563407784617e-09,
+      "loss": 0.8198,
+      "step": 23138
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6694177967716572e-09,
+      "loss": 0.771,
+      "step": 23139
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6429748404426255e-09,
+      "loss": 0.813,
+      "step": 23140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6166274722959637e-09,
+      "loss": 0.8032,
+      "step": 23141
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5903756928368227e-09,
+      "loss": 0.7729,
+      "step": 23142
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5642195025675787e-09,
+      "loss": 0.8452,
+      "step": 23143
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5381589019878314e-09,
+      "loss": 0.7769,
+      "step": 23144
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.512193891596627e-09,
+      "loss": 0.791,
+      "step": 23145
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4863244718907896e-09,
+      "loss": 0.8652,
+      "step": 23146
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.460550643364369e-09,
+      "loss": 0.8521,
+      "step": 23147
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4348724065119687e-09,
+      "loss": 0.832,
+      "step": 23148
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.409289761823753e-09,
+      "loss": 0.877,
+      "step": 23149
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.38380270978933e-09,
+      "loss": 0.8618,
+      "step": 23150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3584112508966425e-09,
+      "loss": 0.8384,
+      "step": 23151
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3331153856314136e-09,
+      "loss": 0.8159,
+      "step": 23152
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3079151144777e-09,
+      "loss": 0.8208,
+      "step": 23153
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2828104379167837e-09,
+      "loss": 0.7861,
+      "step": 23154
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.257801356429946e-09,
+      "loss": 0.8311,
+      "step": 23155
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.232887870495138e-09,
+      "loss": 0.7915,
+      "step": 23156
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.208069980588646e-09,
+      "loss": 0.8496,
+      "step": 23157
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1833476871862e-09,
+      "loss": 0.8677,
+      "step": 23158
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1587209907602e-09,
+      "loss": 0.8457,
+      "step": 23159
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1341898917819357e-09,
+      "loss": 0.8442,
+      "step": 23160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1097543907210313e-09,
+      "loss": 0.8223,
+      "step": 23161
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.085414488044336e-09,
+      "loss": 0.835,
+      "step": 23162
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0611701842175876e-09,
+      "loss": 0.8286,
+      "step": 23163
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0370214797054152e-09,
+      "loss": 0.8794,
+      "step": 23164
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0129683749691165e-09,
+      "loss": 0.8242,
+      "step": 23165
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.989010870468323e-09,
+      "loss": 0.8218,
+      "step": 23166
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9651489666626674e-09,
+      "loss": 0.7529,
+      "step": 23167
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9413826640078967e-09,
+      "loss": 0.8325,
+      "step": 23168
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.917711962959202e-09,
+      "loss": 0.8511,
+      "step": 23169
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8941368639684442e-09,
+      "loss": 0.7632,
+      "step": 23170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8706573674880388e-09,
+      "loss": 0.8018,
+      "step": 23171
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8472734739659615e-09,
+      "loss": 0.8091,
+      "step": 23172
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8239851838501865e-09,
+      "loss": 0.8159,
+      "step": 23173
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8007924975864687e-09,
+      "loss": 0.8628,
+      "step": 23174
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7776954156177873e-09,
+      "loss": 0.8286,
+      "step": 23175
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7546939383865655e-09,
+      "loss": 0.8398,
+      "step": 23176
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7317880663324524e-09,
+      "loss": 0.8828,
+      "step": 23177
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.708977799894541e-09,
+      "loss": 0.8823,
+      "step": 23178
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6862631395080386e-09,
+      "loss": 0.8096,
+      "step": 23179
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6636440856087078e-09,
+      "loss": 0.8931,
+      "step": 23180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.641120638628425e-09,
+      "loss": 0.8408,
+      "step": 23181
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.618692798997957e-09,
+      "loss": 0.7469,
+      "step": 23182
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5963605671469604e-09,
+      "loss": 0.8203,
+      "step": 23183
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.574123943502871e-09,
+      "loss": 0.7915,
+      "step": 23184
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.551982928490904e-09,
+      "loss": 0.793,
+      "step": 23185
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.529937522534054e-09,
+      "loss": 0.8882,
+      "step": 23186
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5079877260542064e-09,
+      "loss": 0.8115,
+      "step": 23187
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.486133539472135e-09,
+      "loss": 0.8164,
+      "step": 23188
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4643749632052848e-09,
+      "loss": 0.8828,
+      "step": 23189
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.442711997670544e-09,
+      "loss": 0.7881,
+      "step": 23190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.42114464328147e-09,
+      "loss": 0.8447,
+      "step": 23191
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3996729004510665e-09,
+      "loss": 0.7861,
+      "step": 23192
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.378296769590116e-09,
+      "loss": 0.8232,
+      "step": 23193
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3570162511077353e-09,
+      "loss": 0.8252,
+      "step": 23194
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3358313454108216e-09,
+      "loss": 0.853,
+      "step": 23195
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3147420529051613e-09,
+      "loss": 0.7534,
+      "step": 23196
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2937483739937647e-09,
+      "loss": 0.8892,
+      "step": 23197
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2728503090779786e-09,
+      "loss": 0.7837,
+      "step": 23198
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2520478585585926e-09,
+      "loss": 0.834,
+      "step": 23199
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2313410228325115e-09,
+      "loss": 0.8579,
+      "step": 23200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2107298022966407e-09,
+      "loss": 0.8062,
+      "step": 23201
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1902141973445535e-09,
+      "loss": 0.8174,
+      "step": 23202
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.169794208369824e-09,
+      "loss": 0.8389,
+      "step": 23203
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1494698357621413e-09,
+      "loss": 0.8726,
+      "step": 23204
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.129241079911193e-09,
+      "loss": 0.8535,
+      "step": 23205
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1091079412027817e-09,
+      "loss": 0.8223,
+      "step": 23206
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0890704200232648e-09,
+      "loss": 0.7896,
+      "step": 23207
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0691285167556697e-09,
+      "loss": 0.8486,
+      "step": 23208
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.049282231781358e-09,
+      "loss": 0.7749,
+      "step": 23209
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0295315654800253e-09,
+      "loss": 0.8281,
+      "step": 23210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.009876518229703e-09,
+      "loss": 0.7896,
+      "step": 23211
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.990317090406202e-09,
+      "loss": 0.8496,
+      "step": 23212
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9708532823836667e-09,
+      "loss": 0.7734,
+      "step": 23213
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9514850945345777e-09,
+      "loss": 0.8403,
+      "step": 23214
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9322125272297488e-09,
+      "loss": 0.834,
+      "step": 23215
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9130355808377743e-09,
+      "loss": 0.8252,
+      "step": 23216
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.893954255725583e-09,
+      "loss": 0.7866,
+      "step": 23217
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.874968552257328e-09,
+      "loss": 0.8145,
+      "step": 23218
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8560784707977175e-09,
+      "loss": 0.835,
+      "step": 23219
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8372840117070189e-09,
+      "loss": 0.8501,
+      "step": 23220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8185851753454996e-09,
+      "loss": 0.835,
+      "step": 23221
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7999819620700965e-09,
+      "loss": 0.7727,
+      "step": 23222
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.781474372237746e-09,
+      "loss": 0.8213,
+      "step": 23223
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.763062406202054e-09,
+      "loss": 0.374,
+      "step": 23224
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7447460643149616e-09,
+      "loss": 0.8447,
+      "step": 23225
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.726525346926744e-09,
+      "loss": 0.7944,
+      "step": 23226
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7084002543871214e-09,
+      "loss": 0.8159,
+      "step": 23227
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6903707870419283e-09,
+      "loss": 0.8506,
+      "step": 23228
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.672436945235889e-09,
+      "loss": 0.7744,
+      "step": 23229
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6545987293131727e-09,
+      "loss": 0.8618,
+      "step": 23230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6368561396140626e-09,
+      "loss": 0.9014,
+      "step": 23231
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.619209176478287e-09,
+      "loss": 0.8521,
+      "step": 23232
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6016578402433536e-09,
+      "loss": 0.8081,
+      "step": 23233
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.584202131245105e-09,
+      "loss": 0.9004,
+      "step": 23234
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5668420498182736e-09,
+      "loss": 0.8379,
+      "step": 23235
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5495775962937054e-09,
+      "loss": 0.8457,
+      "step": 23236
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5324087710028024e-09,
+      "loss": 0.7944,
+      "step": 23237
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5153355742730802e-09,
+      "loss": 0.8218,
+      "step": 23238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4983580064314996e-09,
+      "loss": 0.8945,
+      "step": 23239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4814760678033557e-09,
+      "loss": 0.8584,
+      "step": 23240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4646897587111686e-09,
+      "loss": 0.7998,
+      "step": 23241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4479990794763477e-09,
+      "loss": 0.8833,
+      "step": 23242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4314040304175269e-09,
+      "loss": 0.8545,
+      "step": 23243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4149046118533406e-09,
+      "loss": 0.8276,
+      "step": 23244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3985008240985364e-09,
+      "loss": 0.8794,
+      "step": 23245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.382192667467308e-09,
+      "loss": 0.8013,
+      "step": 23246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3659801422710728e-09,
+      "loss": 0.8516,
+      "step": 23247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3498632488212481e-09,
+      "loss": 0.8574,
+      "step": 23248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.333841987424811e-09,
+      "loss": 0.8608,
+      "step": 23249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3179163583892929e-09,
+      "loss": 0.8716,
+      "step": 23250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.302086362018895e-09,
+      "loss": 0.8052,
+      "step": 23251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2863519986167084e-09,
+      "loss": 0.8379,
+      "step": 23252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2707132684836031e-09,
+      "loss": 0.834,
+      "step": 23253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2551701719182296e-09,
+      "loss": 0.8555,
+      "step": 23254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2397227092192377e-09,
+      "loss": 0.8081,
+      "step": 23255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2243708806808364e-09,
+      "loss": 0.77,
+      "step": 23256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2091146865972347e-09,
+      "loss": 0.8032,
+      "step": 23257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1939541272609767e-09,
+      "loss": 0.8726,
+      "step": 23258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1788892029612754e-09,
+      "loss": 0.813,
+      "step": 23259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1639199139862335e-09,
+      "loss": 0.8428,
+      "step": 23260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1490462606228436e-09,
+      "loss": 0.7656,
+      "step": 23261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1342682431553231e-09,
+      "loss": 0.8433,
+      "step": 23262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1195858618667787e-09,
+      "loss": 0.8184,
+      "step": 23263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1049991170375418e-09,
+      "loss": 0.8496,
+      "step": 23264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0905080089473885e-09,
+      "loss": 0.8765,
+      "step": 23265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0761125378727643e-09,
+      "loss": 0.8271,
+      "step": 23266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0618127040890047e-09,
+      "loss": 0.8203,
+      "step": 23267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0476085078708897e-09,
+      "loss": 0.8887,
+      "step": 23268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0334999494887587e-09,
+      "loss": 0.8164,
+      "step": 23269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.019487029213506e-09,
+      "loss": 0.8374,
+      "step": 23270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0055697473126957e-09,
+      "loss": 0.8779,
+      "step": 23271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.917481040533362e-10,
+      "loss": 0.8237,
+      "step": 23272
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.7802209969855e-10,
+      "loss": 0.8555,
+      "step": 23273
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.643917345120157e-10,
+      "loss": 0.8403,
+      "step": 23274
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.508570087540802e-10,
+      "loss": 0.8115,
+      "step": 23275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.374179226839807e-10,
+      "loss": 0.7778,
+      "step": 23276
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.24074476558734e-10,
+      "loss": 0.8101,
+      "step": 23277
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.108266706331359e-10,
+      "loss": 0.8516,
+      "step": 23278
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.976745051608726e-10,
+      "loss": 0.8018,
+      "step": 23279
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.846179803939647e-10,
+      "loss": 0.8374,
+      "step": 23280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.716570965822124e-10,
+      "loss": 0.8594,
+      "step": 23281
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.587918539726403e-10,
+      "loss": 0.8579,
+      "step": 23282
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.460222528128281e-10,
+      "loss": 0.8516,
+      "step": 23283
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.333482933459147e-10,
+      "loss": 0.3862,
+      "step": 23284
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.207699758144838e-10,
+      "loss": 0.8066,
+      "step": 23285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.08287300460009e-10,
+      "loss": 0.8018,
+      "step": 23286
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.959002675200778e-10,
+      "loss": 0.7954,
+      "step": 23287
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.836088772328332e-10,
+      "loss": 0.7419,
+      "step": 23288
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.714131298325322e-10,
+      "loss": 0.8115,
+      "step": 23289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.59313025553432e-10,
+      "loss": 0.8911,
+      "step": 23290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.473085646264589e-10,
+      "loss": 0.8398,
+      "step": 23291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.353997472808738e-10,
+      "loss": 0.8032,
+      "step": 23292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.235865737453829e-10,
+      "loss": 0.7759,
+      "step": 23293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.118690442453613e-10,
+      "loss": 0.7983,
+      "step": 23294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.002471590050741e-10,
+      "loss": 0.792,
+      "step": 23295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.887209182471211e-10,
+      "loss": 0.8037,
+      "step": 23296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.772903221918814e-10,
+      "loss": 0.3615,
+      "step": 23297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.659553710575139e-10,
+      "loss": 0.8237,
+      "step": 23298
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.547160650621775e-10,
+      "loss": 0.8149,
+      "step": 23299
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.435724044195902e-10,
+      "loss": 0.8276,
+      "step": 23300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.325243893440247e-10,
+      "loss": 0.8872,
+      "step": 23301
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.215720200458686e-10,
+      "loss": 0.8364,
+      "step": 23302
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.107152967349539e-10,
+      "loss": 0.7959,
+      "step": 23303
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.999542196194475e-10,
+      "loss": 0.832,
+      "step": 23304
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.892887889047405e-10,
+      "loss": 0.8037,
+      "step": 23305
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.787190047951141e-10,
+      "loss": 0.7773,
+      "step": 23306
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.682448674926289e-10,
+      "loss": 0.8418,
+      "step": 23307
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.5786637719768e-10,
+      "loss": 0.8662,
+      "step": 23308
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.475835341089975e-10,
+      "loss": 0.8047,
+      "step": 23309
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.373963384230906e-10,
+      "loss": 0.7915,
+      "step": 23310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.273047903348039e-10,
+      "loss": 0.7695,
+      "step": 23311
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.173088900378709e-10,
+      "loss": 0.8472,
+      "step": 23312
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.07408637722695e-10,
+      "loss": 0.8213,
+      "step": 23313
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.976040335791244e-10,
+      "loss": 0.877,
+      "step": 23314
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.878950777942315e-10,
+      "loss": 0.8276,
+      "step": 23315
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.78281770554534e-10,
+      "loss": 0.8711,
+      "step": 23316
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.687641120437736e-10,
+      "loss": 0.7632,
+      "step": 23317
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5934210244347187e-10,
+      "loss": 0.8037,
+      "step": 23318
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.500157419345952e-10,
+      "loss": 0.8452,
+      "step": 23319
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4078503069477916e-10,
+      "loss": 0.811,
+      "step": 23320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3164996890110444e-10,
+      "loss": 0.7949,
+      "step": 23321
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.226105567289862e-10,
+      "loss": 0.8472,
+      "step": 23322
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1366679434995393e-10,
+      "loss": 0.8745,
+      "step": 23323
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.048186819360922e-10,
+      "loss": 0.8535,
+      "step": 23324
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.960662196561549e-10,
+      "loss": 0.7694,
+      "step": 23325
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8740940767834077e-10,
+      "loss": 0.8564,
+      "step": 23326
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.78848246167518e-10,
+      "loss": 0.8677,
+      "step": 23327
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7038273528799963e-10,
+      "loss": 0.8091,
+      "step": 23328
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.62012875201323e-10,
+      "loss": 0.8389,
+      "step": 23329
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.537386660679154e-10,
+      "loss": 0.8462,
+      "step": 23330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.455601080454285e-10,
+      "loss": 0.8496,
+      "step": 23331
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3747720129151395e-10,
+      "loss": 0.8413,
+      "step": 23332
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2948994595993765e-10,
+      "loss": 0.8213,
+      "step": 23333
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2159834220335526e-10,
+      "loss": 0.875,
+      "step": 23334
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1380239017386736e-10,
+      "loss": 0.8589,
+      "step": 23335
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.061020900191336e-10,
+      "loss": 0.8477,
+      "step": 23336
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.984974418873687e-10,
+      "loss": 0.8345,
+      "step": 23337
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9098844592401197e-10,
+      "loss": 0.8218,
+      "step": 23338
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8357510227283725e-10,
+      "loss": 0.9067,
+      "step": 23339
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7625741107539795e-10,
+      "loss": 0.8174,
+      "step": 23340
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6903537247102707e-10,
+      "loss": 0.8555,
+      "step": 23341
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6190898659961273e-10,
+      "loss": 0.8433,
+      "step": 23342
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.54878253596047e-10,
+      "loss": 0.834,
+      "step": 23343
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.47943173595222e-10,
+      "loss": 0.8408,
+      "step": 23344
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.411037467298094e-10,
+      "loss": 0.8521,
+      "step": 23345
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3435997313081547e-10,
+      "loss": 0.8062,
+      "step": 23346
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2771185292702614e-10,
+      "loss": 0.8008,
+      "step": 23347
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.21159386246117e-10,
+      "loss": 0.8149,
+      "step": 23348
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1470257321298815e-10,
+      "loss": 0.873,
+      "step": 23349
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0834141395087437e-10,
+      "loss": 0.8428,
+      "step": 23350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.020759085824553e-10,
+      "loss": 0.8271,
+      "step": 23351
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9590605722652477e-10,
+      "loss": 0.8506,
+      "step": 23352
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.898318600018767e-10,
+      "loss": 0.812,
+      "step": 23353
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8385331702452936e-10,
+      "loss": 0.8066,
+      "step": 23354
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7797042840883572e-10,
+      "loss": 0.7817,
+      "step": 23355
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.721831942669283e-10,
+      "loss": 0.8296,
+      "step": 23356
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6649161471038454e-10,
+      "loss": 0.8032,
+      "step": 23357
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6089568984745118e-10,
+      "loss": 0.8813,
+      "step": 23358
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5539541978526473e-10,
+      "loss": 0.8525,
+      "step": 23359
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4999080462874127e-10,
+      "loss": 0.8545,
+      "step": 23360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4468184448224176e-10,
+      "loss": 0.8618,
+      "step": 23361
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.394685394467965e-10,
+      "loss": 0.3214,
+      "step": 23362
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3435088962177044e-10,
+      "loss": 0.7969,
+      "step": 23363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2932889510597346e-10,
+      "loss": 0.8013,
+      "step": 23364
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.244025559943296e-10,
+      "loss": 0.7759,
+      "step": 23365
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1957187238176293e-10,
+      "loss": 0.7905,
+      "step": 23366
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1483684436097708e-10,
+      "loss": 0.8462,
+      "step": 23367
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1019747202190012e-10,
+      "loss": 0.9019,
+      "step": 23368
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0565375545390499e-10,
+      "loss": 0.7695,
+      "step": 23369
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0120569474303399e-10,
+      "loss": 0.8535,
+      "step": 23370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.68532899753294e-11,
+      "loss": 0.8286,
+      "step": 23371
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.259654123405792e-11,
+      "loss": 0.918,
+      "step": 23372
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.843544859971075e-11,
+      "loss": 0.8701,
+      "step": 23373
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.437001215277907e-11,
+      "loss": 0.8452,
+      "step": 23374
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.040023197042335e-11,
+      "loss": 0.8013,
+      "step": 23375
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.652610812924898e-11,
+      "loss": 0.8564,
+      "step": 23376
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.274764070308582e-11,
+      "loss": 0.7642,
+      "step": 23377
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.906482976409834e-11,
+      "loss": 0.8999,
+      "step": 23378
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.54776753822306e-11,
+      "loss": 0.8237,
+      "step": 23379
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.198617762742665e-11,
+      "loss": 0.7983,
+      "step": 23380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.859033656518964e-11,
+      "loss": 0.7418,
+      "step": 23381
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.529015226157786e-11,
+      "loss": 0.8066,
+      "step": 23382
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.208562477876378e-11,
+      "loss": 0.7905,
+      "step": 23383
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.897675417891989e-11,
+      "loss": 0.8062,
+      "step": 23384
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.596354052088803e-11,
+      "loss": 0.7876,
+      "step": 23385
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3045983862399776e-11,
+      "loss": 0.8213,
+      "step": 23386
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.02240842595214e-11,
+      "loss": 0.8022,
+      "step": 23387
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.749784176609872e-11,
+      "loss": 0.8911,
+      "step": 23388
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.486725643431221e-11,
+      "loss": 0.8696,
+      "step": 23389
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.233232831412192e-11,
+      "loss": 0.8813,
+      "step": 23390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.989305745493276e-11,
+      "loss": 0.8057,
+      "step": 23391
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7549443902818996e-11,
+      "loss": 0.8276,
+      "step": 23392
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5301487702189542e-11,
+      "loss": 0.835,
+      "step": 23393
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.314918889689821e-11,
+      "loss": 0.7935,
+      "step": 23394
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1092547528023256e-11,
+      "loss": 0.8203,
+      "step": 23395
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9131563633867366e-11,
+      "loss": 0.8193,
+      "step": 23396
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7266237253288353e-11,
+      "loss": 0.408,
+      "step": 23397
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.549656842125824e-11,
+      "loss": 0.8394,
+      "step": 23398
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3822557171638828e-11,
+      "loss": 0.7939,
+      "step": 23399
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2244203536626586e-11,
+      "loss": 0.8638,
+      "step": 23400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0761507546752647e-11,
+      "loss": 0.9058,
+      "step": 23401
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.374469229772586e-12,
+      "loss": 0.8564,
+      "step": 23402
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.083088612331758e-12,
+      "loss": 0.75,
+      "step": 23403
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.887365719410177e-12,
+      "loss": 0.8379,
+      "step": 23404
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.787300573767418e-12,
+      "loss": 0.7695,
+      "step": 23405
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.782893195942606e-12,
+      "loss": 0.8486,
+      "step": 23406
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.874143606474867e-12,
+      "loss": 0.7993,
+      "step": 23407
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0610518214624353e-12,
+      "loss": 0.8452,
+      "step": 23408
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3436178564484323e-12,
+      "loss": 0.8706,
+      "step": 23409
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7218417264208698e-12,
+      "loss": 0.8604,
+      "step": 23410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.195723441926866e-12,
+      "loss": 0.8242,
+      "step": 23411
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.652630140686512e-13,
+      "loss": 0.8159,
+      "step": 23412
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.304604500626752e-13,
+      "loss": 0.7573,
+      "step": 23413
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9131575712538765e-13,
+      "loss": 0.8081,
+      "step": 23414
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.782893969768054e-14,
+      "loss": 0.5804,
+      "step": 23415
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.8296,
+      "step": 23416
+    },
+    {
+      "epoch": 1.0,
+      "step": 23416,
+      "total_flos": 6.644263113097373e+19,
+      "train_loss": 0.8662753299026307,
+      "train_runtime": 200906.7429,
+      "train_samples_per_second": 7.459,
+      "train_steps_per_second": 0.117
+    }
+  ],
+  "max_steps": 23416,
+  "num_train_epochs": 1,
+  "total_flos": 6.644263113097373e+19,
+  "trial_name": null,
+  "trial_params": null
+}