diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,104070 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999423331987775,
+  "eval_steps": 500,
+  "global_step": 17340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9193857965451055e-07,
+      "loss": 40166.7656,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.838771593090211e-07,
+      "loss": 73310.4219,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.758157389635317e-07,
+      "loss": 95071.9688,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.677543186180422e-07,
+      "loss": 79584.2422,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.596928982725527e-07,
+      "loss": 103865.5156,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1516314779270634e-06,
+      "loss": 1.9477,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.343570057581574e-06,
+      "loss": 1.7322,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5355086372360844e-06,
+      "loss": 107714.1719,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7274472168905951e-06,
+      "loss": 126261.7031,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9193857965451054e-06,
+      "loss": 108728.9531,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1113243761996164e-06,
+      "loss": 1.9138,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.303262955854127e-06,
+      "loss": 1.9689,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4952015355086374e-06,
+      "loss": 95971.5781,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.687140115163148e-06,
+      "loss": 92665.9688,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.879078694817659e-06,
+      "loss": 1.8886,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.071017274472169e-06,
+      "loss": 82008.8516,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2629558541266794e-06,
+      "loss": 72805.3438,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4548944337811903e-06,
+      "loss": 1.8132,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.646833013435701e-06,
+      "loss": 1.7741,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.838771593090211e-06,
+      "loss": 52450.8867,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.030710172744722e-06,
+      "loss": 65211.1367,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.222648752399233e-06,
+      "loss": 71378.7031,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.414587332053743e-06,
+      "loss": 45700.1172,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.606525911708254e-06,
+      "loss": 46549.3594,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.798464491362764e-06,
+      "loss": 32401.2402,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.990403071017275e-06,
+      "loss": 32168.834,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.182341650671786e-06,
+      "loss": 27860.3672,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.374280230326296e-06,
+      "loss": 34493.3047,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.566218809980806e-06,
+      "loss": 1.8345,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.758157389635318e-06,
+      "loss": 16333.1211,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.950095969289828e-06,
+      "loss": 1.8693,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.142034548944338e-06,
+      "loss": 1.8715,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.333973128598848e-06,
+      "loss": 12766.8525,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.525911708253359e-06,
+      "loss": 1.6624,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.7178502879078705e-06,
+      "loss": 11145.9727,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.909788867562381e-06,
+      "loss": 10891.4502,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.101727447216891e-06,
+      "loss": 8938.0176,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.293666026871402e-06,
+      "loss": 6587.022,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.485604606525912e-06,
+      "loss": 6628.377,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.677543186180422e-06,
+      "loss": 3023.1895,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.869481765834934e-06,
+      "loss": 1.6204,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.061420345489444e-06,
+      "loss": 1.6935,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.253358925143954e-06,
+      "loss": 1.6084,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.445297504798465e-06,
+      "loss": 2018.8033,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.637236084452976e-06,
+      "loss": 1292.0242,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.829174664107486e-06,
+      "loss": 1212.6832,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.021113243761997e-06,
+      "loss": 956.6152,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.213051823416507e-06,
+      "loss": 1025.6605,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.404990403071018e-06,
+      "loss": 840.73,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.596928982725528e-06,
+      "loss": 867.1953,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.78886756238004e-06,
+      "loss": 1.4913,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.98080614203455e-06,
+      "loss": 1.4901,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0172744721689061e-05,
+      "loss": 453.9937,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0364683301343571e-05,
+      "loss": 448.7566,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0556621880998081e-05,
+      "loss": 393.6455,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0748560460652591e-05,
+      "loss": 1.4552,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0940499040307102e-05,
+      "loss": 1.349,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1132437619961612e-05,
+      "loss": 1.3568,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1324376199616123e-05,
+      "loss": 383.8577,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1516314779270635e-05,
+      "loss": 1.2666,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1708253358925145e-05,
+      "loss": 1.2785,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1900191938579655e-05,
+      "loss": 383.9733,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2092130518234165e-05,
+      "loss": 1.2908,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2284069097888675e-05,
+      "loss": 1.3197,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2476007677543186e-05,
+      "loss": 1.2793,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2667946257197696e-05,
+      "loss": 364.9019,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2859884836852207e-05,
+      "loss": 1.1927,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3051823416506717e-05,
+      "loss": 362.1893,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3243761996161231e-05,
+      "loss": 1.2217,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3435700575815741e-05,
+      "loss": 346.6711,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3627639155470251e-05,
+      "loss": 337.8551,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3819577735124761e-05,
+      "loss": 337.3083,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4011516314779271e-05,
+      "loss": 335.0121,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4203454894433781e-05,
+      "loss": 317.3089,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4395393474088293e-05,
+      "loss": 312.041,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4587332053742803e-05,
+      "loss": 294.4712,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4779270633397313e-05,
+      "loss": 287.8751,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4971209213051823e-05,
+      "loss": 1.1509,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5163147792706333e-05,
+      "loss": 1.1098,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5355086372360844e-05,
+      "loss": 1.1595,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5547024952015357e-05,
+      "loss": 539.9817,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5738963531669867e-05,
+      "loss": 1.1271,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5930902111324377e-05,
+      "loss": 291.2214,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6122840690978887e-05,
+      "loss": 303.7813,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6314779270633397e-05,
+      "loss": 253.3305,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6506717850287907e-05,
+      "loss": 250.7164,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.669865642994242e-05,
+      "loss": 239.5748,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.689059500959693e-05,
+      "loss": 1.1427,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.708253358925144e-05,
+      "loss": 1.1159,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.727447216890595e-05,
+      "loss": 245.1028,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.746641074856046e-05,
+      "loss": 1.121,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.765834932821497e-05,
+      "loss": 1.0767,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.785028790786948e-05,
+      "loss": 1.1368,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8042226487523995e-05,
+      "loss": 220.8601,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8234165067178505e-05,
+      "loss": 212.253,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8426103646833015e-05,
+      "loss": 216.5695,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8618042226487525e-05,
+      "loss": 1.0932,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8809980806142035e-05,
+      "loss": 1.0794,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9001919385796545e-05,
+      "loss": 1.0733,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9193857965451055e-05,
+      "loss": 1.0587,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9385796545105565e-05,
+      "loss": 1.1141,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.957773512476008e-05,
+      "loss": 1.0634,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.976967370441459e-05,
+      "loss": 269.3731,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99616122840691e-05,
+      "loss": 205.2335,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.015355086372361e-05,
+      "loss": 189.3826,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0345489443378122e-05,
+      "loss": 193.7715,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0537428023032633e-05,
+      "loss": 191.3852,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0729366602687143e-05,
+      "loss": 1.1293,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0921305182341653e-05,
+      "loss": 184.698,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1113243761996163e-05,
+      "loss": 184.1337,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1305182341650673e-05,
+      "loss": 178.6816,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1497120921305183e-05,
+      "loss": 173.9807,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1689059500959693e-05,
+      "loss": 161.7277,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1880998080614203e-05,
+      "loss": 161.8288,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2072936660268713e-05,
+      "loss": 1.1047,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2264875239923223e-05,
+      "loss": 152.4367,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2456813819577733e-05,
+      "loss": 150.5476,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2648752399232247e-05,
+      "loss": 150.8228,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2840690978886757e-05,
+      "loss": 1.1051,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.303262955854127e-05,
+      "loss": 1.0705,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.322456813819578e-05,
+      "loss": 147.1015,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.341650671785029e-05,
+      "loss": 156.359,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.36084452975048e-05,
+      "loss": 1.0691,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.380038387715931e-05,
+      "loss": 1.0717,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.399232245681382e-05,
+      "loss": 154.4477,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.418426103646833e-05,
+      "loss": 1.0801,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.437619961612284e-05,
+      "loss": 1.1243,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.456813819577735e-05,
+      "loss": 135.3799,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.476007677543186e-05,
+      "loss": 1.07,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.495201535508637e-05,
+      "loss": 1.0863,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.514395393474088e-05,
+      "loss": 138.7148,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.533589251439539e-05,
+      "loss": 137.2117,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5527831094049905e-05,
+      "loss": 1.0775,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5719769673704415e-05,
+      "loss": 131.6057,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5911708253358925e-05,
+      "loss": 131.1676,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6103646833013435e-05,
+      "loss": 129.2896,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6295585412667952e-05,
+      "loss": 1.0441,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6487523992322462e-05,
+      "loss": 1.0824,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6679462571976972e-05,
+      "loss": 128.9653,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6871401151631482e-05,
+      "loss": 126.2059,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7063339731285992e-05,
+      "loss": 1.0538,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7255278310940502e-05,
+      "loss": 1.0359,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7447216890595012e-05,
+      "loss": 1.0504,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7639155470249522e-05,
+      "loss": 127.3239,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7831094049904032e-05,
+      "loss": 124.3072,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8023032629558543e-05,
+      "loss": 1.0591,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8214971209213053e-05,
+      "loss": 1.0041,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8406909788867563e-05,
+      "loss": 132.9683,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8598848368522073e-05,
+      "loss": 1.0375,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8790786948176586e-05,
+      "loss": 121.9531,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8982725527831096e-05,
+      "loss": 125.8824,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9174664107485606e-05,
+      "loss": 118.8859,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9366602687140116e-05,
+      "loss": 1.0368,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9558541266794627e-05,
+      "loss": 114.5575,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9750479846449137e-05,
+      "loss": 1.1135,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9942418426103647e-05,
+      "loss": 113.9053,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0134357005758157e-05,
+      "loss": 1.0075,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0326295585412667e-05,
+      "loss": 116.931,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.051823416506718e-05,
+      "loss": 1.0457,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.071017274472169e-05,
+      "loss": 113.5277,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.09021113243762e-05,
+      "loss": 0.9514,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1094049904030714e-05,
+      "loss": 119.4041,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.128598848368523e-05,
+      "loss": 109.4245,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1477927063339734e-05,
+      "loss": 0.997,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.166986564299425e-05,
+      "loss": 114.5618,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1861804222648754e-05,
+      "loss": 112.6653,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.205374280230327e-05,
+      "loss": 105.2181,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2245681381957774e-05,
+      "loss": 116.1238,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.243761996161229e-05,
+      "loss": 0.9277,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2629558541266795e-05,
+      "loss": 226.9309,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.282149712092131e-05,
+      "loss": 0.9637,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3013435700575815e-05,
+      "loss": 327.947,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.320537428023033e-05,
+      "loss": 183.6508,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.339731285988484e-05,
+      "loss": 109.433,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.358925143953935e-05,
+      "loss": 0.9729,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.378119001919386e-05,
+      "loss": 113.9853,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.397312859884837e-05,
+      "loss": 117.1907,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.416506717850288e-05,
+      "loss": 115.6648,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.435700575815739e-05,
+      "loss": 0.9352,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.45489443378119e-05,
+      "loss": 112.148,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.474088291746641e-05,
+      "loss": 110.4049,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.493282149712092e-05,
+      "loss": 118.2461,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.512476007677543e-05,
+      "loss": 111.5398,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.531669865642994e-05,
+      "loss": 102.907,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.550863723608445e-05,
+      "loss": 101.3712,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.570057581573896e-05,
+      "loss": 108.8732,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5892514395393476e-05,
+      "loss": 0.9414,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.608445297504799e-05,
+      "loss": 0.9671,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6276391554702496e-05,
+      "loss": 0.9478,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.646833013435701e-05,
+      "loss": 100.8114,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6660268714011516e-05,
+      "loss": 98.6187,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.685220729366603e-05,
+      "loss": 0.9533,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.704414587332054e-05,
+      "loss": 0.9157,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.723608445297505e-05,
+      "loss": 115.3182,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7428023032629563e-05,
+      "loss": 107.1674,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.761996161228407e-05,
+      "loss": 0.9178,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7811900191938584e-05,
+      "loss": 100.2813,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.800383877159309e-05,
+      "loss": 0.9535,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8195777351247604e-05,
+      "loss": 0.9141,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.838771593090211e-05,
+      "loss": 99.7219,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8579654510556624e-05,
+      "loss": 94.338,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.877159309021113e-05,
+      "loss": 103.0458,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8963531669865644e-05,
+      "loss": 104.3298,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.915547024952016e-05,
+      "loss": 97.9568,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9347408829174664e-05,
+      "loss": 98.3586,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.953934740882918e-05,
+      "loss": 95.0431,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9731285988483684e-05,
+      "loss": 0.8885,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.99232245681382e-05,
+      "loss": 96.1339,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0115163147792705e-05,
+      "loss": 0.8652,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.030710172744722e-05,
+      "loss": 0.9018,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.049904030710173e-05,
+      "loss": 95.0032,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0690978886756245e-05,
+      "loss": 97.9753,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.088291746641075e-05,
+      "loss": 93.4375,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1074856046065265e-05,
+      "loss": 0.9281,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.126679462571977e-05,
+      "loss": 88.0535,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1458733205374285e-05,
+      "loss": 91.5855,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.165067178502879e-05,
+      "loss": 87.185,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1842610364683305e-05,
+      "loss": 88.2069,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.203454894433781e-05,
+      "loss": 0.855,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2226487523992326e-05,
+      "loss": 89.9853,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.241842610364683e-05,
+      "loss": 93.2096,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2610364683301346e-05,
+      "loss": 0.7879,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.280230326295586e-05,
+      "loss": 85.7833,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2994241842610366e-05,
+      "loss": 86.8817,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.318618042226488e-05,
+      "loss": 92.64,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3378119001919386e-05,
+      "loss": 0.8398,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.35700575815739e-05,
+      "loss": 97.1786,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3761996161228406e-05,
+      "loss": 92.35,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.395393474088292e-05,
+      "loss": 0.8305,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4145873320537426e-05,
+      "loss": 90.8982,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.433781190019194e-05,
+      "loss": 88.2569,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4529750479846447e-05,
+      "loss": 0.8497,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.472168905950096e-05,
+      "loss": 85.8387,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.491362763915547e-05,
+      "loss": 0.8362,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.510556621880998e-05,
+      "loss": 86.8115,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5297504798464494e-05,
+      "loss": 82.9474,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.548944337811901e-05,
+      "loss": 79.5182,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5681381957773514e-05,
+      "loss": 0.9095,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.587332053742803e-05,
+      "loss": 0.8408,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.606525911708254e-05,
+      "loss": 82.9227,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.625719769673705e-05,
+      "loss": 0.8077,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.644913627639156e-05,
+      "loss": 83.2174,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.664107485604607e-05,
+      "loss": 78.9564,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.683301343570058e-05,
+      "loss": 0.8166,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.702495201535509e-05,
+      "loss": 93.194,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.72168905950096e-05,
+      "loss": 88.8424,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.740882917466411e-05,
+      "loss": 86.0309,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.760076775431862e-05,
+      "loss": 86.146,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.779270633397313e-05,
+      "loss": 90.7753,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.798464491362764e-05,
+      "loss": 86.0763,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.817658349328215e-05,
+      "loss": 81.7651,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.836852207293666e-05,
+      "loss": 85.1543,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8560460652591175e-05,
+      "loss": 0.8212,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.875239923224568e-05,
+      "loss": 0.7627,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8944337811900195e-05,
+      "loss": 103.6813,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.91362763915547e-05,
+      "loss": 0.8884,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9328214971209215e-05,
+      "loss": 91.0491,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.952015355086372e-05,
+      "loss": 0.7865,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9712092130518236e-05,
+      "loss": 76.0393,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.990403071017274e-05,
+      "loss": 86.7888,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.009596928982726e-05,
+      "loss": 78.2452,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.028790786948176e-05,
+      "loss": 81.8235,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.047984644913628e-05,
+      "loss": 88.3179,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.067178502879078e-05,
+      "loss": 83.2756,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.08637236084453e-05,
+      "loss": 80.1587,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.105566218809981e-05,
+      "loss": 83.0204,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.124760076775432e-05,
+      "loss": 0.8268,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.143953934740883e-05,
+      "loss": 0.7725,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.163147792706334e-05,
+      "loss": 81.1126,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.182341650671785e-05,
+      "loss": 78.652,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.201535508637236e-05,
+      "loss": 76.3936,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.220729366602687e-05,
+      "loss": 0.7722,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2399232245681383e-05,
+      "loss": 75.1075,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2591170825335904e-05,
+      "loss": 0.8047,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2783109404990404e-05,
+      "loss": 89.6875,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2975047984644924e-05,
+      "loss": 0.8304,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3166986564299424e-05,
+      "loss": 83.3886,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3358925143953944e-05,
+      "loss": 84.9924,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3550863723608444e-05,
+      "loss": 0.7721,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3742802303262964e-05,
+      "loss": 85.9164,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3934740882917464e-05,
+      "loss": 79.7156,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4126679462571984e-05,
+      "loss": 82.8379,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.431861804222649e-05,
+      "loss": 0.7748,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4510556621881004e-05,
+      "loss": 87.2387,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.470249520153551e-05,
+      "loss": 84.4544,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4894433781190025e-05,
+      "loss": 83.1652,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.508637236084453e-05,
+      "loss": 79.1723,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5278310940499045e-05,
+      "loss": 77.1423,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.547024952015355e-05,
+      "loss": 80.9267,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5662188099808065e-05,
+      "loss": 0.7777,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.585412667946257e-05,
+      "loss": 84.6023,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6046065259117085e-05,
+      "loss": 83.0093,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.623800383877159e-05,
+      "loss": 74.7592,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6429942418426105e-05,
+      "loss": 82.1833,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.662188099808061e-05,
+      "loss": 78.8465,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6813819577735125e-05,
+      "loss": 0.8328,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.700575815738963e-05,
+      "loss": 0.7634,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7197696737044146e-05,
+      "loss": 81.1641,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7389635316698666e-05,
+      "loss": 78.7428,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.758157389635317e-05,
+      "loss": 0.7566,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7773512476007686e-05,
+      "loss": 0.7901,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.796545105566219e-05,
+      "loss": 80.7445,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8157389635316706e-05,
+      "loss": 78.9492,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.834932821497121e-05,
+      "loss": 76.977,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8541266794625726e-05,
+      "loss": 79.827,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.873320537428023e-05,
+      "loss": 76.0504,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8925143953934746e-05,
+      "loss": 76.7126,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.911708253358925e-05,
+      "loss": 78.331,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9309021113243767e-05,
+      "loss": 77.2228,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.950095969289827e-05,
+      "loss": 0.7934,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.969289827255279e-05,
+      "loss": 0.8279,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9884836852207293e-05,
+      "loss": 105.152,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.007677543186181e-05,
+      "loss": 112.3333,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0268714011516314e-05,
+      "loss": 0.8013,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.046065259117083e-05,
+      "loss": 85.0341,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0652591170825334e-05,
+      "loss": 0.7942,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.084452975047985e-05,
+      "loss": 0.7712,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.103646833013436e-05,
+      "loss": 82.2815,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.122840690978887e-05,
+      "loss": 0.7876,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.142034548944337e-05,
+      "loss": 77.54,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.16122840690979e-05,
+      "loss": 0.7518,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.18042226487524e-05,
+      "loss": 77.0724,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.199616122840691e-05,
+      "loss": 75.6224,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.218809980806143e-05,
+      "loss": 74.2598,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.238003838771593e-05,
+      "loss": 72.7228,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.257197696737045e-05,
+      "loss": 72.9263,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.276391554702495e-05,
+      "loss": 78.9021,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.295585412667947e-05,
+      "loss": 0.7926,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.314779270633397e-05,
+      "loss": 73.4431,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.33397312859885e-05,
+      "loss": 0.8295,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3531669865643e-05,
+      "loss": 82.0516,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.372360844529751e-05,
+      "loss": 88.9239,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.391554702495202e-05,
+      "loss": 76.7804,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.410748560460654e-05,
+      "loss": 0.7359,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.429942418426104e-05,
+      "loss": 0.7906,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.449136276391555e-05,
+      "loss": 0.7918,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.468330134357006e-05,
+      "loss": 0.7607,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.487523992322458e-05,
+      "loss": 0.7655,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.506717850287908e-05,
+      "loss": 0.7146,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.525911708253359e-05,
+      "loss": 0.7862,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.54510556621881e-05,
+      "loss": 91.1832,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.564299424184262e-05,
+      "loss": 88.8838,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.583493282149712e-05,
+      "loss": 84.5475,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.602687140115163e-05,
+      "loss": 81.2926,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.621880998080614e-05,
+      "loss": 75.3123,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.641074856046066e-05,
+      "loss": 83.9268,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.660268714011516e-05,
+      "loss": 0.852,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.679462571976968e-05,
+      "loss": 190.6337,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.698656429942419e-05,
+      "loss": 234.6956,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.71785028790787e-05,
+      "loss": 0.7736,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.737044145873322e-05,
+      "loss": 158.735,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.756238003838772e-05,
+      "loss": 0.8039,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.775431861804223e-05,
+      "loss": 0.8246,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.794625719769674e-05,
+      "loss": 0.7682,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.813819577735126e-05,
+      "loss": 78.9516,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.833013435700576e-05,
+      "loss": 0.7724,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.852207293666027e-05,
+      "loss": 0.8067,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.871401151631478e-05,
+      "loss": 0.7614,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.89059500959693e-05,
+      "loss": 0.7616,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.90978886756238e-05,
+      "loss": 83.9511,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.928982725527831e-05,
+      "loss": 81.7378,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.948176583493282e-05,
+      "loss": 0.7891,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.967370441458734e-05,
+      "loss": 0.7743,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.986564299424184e-05,
+      "loss": 75.6511,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.005758157389636e-05,
+      "loss": 75.2006,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.024952015355086e-05,
+      "loss": 75.0797,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.044145873320538e-05,
+      "loss": 0.8055,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.063339731285988e-05,
+      "loss": 90.2629,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.08253358925144e-05,
+      "loss": 0.7491,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.10172744721689e-05,
+      "loss": 88.1594,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.120921305182342e-05,
+      "loss": 77.3653,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.140115163147793e-05,
+      "loss": 79.7959,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.159309021113245e-05,
+      "loss": 0.8117,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.178502879078695e-05,
+      "loss": 0.7932,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.197696737044146e-05,
+      "loss": 84.6256,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.216890595009598e-05,
+      "loss": 88.4154,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.236084452975049e-05,
+      "loss": 0.7713,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.255278310940499e-05,
+      "loss": 0.7977,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.27447216890595e-05,
+      "loss": 0.8082,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.293666026871402e-05,
+      "loss": 0.7569,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.312859884836853e-05,
+      "loss": 86.5129,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.332053742802303e-05,
+      "loss": 84.8625,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.351247600767754e-05,
+      "loss": 79.312,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.370441458733206e-05,
+      "loss": 0.7814,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.389635316698657e-05,
+      "loss": 86.2043,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.408829174664109e-05,
+      "loss": 0.8093,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.428023032629558e-05,
+      "loss": 124.0445,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.44721689059501e-05,
+      "loss": 0.8225,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.46641074856046e-05,
+      "loss": 117.2097,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.485604606525913e-05,
+      "loss": 91.5526,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.504798464491363e-05,
+      "loss": 77.6554,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.523992322456814e-05,
+      "loss": 75.9458,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.543186180422265e-05,
+      "loss": 0.7976,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.562380038387717e-05,
+      "loss": 76.1995,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.581573896353167e-05,
+      "loss": 0.8242,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.600767754318618e-05,
+      "loss": 85.1662,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.61996161228407e-05,
+      "loss": 78.9396,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.639155470249521e-05,
+      "loss": 74.7964,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.658349328214971e-05,
+      "loss": 70.8579,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.677543186180422e-05,
+      "loss": 73.7832,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.696737044145874e-05,
+      "loss": 0.8033,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.715930902111325e-05,
+      "loss": 77.8893,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.735124760076777e-05,
+      "loss": 0.803,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.754318618042226e-05,
+      "loss": 0.7583,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.773512476007678e-05,
+      "loss": 74.7665,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.792706333973129e-05,
+      "loss": 68.9323,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.811900191938581e-05,
+      "loss": 75.8076,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.831094049904032e-05,
+      "loss": 73.5018,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.850287907869482e-05,
+      "loss": 0.7905,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.869481765834933e-05,
+      "loss": 0.8641,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.888675623800385e-05,
+      "loss": 99.1582,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.907869481765836e-05,
+      "loss": 110.6431,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.927063339731286e-05,
+      "loss": 0.754,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.946257197696737e-05,
+      "loss": 80.5964,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.965451055662189e-05,
+      "loss": 81.2054,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.98464491362764e-05,
+      "loss": 0.7913,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.00383877159309e-05,
+      "loss": 78.7256,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.023032629558541e-05,
+      "loss": 78.5656,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.042226487523993e-05,
+      "loss": 72.5676,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.061420345489444e-05,
+      "loss": 66.8839,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.080614203454894e-05,
+      "loss": 74.1059,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.099808061420346e-05,
+      "loss": 0.7869,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.119001919385797e-05,
+      "loss": 74.7156,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.138195777351249e-05,
+      "loss": 91.784,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.157389635316698e-05,
+      "loss": 79.93,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.17658349328215e-05,
+      "loss": 76.4785,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.195777351247601e-05,
+      "loss": 0.8145,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.214971209213053e-05,
+      "loss": 77.5992,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.234165067178504e-05,
+      "loss": 78.1748,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.253358925143954e-05,
+      "loss": 73.5875,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.272552783109405e-05,
+      "loss": 0.7707,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.291746641074857e-05,
+      "loss": 71.9324,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.310940499040308e-05,
+      "loss": 0.8131,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.330134357005758e-05,
+      "loss": 0.7346,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.349328214971209e-05,
+      "loss": 0.7521,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.368522072936661e-05,
+      "loss": 93.7439,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.387715930902112e-05,
+      "loss": 0.7795,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.406909788867562e-05,
+      "loss": 0.7316,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.426103646833013e-05,
+      "loss": 0.8339,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.445297504798465e-05,
+      "loss": 107.803,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.464491362763916e-05,
+      "loss": 0.7182,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.483685220729366e-05,
+      "loss": 0.7718,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.502879078694817e-05,
+      "loss": 81.2979,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.522072936660269e-05,
+      "loss": 76.3163,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.54126679462572e-05,
+      "loss": 0.7659,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.560460652591172e-05,
+      "loss": 75.0455,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.579654510556623e-05,
+      "loss": 72.6597,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.598848368522073e-05,
+      "loss": 72.3253,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.618042226487525e-05,
+      "loss": 0.7204,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.637236084452976e-05,
+      "loss": 76.5666,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.656429942418427e-05,
+      "loss": 75.655,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.675623800383877e-05,
+      "loss": 71.8046,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.694817658349329e-05,
+      "loss": 0.7743,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.71401151631478e-05,
+      "loss": 78.1394,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.73320537428023e-05,
+      "loss": 69.7749,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.752399232245681e-05,
+      "loss": 0.8103,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.771593090211133e-05,
+      "loss": 74.7982,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.790786948176584e-05,
+      "loss": 73.3477,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.809980806142035e-05,
+      "loss": 67.4092,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.829174664107485e-05,
+      "loss": 0.7203,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.848368522072937e-05,
+      "loss": 70.1887,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.867562380038388e-05,
+      "loss": 66.4563,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.88675623800384e-05,
+      "loss": 73.0604,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.905950095969289e-05,
+      "loss": 0.7745,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.925143953934741e-05,
+      "loss": 65.0886,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.944337811900192e-05,
+      "loss": 68.1133,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.963531669865644e-05,
+      "loss": 0.7522,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.982725527831093e-05,
+      "loss": 0.7901,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.001919385796545e-05,
+      "loss": 68.3609,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.021113243761996e-05,
+      "loss": 0.7874,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.040307101727448e-05,
+      "loss": 0.7991,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.059500959692899e-05,
+      "loss": 67.737,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.07869481765835e-05,
+      "loss": 0.7644,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.097888675623801e-05,
+      "loss": 67.5394,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.117082533589252e-05,
+      "loss": 0.8193,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.136276391554703e-05,
+      "loss": 0.7158,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.155470249520153e-05,
+      "loss": 68.227,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.174664107485605e-05,
+      "loss": 0.7437,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.193857965451056e-05,
+      "loss": 67.8273,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.213051823416508e-05,
+      "loss": 68.2516,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.232245681381957e-05,
+      "loss": 0.7502,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.25143953934741e-05,
+      "loss": 0.8252,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.27063339731286e-05,
+      "loss": 72.815,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.289827255278312e-05,
+      "loss": 0.7937,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.309021113243761e-05,
+      "loss": 72.992,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.328214971209214e-05,
+      "loss": 72.6453,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.347408829174664e-05,
+      "loss": 0.7746,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.366602687140116e-05,
+      "loss": 72.0008,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.385796545105567e-05,
+      "loss": 0.7733,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.404990403071018e-05,
+      "loss": 72.3265,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.424184261036468e-05,
+      "loss": 0.768,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.44337811900192e-05,
+      "loss": 69.5052,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.462571976967371e-05,
+      "loss": 66.5822,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.481765834932822e-05,
+      "loss": 65.2336,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.500959692898272e-05,
+      "loss": 0.7269,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.520153550863724e-05,
+      "loss": 72.0567,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.539347408829176e-05,
+      "loss": 66.2869,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.558541266794626e-05,
+      "loss": 67.4035,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.577735124760078e-05,
+      "loss": 65.3684,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.596928982725528e-05,
+      "loss": 0.7471,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.61612284069098e-05,
+      "loss": 0.7955,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.63531669865643e-05,
+      "loss": 0.7404,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.654510556621882e-05,
+      "loss": 70.8317,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.673704414587332e-05,
+      "loss": 70.5794,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.692898272552784e-05,
+      "loss": 64.8721,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.712092130518235e-05,
+      "loss": 65.4788,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.731285988483686e-05,
+      "loss": 60.6346,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.750479846449136e-05,
+      "loss": 63.4021,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.769673704414588e-05,
+      "loss": 67.2244,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.788867562380039e-05,
+      "loss": 65.3711,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.80806142034549e-05,
+      "loss": 63.8599,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.82725527831094e-05,
+      "loss": 0.7572,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.846449136276392e-05,
+      "loss": 64.8392,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.865642994241843e-05,
+      "loss": 0.7769,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.884836852207294e-05,
+      "loss": 63.2517,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904030710172744e-05,
+      "loss": 66.8697,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923224568138196e-05,
+      "loss": 64.1417,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.942418426103647e-05,
+      "loss": 0.7287,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.961612284069098e-05,
+      "loss": 0.8089,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980806142034548e-05,
+      "loss": 67.1365,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8008,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999912785687e-05,
+      "loss": 0.7736,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999651142747e-05,
+      "loss": 66.1506,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999215071192e-05,
+      "loss": 63.315,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998604571035e-05,
+      "loss": 61.8157,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9999978196423e-05,
+      "loss": 0.8275,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999686028501e-05,
+      "loss": 62.4379,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999995726499202e-05,
+      "loss": 65.028,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999994418284918e-05,
+      "loss": 65.3495,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999992935642197e-05,
+      "loss": 65.3737,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999991278571095e-05,
+      "loss": 0.8167,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99998944707167e-05,
+      "loss": 63.2766,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999987441143984e-05,
+      "loss": 61.4326,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999985260788107e-05,
+      "loss": 60.4002,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999982906004118e-05,
+      "loss": 0.7934,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999980376792096e-05,
+      "loss": 0.7451,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99997767315213e-05,
+      "loss": 65.8505,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999974795084314e-05,
+      "loss": 0.8436,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99997174258875e-05,
+      "loss": 64.1063,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999968515665543e-05,
+      "loss": 62.3916,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999965114314806e-05,
+      "loss": 0.7798,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999961538536657e-05,
+      "loss": 62.5999,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999957788331222e-05,
+      "loss": 0.8129,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999953863698632e-05,
+      "loss": 62.9053,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999949764639023e-05,
+      "loss": 62.899,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999945491152538e-05,
+      "loss": 0.7486,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999941043239327e-05,
+      "loss": 66.3384,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999936420899543e-05,
+      "loss": 0.7494,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99993162413335e-05,
+      "loss": 63.0179,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999926652940913e-05,
+      "loss": 62.1264,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999921507322407e-05,
+      "loss": 0.7266,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999916187278012e-05,
+      "loss": 0.8292,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99991069280791e-05,
+      "loss": 62.2639,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999905023912298e-05,
+      "loss": 62.3632,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999899180591368e-05,
+      "loss": 63.7898,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999893162845328e-05,
+      "loss": 0.7871,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999886970674388e-05,
+      "loss": 58.4963,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999880604078762e-05,
+      "loss": 61.0342,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999874063058674e-05,
+      "loss": 59.7577,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999867347614349e-05,
+      "loss": 0.8256,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999860457746025e-05,
+      "loss": 61.1215,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999853393453939e-05,
+      "loss": 0.7281,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999846154738342e-05,
+      "loss": 0.7309,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999838741599482e-05,
+      "loss": 61.3299,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999831154037619e-05,
+      "loss": 0.7603,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999823392053018e-05,
+      "loss": 59.3014,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999815455645951e-05,
+      "loss": 61.1229,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999807344816693e-05,
+      "loss": 55.8343,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999799059565527e-05,
+      "loss": 61.0488,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999790599892744e-05,
+      "loss": 0.7761,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999781965798635e-05,
+      "loss": 65.8781,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999773157283508e-05,
+      "loss": 0.777,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999764174347664e-05,
+      "loss": 60.1417,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99975501699142e-05,
+      "loss": 0.7407,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999745685215092e-05,
+      "loss": 57.5513,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999736179019008e-05,
+      "loss": 62.6229,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999726498403498e-05,
+      "loss": 0.808,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999716643368904e-05,
+      "loss": 63.6162,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999706613915566e-05,
+      "loss": 63.1468,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999696410043835e-05,
+      "loss": 61.8665,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999686031754065e-05,
+      "loss": 60.1932,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999675479046622e-05,
+      "loss": 59.3034,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99966475192187e-05,
+      "loss": 62.5836,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999653850380186e-05,
+      "loss": 59.8679,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999642774421949e-05,
+      "loss": 59.366,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999631524047546e-05,
+      "loss": 54.4645,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999620099257367e-05,
+      "loss": 64.8119,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999608500051816e-05,
+      "loss": 60.1273,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999596726431292e-05,
+      "loss": 0.7571,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99958477839621e-05,
+      "loss": 59.9806,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999572655946985e-05,
+      "loss": 56.3806,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999560359084039e-05,
+      "loss": 59.7176,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999547887807803e-05,
+      "loss": 54.9886,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999535242118709e-05,
+      "loss": 57.6693,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999522422017202e-05,
+      "loss": 60.7395,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999509427503727e-05,
+      "loss": 0.7983,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999496258578738e-05,
+      "loss": 59.4488,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999482915242693e-05,
+      "loss": 59.6845,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999469397496059e-05,
+      "loss": 59.019,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999455705339307e-05,
+      "loss": 57.335,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999441838772914e-05,
+      "loss": 58.1484,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999427797797367e-05,
+      "loss": 59.2674,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99941358241315e-05,
+      "loss": 0.7402,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999399192620767e-05,
+      "loss": 0.7264,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999384628420712e-05,
+      "loss": 58.3062,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999369889813497e-05,
+      "loss": 0.7934,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999354976799635e-05,
+      "loss": 57.0756,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999339889379647e-05,
+      "loss": 56.4437,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999324627554059e-05,
+      "loss": 0.7377,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999309191323404e-05,
+      "loss": 54.1689,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999293580688221e-05,
+      "loss": 53.4719,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999277795649052e-05,
+      "loss": 0.7615,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999261836206449e-05,
+      "loss": 0.7405,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999245702360971e-05,
+      "loss": 0.7799,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999229394113177e-05,
+      "loss": 0.8087,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99921291146364e-05,
+      "loss": 0.7843,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999196254412931e-05,
+      "loss": 0.7419,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999179422961633e-05,
+      "loss": 125.9519,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999162417110335e-05,
+      "loss": 101.596,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999145236859626e-05,
+      "loss": 71.4846,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999127882210108e-05,
+      "loss": 58.4715,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999110353162389e-05,
+      "loss": 59.1621,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999092649717074e-05,
+      "loss": 64.9901,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999074771874786e-05,
+      "loss": 55.4625,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999056719636149e-05,
+      "loss": 57.919,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999038493001787e-05,
+      "loss": 55.5458,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999020091972341e-05,
+      "loss": 57.9309,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999001516548452e-05,
+      "loss": 55.9897,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998982766730766e-05,
+      "loss": 57.7613,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998963842519942e-05,
+      "loss": 0.7882,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998944743916633e-05,
+      "loss": 0.8263,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998925470921511e-05,
+      "loss": 53.4802,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998906023535247e-05,
+      "loss": 0.7787,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998886401758519e-05,
+      "loss": 55.4178,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998866605592009e-05,
+      "loss": 56.5564,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998846635036412e-05,
+      "loss": 0.787,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998826490092421e-05,
+      "loss": 57.1944,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998806170760741e-05,
+      "loss": 56.2309,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998785677042082e-05,
+      "loss": 0.8344,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998765008937155e-05,
+      "loss": 0.735,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998744166446686e-05,
+      "loss": 54.9132,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998723149571396e-05,
+      "loss": 58.6068,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998701958312022e-05,
+      "loss": 54.8101,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998680592669304e-05,
+      "loss": 0.7657,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998659052643987e-05,
+      "loss": 56.6903,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99863733823682e-05,
+      "loss": 65.6481,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998615449448565e-05,
+      "loss": 57.3414,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998593386279979e-05,
+      "loss": 0.7977,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998571148731837e-05,
+      "loss": 56.7333,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998548736804913e-05,
+      "loss": 0.7287,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998526150499989e-05,
+      "loss": 58.6507,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998503389817855e-05,
+      "loss": 58.0874,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9984804547593e-05,
+      "loss": 53.7582,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99845734532513e-05,
+      "loss": 54.4274,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998434061516147e-05,
+      "loss": 55.7145,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998410603333163e-05,
+      "loss": 58.6054,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998386970777e-05,
+      "loss": 0.7883,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99836316384848e-05,
+      "loss": 54.1392,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998339182548433e-05,
+      "loss": 0.7215,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998315026877698e-05,
+      "loss": 54.6837,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998290696837115e-05,
+      "loss": 57.2754,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998266192427534e-05,
+      "loss": 0.8524,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998241513649811e-05,
+      "loss": 0.7995,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998216660504804e-05,
+      "loss": 56.7577,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998191632993384e-05,
+      "loss": 57.1246,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99816643111642e-05,
+      "loss": 55.8684,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998141054874794e-05,
+      "loss": 58.0172,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99811550426939e-05,
+      "loss": 50.9315,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9980897793011e-05,
+      "loss": 52.9566,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99806387997082e-05,
+      "loss": 55.7927,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998037806279457e-05,
+      "loss": 51.4601,
+      "step": 671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998011558227916e-05,
+      "loss": 52.8647,
+      "step": 672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997985135817116e-05,
+      "loss": 53.2382,
+      "step": 673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997958539047978e-05,
+      "loss": 54.7912,
+      "step": 674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997931767921429e-05,
+      "loss": 0.7546,
+      "step": 675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997904822438403e-05,
+      "loss": 54.6195,
+      "step": 676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99787770259984e-05,
+      "loss": 0.808,
+      "step": 677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997850408406688e-05,
+      "loss": 0.8002,
+      "step": 678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997822939859898e-05,
+      "loss": 0.7663,
+      "step": 679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997795296960427e-05,
+      "loss": 52.9767,
+      "step": 680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997767479709241e-05,
+      "loss": 0.7674,
+      "step": 681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99773948810731e-05,
+      "loss": 0.7453,
+      "step": 682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99771132215561e-05,
+      "loss": 0.81,
+      "step": 683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997682981855123e-05,
+      "loss": 56.3825,
+      "step": 684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997654467206839e-05,
+      "loss": 0.77,
+      "step": 685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997625778211752e-05,
+      "loss": 57.3114,
+      "step": 686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997596914870864e-05,
+      "loss": 0.8152,
+      "step": 687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997567877185181e-05,
+      "loss": 0.7868,
+      "step": 688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997538665155716e-05,
+      "loss": 52.7801,
+      "step": 689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997509278783487e-05,
+      "loss": 57.2091,
+      "step": 690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997479718069522e-05,
+      "loss": 47.1415,
+      "step": 691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997449983014849e-05,
+      "loss": 0.7372,
+      "step": 692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997420073620509e-05,
+      "loss": 0.8105,
+      "step": 693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997389989887541e-05,
+      "loss": 0.7464,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997359731816998e-05,
+      "loss": 0.7455,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997329299409935e-05,
+      "loss": 61.3629,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997298692667412e-05,
+      "loss": 56.6886,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997267911590497e-05,
+      "loss": 52.5793,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997236956180265e-05,
+      "loss": 50.8993,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997205826437797e-05,
+      "loss": 50.3085,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997174522364177e-05,
+      "loss": 0.7759,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997143043960496e-05,
+      "loss": 55.4385,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997111391227856e-05,
+      "loss": 59.935,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997079564167358e-05,
+      "loss": 0.7342,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997047562780112e-05,
+      "loss": 55.5031,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997015387067237e-05,
+      "loss": 50.3746,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996983037029853e-05,
+      "loss": 0.7519,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99695051266909e-05,
+      "loss": 52.3369,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996917813986082e-05,
+      "loss": 0.7788,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99688494098197e-05,
+      "loss": 51.2265,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996851893657901e-05,
+      "loss": 54.5366,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996818672015029e-05,
+      "loss": 0.7515,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99678527605451e-05,
+      "loss": 0.7514,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996751705777512e-05,
+      "loss": 0.8181,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996717961185203e-05,
+      "loss": 52.1796,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996684042278763e-05,
+      "loss": 51.1849,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996649949059375e-05,
+      "loss": 0.798,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996615681528227e-05,
+      "loss": 49.9501,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996581239686516e-05,
+      "loss": 48.0634,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996546623535441e-05,
+      "loss": 49.7409,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996511833076212e-05,
+      "loss": 52.1311,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99647686831004e-05,
+      "loss": 48.7004,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996441729238151e-05,
+      "loss": 53.0331,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996406415861763e-05,
+      "loss": 0.7941,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996370928182113e-05,
+      "loss": 0.7408,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996335266200437e-05,
+      "loss": 0.7238,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996299429917979e-05,
+      "loss": 48.2118,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996263419335991e-05,
+      "loss": 0.8319,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996227234455728e-05,
+      "loss": 0.8474,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996190875278452e-05,
+      "loss": 52.4579,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99615434180543e-05,
+      "loss": 49.4804,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99611763403794e-05,
+      "loss": 50.3111,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996080751977262e-05,
+      "loss": 45.384,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996043695624679e-05,
+      "loss": 0.7903,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996006464981489e-05,
+      "loss": 48.2078,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995969060048986e-05,
+      "loss": 0.7957,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995931480828477e-05,
+      "loss": 52.5893,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995893727321274e-05,
+      "loss": 0.7781,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995855799528693e-05,
+      "loss": 48.7344,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995817697452056e-05,
+      "loss": 49.8127,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995779421092694e-05,
+      "loss": 47.4656,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995740970451941e-05,
+      "loss": 0.7671,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99570234553114e-05,
+      "loss": 49.7953,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995663546331637e-05,
+      "loss": 51.3356,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995624572854786e-05,
+      "loss": 0.7657,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995585425101948e-05,
+      "loss": 0.8222,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995546103074486e-05,
+      "loss": 0.7716,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995506606773772e-05,
+      "loss": 0.7452,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995466936201188e-05,
+      "loss": 52.9146,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995427091358113e-05,
+      "loss": 52.7948,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995387072245938e-05,
+      "loss": 47.0175,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995346878866062e-05,
+      "loss": 46.8758,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995306511219885e-05,
+      "loss": 51.6157,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995265969308814e-05,
+      "loss": 50.3059,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995225253134266e-05,
+      "loss": 50.4113,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995184362697659e-05,
+      "loss": 43.1886,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995143298000422e-05,
+      "loss": 47.6453,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995102059043985e-05,
+      "loss": 48.4504,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99506064582979e-05,
+      "loss": 0.7967,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995019058359279e-05,
+      "loss": 0.8312,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994977296633902e-05,
+      "loss": 50.7135,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994935360655117e-05,
+      "loss": 0.7972,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994893250424389e-05,
+      "loss": 49.7,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994850965943186e-05,
+      "loss": 54.7795,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99480850721298e-05,
+      "loss": 42.4584,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994765874235257e-05,
+      "loss": 0.7856,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9947230670115e-05,
+      "loss": 0.761,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994680085543205e-05,
+      "loss": 51.0541,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994636929831873e-05,
+      "loss": 49.8013,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994593599879004e-05,
+      "loss": 0.7326,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994550095686115e-05,
+      "loss": 50.0045,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99450641725472e-05,
+      "loss": 47.4688,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994462564586345e-05,
+      "loss": 0.7263,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99441853768252e-05,
+      "loss": 48.4119,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994374336544779e-05,
+      "loss": 43.8159,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994329961174666e-05,
+      "loss": 47.5163,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994285411573727e-05,
+      "loss": 48.7932,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994240687743519e-05,
+      "loss": 0.7882,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994195789685599e-05,
+      "loss": 0.7866,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994150717401533e-05,
+      "loss": 0.7579,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994105470892898e-05,
+      "loss": 0.7801,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994060050161269e-05,
+      "loss": 0.7915,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994014455208232e-05,
+      "loss": 50.4718,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993968686035376e-05,
+      "loss": 46.7655,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993922742644297e-05,
+      "loss": 48.141,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9938766250366e-05,
+      "loss": 44.8427,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993830333213894e-05,
+      "loss": 41.6916,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993783867177793e-05,
+      "loss": 0.8203,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993737226929918e-05,
+      "loss": 0.7388,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993690412471894e-05,
+      "loss": 48.3937,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993643423805358e-05,
+      "loss": 50.8662,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993596260931948e-05,
+      "loss": 46.743,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993548923853309e-05,
+      "loss": 44.8218,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993501412571091e-05,
+      "loss": 0.7528,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993453727086953e-05,
+      "loss": 0.7672,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993405867402558e-05,
+      "loss": 0.8114,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993357833519575e-05,
+      "loss": 0.7664,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993309625439682e-05,
+      "loss": 48.8514,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993261243164558e-05,
+      "loss": 52.7252,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993212686695893e-05,
+      "loss": 45.9095,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99316395603538e-05,
+      "loss": 43.7492,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993115051184718e-05,
+      "loss": 0.7253,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993065972145616e-05,
+      "loss": 0.8108,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993016718919783e-05,
+      "loss": 0.726,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992967291508938e-05,
+      "loss": 52.2591,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992917689914808e-05,
+      "loss": 48.0901,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992867914139119e-05,
+      "loss": 46.3175,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992817964183611e-05,
+      "loss": 0.7415,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992767840050025e-05,
+      "loss": 0.7729,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99271754174011e-05,
+      "loss": 48.6891,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992667069255619e-05,
+      "loss": 0.7697,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992616422598316e-05,
+      "loss": 44.9471,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992565601769965e-05,
+      "loss": 47.9235,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992514606772342e-05,
+      "loss": 43.7468,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992463437607222e-05,
+      "loss": 45.1439,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992412094276393e-05,
+      "loss": 0.7474,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992360576781645e-05,
+      "loss": 0.702,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992308885124777e-05,
+      "loss": 48.7061,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99225701930759e-05,
+      "loss": 0.7617,
+      "step": 819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992204979331893e-05,
+      "loss": 0.7181,
+      "step": 820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992152765199504e-05,
+      "loss": 0.7061,
+      "step": 821
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992100376912242e-05,
+      "loss": 46.147,
+      "step": 822
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992047814471937e-05,
+      "loss": 0.8042,
+      "step": 823
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99199507788042e-05,
+      "loss": 47.3979,
+      "step": 824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991942167139533e-05,
+      "loss": 45.9539,
+      "step": 825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99188908225112e-05,
+      "loss": 40.5623,
+      "step": 826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991835823217036e-05,
+      "loss": 45.8353,
+      "step": 827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991782390039136e-05,
+      "loss": 0.7444,
+      "step": 828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991728782719285e-05,
+      "loss": 46.2001,
+      "step": 829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991675001259353e-05,
+      "loss": 45.9401,
+      "step": 830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991621045661216e-05,
+      "loss": 46.237,
+      "step": 831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991566915926758e-05,
+      "loss": 41.4024,
+      "step": 832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991512612057864e-05,
+      "loss": 0.833,
+      "step": 833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99145813405643e-05,
+      "loss": 44.1805,
+      "step": 834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991403481924359e-05,
+      "loss": 44.0125,
+      "step": 835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991348655663555e-05,
+      "loss": 42.7461,
+      "step": 836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991293655275931e-05,
+      "loss": 0.8047,
+      "step": 837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991238480763406e-05,
+      "loss": 40.666,
+      "step": 838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991183132127904e-05,
+      "loss": 0.7954,
+      "step": 839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991127609371356e-05,
+      "loss": 0.8454,
+      "step": 840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991071912495701e-05,
+      "loss": 44.4041,
+      "step": 841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991016041502879e-05,
+      "loss": 0.7023,
+      "step": 842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990959996394842e-05,
+      "loss": 0.7626,
+      "step": 843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990903777173544e-05,
+      "loss": 42.4345,
+      "step": 844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990847383840945e-05,
+      "loss": 44.6955,
+      "step": 845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990790816399014e-05,
+      "loss": 40.9244,
+      "step": 846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990734074849723e-05,
+      "loss": 0.7577,
+      "step": 847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990677159195052e-05,
+      "loss": 44.3864,
+      "step": 848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990620069436987e-05,
+      "loss": 45.2435,
+      "step": 849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99056280557752e-05,
+      "loss": 39.6284,
+      "step": 850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990505367618647e-05,
+      "loss": 44.3855,
+      "step": 851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990447755562375e-05,
+      "loss": 44.1331,
+      "step": 852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990389969410709e-05,
+      "loss": 44.9319,
+      "step": 853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990332009165671e-05,
+      "loss": 0.7863,
+      "step": 854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990273874829277e-05,
+      "loss": 0.7203,
+      "step": 855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990215566403558e-05,
+      "loss": 41.4924,
+      "step": 856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990157083890547e-05,
+      "loss": 40.2055,
+      "step": 857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990098427292287e-05,
+      "loss": 44.2364,
+      "step": 858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990039596610821e-05,
+      "loss": 41.1213,
+      "step": 859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989980591848201e-05,
+      "loss": 44.8847,
+      "step": 860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989921413006489e-05,
+      "loss": 42.736,
+      "step": 861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989862060087746e-05,
+      "loss": 0.7793,
+      "step": 862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989802533094045e-05,
+      "loss": 40.7473,
+      "step": 863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989742832027464e-05,
+      "loss": 0.7818,
+      "step": 864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98968295689008e-05,
+      "loss": 45.0792,
+      "step": 865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989622907683989e-05,
+      "loss": 0.7786,
+      "step": 866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98956268441128e-05,
+      "loss": 0.7692,
+      "step": 867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989502287074056e-05,
+      "loss": 0.74,
+      "step": 868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989441715674422e-05,
+      "loss": 0.7994,
+      "step": 869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989380970214495e-05,
+      "loss": 47.0048,
+      "step": 870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989320050696392e-05,
+      "loss": 47.0812,
+      "step": 871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989258957122238e-05,
+      "loss": 0.8012,
+      "step": 872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989197689494166e-05,
+      "loss": 0.7624,
+      "step": 873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989136247814312e-05,
+      "loss": 41.6935,
+      "step": 874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989074632084819e-05,
+      "loss": 39.433,
+      "step": 875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989012842307837e-05,
+      "loss": 41.7339,
+      "step": 876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988950878485522e-05,
+      "loss": 47.3922,
+      "step": 877
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988888740620035e-05,
+      "loss": 42.5688,
+      "step": 878
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988826428713545e-05,
+      "loss": 0.7618,
+      "step": 879
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988763942768224e-05,
+      "loss": 0.744,
+      "step": 880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988701282786253e-05,
+      "loss": 0.7827,
+      "step": 881
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988638448769818e-05,
+      "loss": 0.765,
+      "step": 882
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988575440721108e-05,
+      "loss": 0.7421,
+      "step": 883
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988512258642327e-05,
+      "loss": 0.7665,
+      "step": 884
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988448902535674e-05,
+      "loss": 45.9746,
+      "step": 885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988385372403363e-05,
+      "loss": 48.3357,
+      "step": 886
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988321668247607e-05,
+      "loss": 48.3038,
+      "step": 887
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98825779007063e-05,
+      "loss": 41.2728,
+      "step": 888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98819373787466e-05,
+      "loss": 42.1394,
+      "step": 889
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988129511661932e-05,
+      "loss": 43.8003,
+      "step": 890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988065111434687e-05,
+      "loss": 0.8144,
+      "step": 891
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98800053719517e-05,
+      "loss": 0.7747,
+      "step": 892
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987935788945635e-05,
+      "loss": 0.7344,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98787086668834e-05,
+      "loss": 0.7713,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987805770425551e-05,
+      "loss": 0.7513,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987740500159536e-05,
+      "loss": 47.4987,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987675055892578e-05,
+      "loss": 0.8129,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987609437626955e-05,
+      "loss": 41.6031,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987543645364956e-05,
+      "loss": 0.7866,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987477679108878e-05,
+      "loss": 41.7895,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987411538861022e-05,
+      "loss": 42.6241,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987345224623697e-05,
+      "loss": 37.9824,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987278736399213e-05,
+      "loss": 0.7884,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987212074189892e-05,
+      "loss": 48.9618,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98714523799806e-05,
+      "loss": 43.1474,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987078227826044e-05,
+      "loss": 46.0163,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987011043676187e-05,
+      "loss": 0.7027,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986943685550831e-05,
+      "loss": 42.4935,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986876153452325e-05,
+      "loss": 44.8715,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986808447383027e-05,
+      "loss": 38.038,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986740567345296e-05,
+      "loss": 37.2348,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986672513341503e-05,
+      "loss": 0.7623,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98660428537402e-05,
+      "loss": 43.546,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986535883445228e-05,
+      "loss": 45.9974,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986467307557514e-05,
+      "loss": 41.1308,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986398557713269e-05,
+      "loss": 40.75,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98632963391489e-05,
+      "loss": 41.2052,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986260536164787e-05,
+      "loss": 40.3087,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986191264465363e-05,
+      "loss": 42.1358,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986121818819041e-05,
+      "loss": 37.5765,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98605219922824e-05,
+      "loss": 40.8062,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985982405695391e-05,
+      "loss": 42.2767,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985912438222927e-05,
+      "loss": 38.9871,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985842296813288e-05,
+      "loss": 41.2452,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985771981468925e-05,
+      "loss": 0.7456,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985701492192287e-05,
+      "loss": 40.1287,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985630828985835e-05,
+      "loss": 0.7466,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985559991852034e-05,
+      "loss": 40.0467,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985488980793354e-05,
+      "loss": 39.2552,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985417795812274e-05,
+      "loss": 38.9202,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985346436911275e-05,
+      "loss": 0.8101,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985274904092848e-05,
+      "loss": 38.4846,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985203197359489e-05,
+      "loss": 0.7459,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9851313167137e-05,
+      "loss": 39.012,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985059262157986e-05,
+      "loss": 0.7563,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984987033694864e-05,
+      "loss": 0.7926,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984914631326849e-05,
+      "loss": 0.763,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984842055056472e-05,
+      "loss": 36.629,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984769304886261e-05,
+      "loss": 39.0825,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984696380818758e-05,
+      "loss": 36.9263,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984623282856503e-05,
+      "loss": 39.3563,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984550011002047e-05,
+      "loss": 0.7647,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984476565257948e-05,
+      "loss": 38.6829,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984402945626767e-05,
+      "loss": 39.9685,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984329152111072e-05,
+      "loss": 0.7529,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984255184713435e-05,
+      "loss": 37.2471,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984181043436443e-05,
+      "loss": 0.7142,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984106728282677e-05,
+      "loss": 39.8536,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98403223925473e-05,
+      "loss": 0.7589,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983957576355203e-05,
+      "loss": 41.3238,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983882739586699e-05,
+      "loss": 40.8187,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983807728951828e-05,
+      "loss": 0.796,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983732544453208e-05,
+      "loss": 41.7287,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983657186093462e-05,
+      "loss": 38.8571,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98358165387522e-05,
+      "loss": 40.6346,
+      "step": 955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983505947801115e-05,
+      "loss": 37.3306,
+      "step": 956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983430067873787e-05,
+      "loss": 38.8975,
+      "step": 957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983354014095887e-05,
+      "loss": 35.9061,
+      "step": 958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983277786470065e-05,
+      "loss": 0.7644,
+      "step": 959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98320138499898e-05,
+      "loss": 0.7597,
+      "step": 960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9831248096853e-05,
+      "loss": 36.419,
+      "step": 961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983048060531695e-05,
+      "loss": 43.5963,
+      "step": 962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982971137540842e-05,
+      "loss": 35.8552,
+      "step": 963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982894040715424e-05,
+      "loss": 38.9692,
+      "step": 964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982816770058135e-05,
+      "loss": 38.7534,
+      "step": 965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982739325571664e-05,
+      "loss": 0.7906,
+      "step": 966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982661707258716e-05,
+      "loss": 39.2613,
+      "step": 967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982583915121999e-05,
+      "loss": 38.3263,
+      "step": 968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982505949164227e-05,
+      "loss": 38.3248,
+      "step": 969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98242780938812e-05,
+      "loss": 38.0472,
+      "step": 970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982349495796401e-05,
+      "loss": 0.8325,
+      "step": 971
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982271008391806e-05,
+      "loss": 38.7339,
+      "step": 972
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982192347177071e-05,
+      "loss": 0.7627,
+      "step": 973
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98211351215494e-05,
+      "loss": 0.7631,
+      "step": 974
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982034503328165e-05,
+      "loss": 38.4905,
+      "step": 975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981955320699501e-05,
+      "loss": 38.0299,
+      "step": 976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98187596427171e-05,
+      "loss": 0.7973,
+      "step": 977
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98179643404756e-05,
+      "loss": 37.0636,
+      "step": 978
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981716730029827e-05,
+      "loss": 36.1205,
+      "step": 979
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981636852221292e-05,
+      "loss": 0.7122,
+      "step": 980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981556800624739e-05,
+      "loss": 36.603,
+      "step": 981
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981476575242962e-05,
+      "loss": 38.56,
+      "step": 982
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981396176078762e-05,
+      "loss": 0.7775,
+      "step": 983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981315603134938e-05,
+      "loss": 39.5083,
+      "step": 984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981234856414307e-05,
+      "loss": 36.5258,
+      "step": 985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981153935919683e-05,
+      "loss": 0.7105,
+      "step": 986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981072841653889e-05,
+      "loss": 0.8001,
+      "step": 987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980991573619753e-05,
+      "loss": 39.3335,
+      "step": 988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980910131820114e-05,
+      "loss": 0.7813,
+      "step": 989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98082851625781e-05,
+      "loss": 38.6464,
+      "step": 990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980746726935688e-05,
+      "loss": 37.0383,
+      "step": 991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980664763856603e-05,
+      "loss": 0.8406,
+      "step": 992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980582627023412e-05,
+      "loss": 35.2625,
+      "step": 993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980500316438982e-05,
+      "loss": 0.7217,
+      "step": 994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980417832106185e-05,
+      "loss": 0.7308,
+      "step": 995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980335174027897e-05,
+      "loss": 36.0719,
+      "step": 996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980252342207004e-05,
+      "loss": 35.0012,
+      "step": 997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980169336646392e-05,
+      "loss": 0.7866,
+      "step": 998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98008615734896e-05,
+      "loss": 33.6224,
+      "step": 999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98000280431761e-05,
+      "loss": 0.7909,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979919277555246e-05,
+      "loss": 38.0301,
+      "step": 1001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979835577064785e-05,
+      "loss": 0.8233,
+      "step": 1002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979751702849147e-05,
+      "loss": 0.7412,
+      "step": 1003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979667654911257e-05,
+      "loss": 37.1845,
+      "step": 1004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979583433254048e-05,
+      "loss": 34.3108,
+      "step": 1005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979499037880455e-05,
+      "loss": 0.7506,
+      "step": 1006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979414468793427e-05,
+      "loss": 36.0232,
+      "step": 1007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979329725995909e-05,
+      "loss": 0.8343,
+      "step": 1008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979244809490863e-05,
+      "loss": 0.812,
+      "step": 1009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979159719281247e-05,
+      "loss": 0.7501,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979074455370031e-05,
+      "loss": 0.7441,
+      "step": 1011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978989017760192e-05,
+      "loss": 0.7579,
+      "step": 1012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978903406454706e-05,
+      "loss": 0.7306,
+      "step": 1013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978817621456562e-05,
+      "loss": 39.0051,
+      "step": 1014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978731662768753e-05,
+      "loss": 34.1403,
+      "step": 1015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978645530394276e-05,
+      "loss": 0.7713,
+      "step": 1016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978559224336138e-05,
+      "loss": 0.7061,
+      "step": 1017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978472744597347e-05,
+      "loss": 39.9444,
+      "step": 1018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978386091180923e-05,
+      "loss": 37.8633,
+      "step": 1019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978299264089888e-05,
+      "loss": 35.2702,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978212263327269e-05,
+      "loss": 33.8156,
+      "step": 1021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978125088896103e-05,
+      "loss": 0.7909,
+      "step": 1022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978037740799431e-05,
+      "loss": 0.7834,
+      "step": 1023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977950219040301e-05,
+      "loss": 0.8465,
+      "step": 1024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977862523621765e-05,
+      "loss": 0.7876,
+      "step": 1025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977774654546883e-05,
+      "loss": 0.8057,
+      "step": 1026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977686611818718e-05,
+      "loss": 61.3097,
+      "step": 1027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977598395440346e-05,
+      "loss": 65.2611,
+      "step": 1028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97751000541484e-05,
+      "loss": 52.293,
+      "step": 1029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977421441745287e-05,
+      "loss": 0.7084,
+      "step": 1030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977332704434773e-05,
+      "loss": 40.2907,
+      "step": 1031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977243793486397e-05,
+      "loss": 39.3646,
+      "step": 1032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97715470890326e-05,
+      "loss": 39.1512,
+      "step": 1033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977065450688467e-05,
+      "loss": 41.1173,
+      "step": 1034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976976018845137e-05,
+      "loss": 0.7139,
+      "step": 1035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976886413376386e-05,
+      "loss": 38.212,
+      "step": 1036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976796634285342e-05,
+      "loss": 0.7114,
+      "step": 1037
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976706681575135e-05,
+      "loss": 34.5505,
+      "step": 1038
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976616555248901e-05,
+      "loss": 37.9683,
+      "step": 1039
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97652625530979e-05,
+      "loss": 38.4048,
+      "step": 1040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97643578176095e-05,
+      "loss": 37.2291,
+      "step": 1041
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976345134605537e-05,
+      "loss": 36.3072,
+      "step": 1042
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97625431384671e-05,
+      "loss": 34.21,
+      "step": 1043
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976163319487641e-05,
+      "loss": 38.1845,
+      "step": 1044
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976072151531506e-05,
+      "loss": 0.7436,
+      "step": 1045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975980809981481e-05,
+      "loss": 38.9623,
+      "step": 1046
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975889294840755e-05,
+      "loss": 36.929,
+      "step": 1047
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975797606112521e-05,
+      "loss": 38.917,
+      "step": 1048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975705743799975e-05,
+      "loss": 33.3681,
+      "step": 1049
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975613707906324e-05,
+      "loss": 37.368,
+      "step": 1050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975521498434778e-05,
+      "loss": 0.7773,
+      "step": 1051
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975429115388555e-05,
+      "loss": 0.7708,
+      "step": 1052
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975336558770875e-05,
+      "loss": 0.7742,
+      "step": 1053
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975243828584971e-05,
+      "loss": 36.7888,
+      "step": 1054
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975150924834076e-05,
+      "loss": 39.229,
+      "step": 1055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975057847521429e-05,
+      "loss": 34.9099,
+      "step": 1056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974964596650279e-05,
+      "loss": 37.493,
+      "step": 1057
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974871172223878e-05,
+      "loss": 0.7823,
+      "step": 1058
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974777574245488e-05,
+      "loss": 0.777,
+      "step": 1059
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974683802718372e-05,
+      "loss": 0.7408,
+      "step": 1060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974589857645802e-05,
+      "loss": 36.1275,
+      "step": 1061
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974495739031053e-05,
+      "loss": 0.777,
+      "step": 1062
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974401446877413e-05,
+      "loss": 39.5786,
+      "step": 1063
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974306981188167e-05,
+      "loss": 37.5633,
+      "step": 1064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974212341966612e-05,
+      "loss": 33.8944,
+      "step": 1065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97411752921605e-05,
+      "loss": 0.79,
+      "step": 1066
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974022542939789e-05,
+      "loss": 37.1766,
+      "step": 1067
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973927383141143e-05,
+      "loss": 38.8413,
+      "step": 1068
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973832049823429e-05,
+      "loss": 37.9531,
+      "step": 1069
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973736542989976e-05,
+      "loss": 0.719,
+      "step": 1070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973640862644114e-05,
+      "loss": 36.5471,
+      "step": 1071
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973545008789181e-05,
+      "loss": 37.0036,
+      "step": 1072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973448981428522e-05,
+      "loss": 36.3739,
+      "step": 1073
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973352780565486e-05,
+      "loss": 33.7922,
+      "step": 1074
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973256406203429e-05,
+      "loss": 34.2224,
+      "step": 1075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973159858345713e-05,
+      "loss": 37.7252,
+      "step": 1076
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973063136995708e-05,
+      "loss": 0.7768,
+      "step": 1077
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972966242156786e-05,
+      "loss": 39.4132,
+      "step": 1078
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972869173832327e-05,
+      "loss": 35.4671,
+      "step": 1079
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97277193202572e-05,
+      "loss": 36.511,
+      "step": 1080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972674516740353e-05,
+      "loss": 37.1352,
+      "step": 1081
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972576927979629e-05,
+      "loss": 34.9714,
+      "step": 1082
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97247916574695e-05,
+      "loss": 35.0376,
+      "step": 1083
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972381230045727e-05,
+      "loss": 0.7793,
+      "step": 1084
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972283120879376e-05,
+      "loss": 37.0188,
+      "step": 1085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972184838251321e-05,
+      "loss": 38.5488,
+      "step": 1086
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97208638216499e-05,
+      "loss": 35.8355,
+      "step": 1087
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971987752623817e-05,
+      "loss": 40.5661,
+      "step": 1088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971888949631243e-05,
+      "loss": 0.7402,
+      "step": 1089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971789973190715e-05,
+      "loss": 31.7359,
+      "step": 1090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971690823305686e-05,
+      "loss": 0.7678,
+      "step": 1091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971591499979616e-05,
+      "loss": 36.7627,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971492003215967e-05,
+      "loss": 0.7212,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971392333018214e-05,
+      "loss": 39.6192,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971292489389831e-05,
+      "loss": 0.7873,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9711924723343e-05,
+      "loss": 0.7896,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971092281855115e-05,
+      "loss": 36.5608,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970991917955768e-05,
+      "loss": 39.6488,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970891380639759e-05,
+      "loss": 0.7771,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970790669910599e-05,
+      "loss": 0.8105,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970689785771798e-05,
+      "loss": 0.6936,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970588728226877e-05,
+      "loss": 36.5358,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970487497279362e-05,
+      "loss": 37.4523,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970386092932785e-05,
+      "loss": 33.9085,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97028451519068e-05,
+      "loss": 36.0548,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970182764056593e-05,
+      "loss": 34.5384,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970080839534074e-05,
+      "loss": 0.8123,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969978741626679e-05,
+      "loss": 0.7778,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969876470337969e-05,
+      "loss": 35.5599,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96977402567151e-05,
+      "loss": 37.736,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96967140763088e-05,
+      "loss": 35.623,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969568616219656e-05,
+      "loss": 0.793,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969465651441424e-05,
+      "loss": 35.0513,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969362513299779e-05,
+      "loss": 33.378,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969259201798313e-05,
+      "loss": 0.7366,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969155716940636e-05,
+      "loss": 38.745,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969052058730355e-05,
+      "loss": 37.175,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968948227171088e-05,
+      "loss": 0.749,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968844222266456e-05,
+      "loss": 34.1751,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968740044020086e-05,
+      "loss": 35.7992,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968635692435617e-05,
+      "loss": 0.7031,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968531167516684e-05,
+      "loss": 36.8526,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968426469266936e-05,
+      "loss": 35.3369,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968321597690025e-05,
+      "loss": 34.865,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96821655278961e-05,
+      "loss": 33.6427,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968111334569357e-05,
+      "loss": 32.9011,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968005943032935e-05,
+      "loss": 0.7425,
+      "step": 1127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967900378184019e-05,
+      "loss": 0.7364,
+      "step": 1128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967794640026293e-05,
+      "loss": 38.5608,
+      "step": 1129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967688728563446e-05,
+      "loss": 0.7291,
+      "step": 1130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967582643799174e-05,
+      "loss": 35.3991,
+      "step": 1131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967476385737176e-05,
+      "loss": 34.6453,
+      "step": 1132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96736995438116e-05,
+      "loss": 31.3299,
+      "step": 1133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967263349734837e-05,
+      "loss": 33.7752,
+      "step": 1134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96715657180193e-05,
+      "loss": 0.7996,
+      "step": 1135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96704962058616e-05,
+      "loss": 0.7516,
+      "step": 1136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96694249609126e-05,
+      "loss": 34.2218,
+      "step": 1137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966835198320966e-05,
+      "loss": 38.0752,
+      "step": 1138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966727727279022e-05,
+      "loss": 0.7741,
+      "step": 1139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966620082969178e-05,
+      "loss": 30.7811,
+      "step": 1140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966512265395188e-05,
+      "loss": 37.8192,
+      "step": 1141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966404274560813e-05,
+      "loss": 34.3281,
+      "step": 1142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966296110469821e-05,
+      "loss": 33.2368,
+      "step": 1143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966187773125987e-05,
+      "loss": 0.806,
+      "step": 1144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966079262533087e-05,
+      "loss": 38.1086,
+      "step": 1145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965970578694908e-05,
+      "loss": 32.3057,
+      "step": 1146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965861721615244e-05,
+      "loss": 35.1885,
+      "step": 1147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965752691297888e-05,
+      "loss": 0.7205,
+      "step": 1148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965643487746647e-05,
+      "loss": 34.7598,
+      "step": 1149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96553411096533e-05,
+      "loss": 36.6135,
+      "step": 1150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96542456095775e-05,
+      "loss": 0.7834,
+      "step": 1151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965314837727733e-05,
+      "loss": 36.2269,
+      "step": 1152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965204941279105e-05,
+      "loss": 33.2311,
+      "step": 1153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9650948716157e-05,
+      "loss": 33.1762,
+      "step": 1154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964984628741354e-05,
+      "loss": 32.3484,
+      "step": 1155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964874212659919e-05,
+      "loss": 32.8516,
+      "step": 1156
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964763623375244e-05,
+      "loss": 33.6558,
+      "step": 1157
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964652860891187e-05,
+      "loss": 34.6232,
+      "step": 1158
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964541925211612e-05,
+      "loss": 33.9009,
+      "step": 1159
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96443081634039e-05,
+      "loss": 31.0008,
+      "step": 1160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964319534281397e-05,
+      "loss": 33.2357,
+      "step": 1161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964208079038513e-05,
+      "loss": 0.7911,
+      "step": 1162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96409645061563e-05,
+      "loss": 33.2195,
+      "step": 1163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963984649016637e-05,
+      "loss": 34.1875,
+      "step": 1164
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963872674245441e-05,
+      "loss": 0.7839,
+      "step": 1165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963760526305942e-05,
+      "loss": 32.9812,
+      "step": 1166
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963648205202057e-05,
+      "loss": 31.3622,
+      "step": 1167
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9635357109377e-05,
+      "loss": 35.5625,
+      "step": 1168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963423043516799e-05,
+      "loss": 33.2915,
+      "step": 1169
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963310202943284e-05,
+      "loss": 31.8328,
+      "step": 1170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963197189221091e-05,
+      "loss": 0.7711,
+      "step": 1171
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963084002354162e-05,
+      "loss": 36.7669,
+      "step": 1172
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962970642346445e-05,
+      "loss": 0.7782,
+      "step": 1173
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962857109201897e-05,
+      "loss": 32.3001,
+      "step": 1174
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962743402924477e-05,
+      "loss": 35.0642,
+      "step": 1175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962629523518151e-05,
+      "loss": 0.716,
+      "step": 1176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962515470986895e-05,
+      "loss": 0.7986,
+      "step": 1177
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962401245334685e-05,
+      "loss": 31.8786,
+      "step": 1178
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962286846565508e-05,
+      "loss": 34.5701,
+      "step": 1179
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96217227468335e-05,
+      "loss": 35.4613,
+      "step": 1180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962057529692212e-05,
+      "loss": 31.8869,
+      "step": 1181
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9619426115961e-05,
+      "loss": 32.4353,
+      "step": 1182
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961827520399015e-05,
+      "loss": 0.7615,
+      "step": 1183
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961712256104977e-05,
+      "loss": 0.7231,
+      "step": 1184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961596818718009e-05,
+      "loss": 0.8279,
+      "step": 1185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961481208242134e-05,
+      "loss": 34.6512,
+      "step": 1186
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961365424681385e-05,
+      "loss": 35.7029,
+      "step": 1187
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961249468039807e-05,
+      "loss": 31.9616,
+      "step": 1188
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961133338321438e-05,
+      "loss": 30.5549,
+      "step": 1189
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961017035530334e-05,
+      "loss": 0.7308,
+      "step": 1190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96090055967055e-05,
+      "loss": 33.887,
+      "step": 1191
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960783910746149e-05,
+      "loss": 30.7291,
+      "step": 1192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960667088761202e-05,
+      "loss": 30.7366,
+      "step": 1193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960550093719784e-05,
+      "loss": 0.8317,
+      "step": 1194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960432925625977e-05,
+      "loss": 0.7381,
+      "step": 1195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960315584483866e-05,
+      "loss": 0.7647,
+      "step": 1196
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960198070297548e-05,
+      "loss": 32.7556,
+      "step": 1197
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96008038307112e-05,
+      "loss": 31.411,
+      "step": 1198
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959962522808691e-05,
+      "loss": 0.7361,
+      "step": 1199
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959844489514368e-05,
+      "loss": 33.5219,
+      "step": 1200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95972628319227e-05,
+      "loss": 33.7455,
+      "step": 1201
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959607903846521e-05,
+      "loss": 0.7103,
+      "step": 1202
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959489351481255e-05,
+      "loss": 0.7644,
+      "step": 1203
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959370626100601e-05,
+      "loss": 35.4873,
+      "step": 1204
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959251727708705e-05,
+      "loss": 30.9286,
+      "step": 1205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959132656309714e-05,
+      "loss": 0.7517,
+      "step": 1206
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959013411907781e-05,
+      "loss": 32.9335,
+      "step": 1207
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958893994507065e-05,
+      "loss": 33.976,
+      "step": 1208
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958774404111735e-05,
+      "loss": 34.6583,
+      "step": 1209
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958654640725963e-05,
+      "loss": 31.8389,
+      "step": 1210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958534704353924e-05,
+      "loss": 0.8187,
+      "step": 1211
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958414594999803e-05,
+      "loss": 31.7638,
+      "step": 1212
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958294312667793e-05,
+      "loss": 0.7671,
+      "step": 1213
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958173857362085e-05,
+      "loss": 33.6291,
+      "step": 1214
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958053229086887e-05,
+      "loss": 31.8604,
+      "step": 1215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957932427846402e-05,
+      "loss": 0.7896,
+      "step": 1216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957811453644847e-05,
+      "loss": 0.7885,
+      "step": 1217
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957690306486442e-05,
+      "loss": 33.2548,
+      "step": 1218
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957568986375414e-05,
+      "loss": 0.7595,
+      "step": 1219
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957447493315992e-05,
+      "loss": 0.7729,
+      "step": 1220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957325827312419e-05,
+      "loss": 0.7837,
+      "step": 1221
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957203988368937e-05,
+      "loss": 0.8078,
+      "step": 1222
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957081976489796e-05,
+      "loss": 34.87,
+      "step": 1223
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956959791679253e-05,
+      "loss": 0.7213,
+      "step": 1224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95683743394157e-05,
+      "loss": 35.688,
+      "step": 1225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956714903281017e-05,
+      "loss": 0.7364,
+      "step": 1226
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956592199701868e-05,
+      "loss": 0.7603,
+      "step": 1227
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956469323208403e-05,
+      "loss": 30.1385,
+      "step": 1228
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956346273804909e-05,
+      "loss": 35.0577,
+      "step": 1229
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956223051495678e-05,
+      "loss": 35.8497,
+      "step": 1230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95609965628501e-05,
+      "loss": 0.7441,
+      "step": 1231
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955976088177208e-05,
+      "loss": 34.263,
+      "step": 1232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955852347176584e-05,
+      "loss": 0.8446,
+      "step": 1233
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955728433287456e-05,
+      "loss": 0.8132,
+      "step": 1234
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955604346514145e-05,
+      "loss": 32.4682,
+      "step": 1235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95548008686098e-05,
+      "loss": 31.4907,
+      "step": 1236
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955355654332294e-05,
+      "loss": 35.2979,
+      "step": 1237
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955231048932434e-05,
+      "loss": 32.0391,
+      "step": 1238
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955106270665741e-05,
+      "loss": 33.1883,
+      "step": 1239
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954981319536569e-05,
+      "loss": 31.9835,
+      "step": 1240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95485619554928e-05,
+      "loss": 33.5871,
+      "step": 1241
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954730898708237e-05,
+      "loss": 0.7731,
+      "step": 1242
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954605429017809e-05,
+      "loss": 0.774,
+      "step": 1243
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954479786482378e-05,
+      "loss": 36.2369,
+      "step": 1244
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954353971106323e-05,
+      "loss": 0.7539,
+      "step": 1245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954227982894034e-05,
+      "loss": 0.7632,
+      "step": 1246
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954101821849908e-05,
+      "loss": 0.7715,
+      "step": 1247
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953975487978345e-05,
+      "loss": 32.2305,
+      "step": 1248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953848981283751e-05,
+      "loss": 34.8954,
+      "step": 1249
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953722301770541e-05,
+      "loss": 34.8862,
+      "step": 1250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953595449443134e-05,
+      "loss": 0.7305,
+      "step": 1251
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953468424305954e-05,
+      "loss": 32.7877,
+      "step": 1252
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953341226363435e-05,
+      "loss": 0.7686,
+      "step": 1253
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953213855620012e-05,
+      "loss": 34.4383,
+      "step": 1254
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953086312080129e-05,
+      "loss": 0.7508,
+      "step": 1255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952958595748237e-05,
+      "loss": 0.7388,
+      "step": 1256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952830706628788e-05,
+      "loss": 0.769,
+      "step": 1257
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952702644726248e-05,
+      "loss": 32.4464,
+      "step": 1258
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952574410045082e-05,
+      "loss": 33.2146,
+      "step": 1259
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952446002589763e-05,
+      "loss": 0.7396,
+      "step": 1260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952317422364772e-05,
+      "loss": 34.9748,
+      "step": 1261
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952188669374593e-05,
+      "loss": 0.726,
+      "step": 1262
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95205974362372e-05,
+      "loss": 0.7816,
+      "step": 1263
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951930645116648e-05,
+      "loss": 0.7955,
+      "step": 1264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951801373857883e-05,
+      "loss": 30.11,
+      "step": 1265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951671929851933e-05,
+      "loss": 29.665,
+      "step": 1266
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951542313103315e-05,
+      "loss": 29.6824,
+      "step": 1267
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95141252361655e-05,
+      "loss": 0.7462,
+      "step": 1268
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951282561396167e-05,
+      "loss": 32.2875,
+      "step": 1269
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951152426446698e-05,
+      "loss": 27.524,
+      "step": 1270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951022118772684e-05,
+      "loss": 31.292,
+      "step": 1271
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95089163837867e-05,
+      "loss": 34.4362,
+      "step": 1272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950760985269209e-05,
+      "loss": 33.1613,
+      "step": 1273
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950630159448859e-05,
+      "loss": 0.6818,
+      "step": 1274
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950499160922183e-05,
+      "loss": 0.794,
+      "step": 1275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950367989693752e-05,
+      "loss": 0.7939,
+      "step": 1276
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95023664576814e-05,
+      "loss": 0.7565,
+      "step": 1277
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950105129149931e-05,
+      "loss": 30.8134,
+      "step": 1278
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949973439843713e-05,
+      "loss": 32.1406,
+      "step": 1279
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949841577854079e-05,
+      "loss": 0.7664,
+      "step": 1280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94970954318563e-05,
+      "loss": 32.0306,
+      "step": 1281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949577335842973e-05,
+      "loss": 0.7845,
+      "step": 1282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949444955830716e-05,
+      "loss": 0.8732,
+      "step": 1283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949312403153482e-05,
+      "loss": 0.7247,
+      "step": 1284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949179677815891e-05,
+      "loss": 0.8347,
+      "step": 1285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94904677982258e-05,
+      "loss": 34.3108,
+      "step": 1286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948913709178177e-05,
+      "loss": 0.7772,
+      "step": 1287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94878046588733e-05,
+      "loss": 33.0494,
+      "step": 1288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948647049954685e-05,
+      "loss": 0.6828,
+      "step": 1289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948513461384896e-05,
+      "loss": 32.6104,
+      "step": 1290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948379700182624e-05,
+      "loss": 32.4268,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948245766352536e-05,
+      "loss": 34.5091,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948111659899301e-05,
+      "loss": 34.9759,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947977380827603e-05,
+      "loss": 0.7391,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947842929142123e-05,
+      "loss": 0.7601,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947708304847552e-05,
+      "loss": 0.7489,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947573507948585e-05,
+      "loss": 0.7417,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947438538449927e-05,
+      "loss": 34.0326,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947303396356287e-05,
+      "loss": 35.4293,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947168081672375e-05,
+      "loss": 0.7524,
+      "step": 1300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.947032594402917e-05,
+      "loss": 0.7313,
+      "step": 1301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946896934552635e-05,
+      "loss": 0.7136,
+      "step": 1302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946761102126265e-05,
+      "loss": 31.5197,
+      "step": 1303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946625097128543e-05,
+      "loss": 35.8017,
+      "step": 1304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946488919564216e-05,
+      "loss": 34.5114,
+      "step": 1305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946352569438034e-05,
+      "loss": 32.1719,
+      "step": 1306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946216046754752e-05,
+      "loss": 33.8104,
+      "step": 1307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946079351519135e-05,
+      "loss": 32.3694,
+      "step": 1308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94594248373595e-05,
+      "loss": 31.8187,
+      "step": 1309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945805443409973e-05,
+      "loss": 32.9493,
+      "step": 1310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945668230545983e-05,
+      "loss": 29.1888,
+      "step": 1311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945530845148769e-05,
+      "loss": 0.7767,
+      "step": 1312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945393287223124e-05,
+      "loss": 0.6808,
+      "step": 1313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945255556773843e-05,
+      "loss": 0.7367,
+      "step": 1314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945117653805735e-05,
+      "loss": 0.7259,
+      "step": 1315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944979578323608e-05,
+      "loss": 0.7005,
+      "step": 1316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94484133033228e-05,
+      "loss": 38.5158,
+      "step": 1317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944702909836574e-05,
+      "loss": 35.7147,
+      "step": 1318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944564316841319e-05,
+      "loss": 32.5956,
+      "step": 1319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944425551351351e-05,
+      "loss": 34.3188,
+      "step": 1320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944286613371507e-05,
+      "loss": 31.1588,
+      "step": 1321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944147502906639e-05,
+      "loss": 33.2539,
+      "step": 1322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944008219961595e-05,
+      "loss": 36.057,
+      "step": 1323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943868764541239e-05,
+      "loss": 34.9231,
+      "step": 1324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943729136650433e-05,
+      "loss": 32.9369,
+      "step": 1325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94358933629405e-05,
+      "loss": 0.7202,
+      "step": 1326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943449363476963e-05,
+      "loss": 33.2934,
+      "step": 1327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943309218204059e-05,
+      "loss": 33.1534,
+      "step": 1328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943168900480226e-05,
+      "loss": 36.1792,
+      "step": 1329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943028410310358e-05,
+      "loss": 33.0298,
+      "step": 1330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942887747699357e-05,
+      "loss": 33.187,
+      "step": 1331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94274691265213e-05,
+      "loss": 0.7685,
+      "step": 1332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942605905173592e-05,
+      "loss": 30.9668,
+      "step": 1333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942464725268659e-05,
+      "loss": 30.8934,
+      "step": 1334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942323372942258e-05,
+      "loss": 33.0713,
+      "step": 1335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942181848199318e-05,
+      "loss": 0.8191,
+      "step": 1336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942040151044778e-05,
+      "loss": 34.9832,
+      "step": 1337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941898281483582e-05,
+      "loss": 37.7695,
+      "step": 1338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941756239520678e-05,
+      "loss": 33.8458,
+      "step": 1339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941614025161024e-05,
+      "loss": 31.101,
+      "step": 1340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941471638409577e-05,
+      "loss": 0.7453,
+      "step": 1341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941329079271305e-05,
+      "loss": 30.0624,
+      "step": 1342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941186347751184e-05,
+      "loss": 29.9807,
+      "step": 1343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94104344385419e-05,
+      "loss": 0.7891,
+      "step": 1344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94090036758531e-05,
+      "loss": 30.8794,
+      "step": 1345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940757118949537e-05,
+      "loss": 30.0098,
+      "step": 1346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940613697951865e-05,
+      "loss": 0.6976,
+      "step": 1347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9404701045973e-05,
+      "loss": 30.9407,
+      "step": 1348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940326338890849e-05,
+      "loss": 32.4642,
+      "step": 1349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94018240083753e-05,
+      "loss": 28.8362,
+      "step": 1350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940038290442361e-05,
+      "loss": 0.7588,
+      "step": 1351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939894007710374e-05,
+      "loss": 0.7191,
+      "step": 1352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9397495526466e-05,
+      "loss": 30.4913,
+      "step": 1353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939604925256077e-05,
+      "loss": 33.6085,
+      "step": 1354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939460125543851e-05,
+      "loss": 31.8808,
+      "step": 1355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939315153514975e-05,
+      "loss": 0.7463,
+      "step": 1356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939170009174506e-05,
+      "loss": 30.5306,
+      "step": 1357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939024692527505e-05,
+      "loss": 0.7765,
+      "step": 1358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938879203579046e-05,
+      "loss": 0.7581,
+      "step": 1359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938733542334202e-05,
+      "loss": 0.7403,
+      "step": 1360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938587708798053e-05,
+      "loss": 33.4127,
+      "step": 1361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938441702975689e-05,
+      "loss": 32.2939,
+      "step": 1362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938295524872203e-05,
+      "loss": 0.7238,
+      "step": 1363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938149174492693e-05,
+      "loss": 31.046,
+      "step": 1364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938002651842266e-05,
+      "loss": 32.8841,
+      "step": 1365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937855956926034e-05,
+      "loss": 32.0785,
+      "step": 1366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937709089749112e-05,
+      "loss": 0.7738,
+      "step": 1367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937562050316627e-05,
+      "loss": 33.1618,
+      "step": 1368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937414838633705e-05,
+      "loss": 33.726,
+      "step": 1369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937267454705486e-05,
+      "loss": 32.8496,
+      "step": 1370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937119898537106e-05,
+      "loss": 31.7068,
+      "step": 1371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936972170133718e-05,
+      "loss": 29.0642,
+      "step": 1372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936824269500473e-05,
+      "loss": 0.7933,
+      "step": 1373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936676196642531e-05,
+      "loss": 0.7356,
+      "step": 1374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936527951565056e-05,
+      "loss": 30.7471,
+      "step": 1375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936379534273223e-05,
+      "loss": 0.7047,
+      "step": 1376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936230944772207e-05,
+      "loss": 32.2134,
+      "step": 1377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936082183067193e-05,
+      "loss": 31.8743,
+      "step": 1378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93593324916337e-05,
+      "loss": 31.5227,
+      "step": 1379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935784143065931e-05,
+      "loss": 0.8141,
+      "step": 1380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935634864780083e-05,
+      "loss": 0.732,
+      "step": 1381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935485414311032e-05,
+      "loss": 35.6381,
+      "step": 1382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93533579166399e-05,
+      "loss": 35.1825,
+      "step": 1383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935185996844177e-05,
+      "loss": 32.3992,
+      "step": 1384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935036029856819e-05,
+      "loss": 0.7246,
+      "step": 1385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93488589070715e-05,
+      "loss": 0.7226,
+      "step": 1386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934735579400404e-05,
+      "loss": 0.7312,
+      "step": 1387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934585095941826e-05,
+      "loss": 0.7634,
+      "step": 1388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934434440336667e-05,
+      "loss": 0.7702,
+      "step": 1389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93428361259018e-05,
+      "loss": 40.6792,
+      "step": 1390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934132612707632e-05,
+      "loss": 0.7232,
+      "step": 1391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933981440694285e-05,
+      "loss": 35.6297,
+      "step": 1392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933830096555414e-05,
+      "loss": 0.7421,
+      "step": 1393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933678580296302e-05,
+      "loss": 0.8154,
+      "step": 1394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933526891922231e-05,
+      "loss": 35.0311,
+      "step": 1395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933375031438497e-05,
+      "loss": 34.6004,
+      "step": 1396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933222998850393e-05,
+      "loss": 0.7608,
+      "step": 1397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933070794163224e-05,
+      "loss": 35.2873,
+      "step": 1398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932918417382303e-05,
+      "loss": 0.8048,
+      "step": 1399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932765868512941e-05,
+      "loss": 0.7448,
+      "step": 1400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932613147560463e-05,
+      "loss": 34.3687,
+      "step": 1401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932460254530196e-05,
+      "loss": 0.7716,
+      "step": 1402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932307189427474e-05,
+      "loss": 34.5618,
+      "step": 1403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932153952257636e-05,
+      "loss": 32.1236,
+      "step": 1404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932000543026028e-05,
+      "loss": 28.6573,
+      "step": 1405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931846961738003e-05,
+      "loss": 0.6967,
+      "step": 1406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931693208398918e-05,
+      "loss": 0.75,
+      "step": 1407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931539283014135e-05,
+      "loss": 31.4699,
+      "step": 1408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931385185589029e-05,
+      "loss": 33.3726,
+      "step": 1409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931230916128968e-05,
+      "loss": 32.915,
+      "step": 1410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93107647463934e-05,
+      "loss": 0.7551,
+      "step": 1411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930921861125531e-05,
+      "loss": 33.0645,
+      "step": 1412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930767075592935e-05,
+      "loss": 32.4392,
+      "step": 1413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930612118046951e-05,
+      "loss": 35.2833,
+      "step": 1414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930456988492987e-05,
+      "loss": 33.3836,
+      "step": 1415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930301686936452e-05,
+      "loss": 29.4361,
+      "step": 1416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930146213382764e-05,
+      "loss": 0.7756,
+      "step": 1417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929990567837348e-05,
+      "loss": 34.9401,
+      "step": 1418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929834750305633e-05,
+      "loss": 0.822,
+      "step": 1419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929678760793057e-05,
+      "loss": 0.7487,
+      "step": 1420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929522599305058e-05,
+      "loss": 0.7835,
+      "step": 1421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929366265847089e-05,
+      "loss": 0.7906,
+      "step": 1422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929209760424599e-05,
+      "loss": 35.1459,
+      "step": 1423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92905308304305e-05,
+      "loss": 0.7447,
+      "step": 1424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928896233707907e-05,
+      "loss": 0.7181,
+      "step": 1425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928739212424643e-05,
+      "loss": 31.1925,
+      "step": 1426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928582019198733e-05,
+      "loss": 30.1713,
+      "step": 1427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928424654035665e-05,
+      "loss": 0.7858,
+      "step": 1428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928267116940925e-05,
+      "loss": 30.9191,
+      "step": 1429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928109407920009e-05,
+      "loss": 0.8012,
+      "step": 1430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927951526978423e-05,
+      "loss": 32.0978,
+      "step": 1431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927793474121671e-05,
+      "loss": 30.7737,
+      "step": 1432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927635249355269e-05,
+      "loss": 0.7282,
+      "step": 1433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927476852684734e-05,
+      "loss": 0.7601,
+      "step": 1434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927318284115595e-05,
+      "loss": 28.9591,
+      "step": 1435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92715954365338e-05,
+      "loss": 0.7661,
+      "step": 1436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927000631303629e-05,
+      "loss": 0.7835,
+      "step": 1437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926841547071886e-05,
+      "loss": 31.7991,
+      "step": 1438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926682290963702e-05,
+      "loss": 29.9203,
+      "step": 1439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926522862984629e-05,
+      "loss": 0.6856,
+      "step": 1440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926363263140233e-05,
+      "loss": 28.7442,
+      "step": 1441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926203491436076e-05,
+      "loss": 32.6134,
+      "step": 1442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92604354787774e-05,
+      "loss": 0.721,
+      "step": 1443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925883432470797e-05,
+      "loss": 31.9048,
+      "step": 1444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925723145220838e-05,
+      "loss": 30.2888,
+      "step": 1445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925562686133451e-05,
+      "loss": 0.6606,
+      "step": 1446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925402055214237e-05,
+      "loss": 33.7318,
+      "step": 1447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925241252468797e-05,
+      "loss": 30.7776,
+      "step": 1448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925080277902743e-05,
+      "loss": 29.559,
+      "step": 1449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924919131521688e-05,
+      "loss": 31.3869,
+      "step": 1450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924757813331255e-05,
+      "loss": 0.7419,
+      "step": 1451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924596323337074e-05,
+      "loss": 30.5053,
+      "step": 1452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924434661544776e-05,
+      "loss": 29.6631,
+      "step": 1453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924272827960002e-05,
+      "loss": 0.7748,
+      "step": 1454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924110822588395e-05,
+      "loss": 0.6941,
+      "step": 1455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92394864543561e-05,
+      "loss": 34.0309,
+      "step": 1456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923786296507303e-05,
+      "loss": 0.7487,
+      "step": 1457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923623775809138e-05,
+      "loss": 29.7698,
+      "step": 1458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923461083346785e-05,
+      "loss": 29.7378,
+      "step": 1459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923298219125921e-05,
+      "loss": 32.217,
+      "step": 1460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923135183152224e-05,
+      "loss": 29.8082,
+      "step": 1461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922971975431383e-05,
+      "loss": 30.2258,
+      "step": 1462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922808595969094e-05,
+      "loss": 28.9877,
+      "step": 1463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922645044771053e-05,
+      "loss": 0.7677,
+      "step": 1464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922481321842969e-05,
+      "loss": 30.8326,
+      "step": 1465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922317427190551e-05,
+      "loss": 30.6791,
+      "step": 1466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922153360819518e-05,
+      "loss": 32.1056,
+      "step": 1467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921989122735593e-05,
+      "loss": 31.6415,
+      "step": 1468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921824712944507e-05,
+      "loss": 0.7098,
+      "step": 1469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921660131451993e-05,
+      "loss": 0.745,
+      "step": 1470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921495378263794e-05,
+      "loss": 0.7629,
+      "step": 1471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921330453385657e-05,
+      "loss": 0.7517,
+      "step": 1472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921165356823337e-05,
+      "loss": 32.6513,
+      "step": 1473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.921000088582593e-05,
+      "loss": 30.302,
+      "step": 1474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920834648669188e-05,
+      "loss": 0.7782,
+      "step": 1475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920669037088897e-05,
+      "loss": 34.2425,
+      "step": 1476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920503253847494e-05,
+      "loss": 32.5333,
+      "step": 1477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920337298950765e-05,
+      "loss": 29.6147,
+      "step": 1478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9201711724045e-05,
+      "loss": 0.7708,
+      "step": 1479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920004874214492e-05,
+      "loss": 30.1298,
+      "step": 1480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919838404386546e-05,
+      "loss": 0.7167,
+      "step": 1481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919671762926465e-05,
+      "loss": 31.1879,
+      "step": 1482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919504949840065e-05,
+      "loss": 29.1879,
+      "step": 1483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919337965133164e-05,
+      "loss": 30.8451,
+      "step": 1484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919170808811591e-05,
+      "loss": 0.7133,
+      "step": 1485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919003480881172e-05,
+      "loss": 29.3187,
+      "step": 1486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918835981347749e-05,
+      "loss": 30.0341,
+      "step": 1487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918668310217161e-05,
+      "loss": 34.196,
+      "step": 1488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918500467495262e-05,
+      "loss": 0.7025,
+      "step": 1489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918332453187905e-05,
+      "loss": 30.4607,
+      "step": 1490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91816426730095e-05,
+      "loss": 31.3834,
+      "step": 1491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917995909840265e-05,
+      "loss": 0.7725,
+      "step": 1492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917827380811726e-05,
+      "loss": 30.7455,
+      "step": 1493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917658680221209e-05,
+      "loss": 31.8964,
+      "step": 1494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917489808074601e-05,
+      "loss": 0.7999,
+      "step": 1495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917320764377791e-05,
+      "loss": 32.6504,
+      "step": 1496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91715154913668e-05,
+      "loss": 32.3112,
+      "step": 1497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916982162357167e-05,
+      "loss": 31.0739,
+      "step": 1498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916812604045165e-05,
+      "loss": 31.5495,
+      "step": 1499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916642874206586e-05,
+      "loss": 28.6267,
+      "step": 1500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916472972847353e-05,
+      "loss": 0.7009,
+      "step": 1501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916302899973393e-05,
+      "loss": 29.5874,
+      "step": 1502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916132655590639e-05,
+      "loss": 31.6387,
+      "step": 1503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915962239705028e-05,
+      "loss": 0.7105,
+      "step": 1504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91579165232251e-05,
+      "loss": 33.0997,
+      "step": 1505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915620893449031e-05,
+      "loss": 31.0738,
+      "step": 1506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91544996309055e-05,
+      "loss": 0.6941,
+      "step": 1507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91527886125303e-05,
+      "loss": 35.1793,
+      "step": 1508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915107587942442e-05,
+      "loss": 31.5741,
+      "step": 1509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914936143164757e-05,
+      "loss": 31.1059,
+      "step": 1510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914764526925959e-05,
+      "loss": 0.7165,
+      "step": 1511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914592739232034e-05,
+      "loss": 0.7843,
+      "step": 1512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914420780088975e-05,
+      "loss": 33.1519,
+      "step": 1513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91424864950278e-05,
+      "loss": 33.4001,
+      "step": 1514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914076347479455e-05,
+      "loss": 0.7692,
+      "step": 1515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913903874025012e-05,
+      "loss": 30.1978,
+      "step": 1516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913731229145466e-05,
+      "loss": 31.3615,
+      "step": 1517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913558412846839e-05,
+      "loss": 32.0158,
+      "step": 1518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913385425135162e-05,
+      "loss": 34.4613,
+      "step": 1519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91321226601647e-05,
+      "loss": 33.1856,
+      "step": 1520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9130389354968e-05,
+      "loss": 30.1868,
+      "step": 1521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912865433582205e-05,
+      "loss": 28.6071,
+      "step": 1522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912691760278733e-05,
+      "loss": 30.0478,
+      "step": 1523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912517915592443e-05,
+      "loss": 0.8121,
+      "step": 1524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912343899529402e-05,
+      "loss": 0.777,
+      "step": 1525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912169712095677e-05,
+      "loss": 25.4907,
+      "step": 1526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911995353297351e-05,
+      "loss": 0.7917,
+      "step": 1527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9118208231405e-05,
+      "loss": 31.4792,
+      "step": 1528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911646121631215e-05,
+      "loss": 31.6184,
+      "step": 1529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911471248775593e-05,
+      "loss": 30.9424,
+      "step": 1530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911296204579731e-05,
+      "loss": 27.9485,
+      "step": 1531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911120989049736e-05,
+      "loss": 30.1301,
+      "step": 1532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910945602191724e-05,
+      "loss": 0.7802,
+      "step": 1533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910770044011808e-05,
+      "loss": 30.4649,
+      "step": 1534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910594314516118e-05,
+      "loss": 28.5831,
+      "step": 1535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91041841371078e-05,
+      "loss": 29.0006,
+      "step": 1536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910242341601934e-05,
+      "loss": 0.872,
+      "step": 1537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91006609819572e-05,
+      "loss": 30.2008,
+      "step": 1538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909889683498287e-05,
+      "loss": 29.4054,
+      "step": 1539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909713097515789e-05,
+      "loss": 0.7117,
+      "step": 1540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909536340254388e-05,
+      "loss": 0.7794,
+      "step": 1541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909359411720248e-05,
+      "loss": 0.7907,
+      "step": 1542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909182311919542e-05,
+      "loss": 0.7806,
+      "step": 1543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90900504085845e-05,
+      "loss": 31.1481,
+      "step": 1544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908827598543155e-05,
+      "loss": 33.2921,
+      "step": 1545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908649984979847e-05,
+      "loss": 30.4761,
+      "step": 1546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908472200174721e-05,
+      "loss": 32.0609,
+      "step": 1547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908294244133982e-05,
+      "loss": 0.7062,
+      "step": 1548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908116116863835e-05,
+      "loss": 0.7181,
+      "step": 1549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907937818370496e-05,
+      "loss": 30.2477,
+      "step": 1550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907759348660186e-05,
+      "loss": 0.7061,
+      "step": 1551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907580707739128e-05,
+      "loss": 30.2943,
+      "step": 1552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907401895613558e-05,
+      "loss": 0.7347,
+      "step": 1553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907222912289709e-05,
+      "loss": 28.6564,
+      "step": 1554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907043757773828e-05,
+      "loss": 0.7779,
+      "step": 1555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906864432072168e-05,
+      "loss": 26.6246,
+      "step": 1556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906684935190978e-05,
+      "loss": 28.7377,
+      "step": 1557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906505267136525e-05,
+      "loss": 29.3534,
+      "step": 1558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906325427915075e-05,
+      "loss": 0.747,
+      "step": 1559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906145417532901e-05,
+      "loss": 31.0025,
+      "step": 1560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905965235996286e-05,
+      "loss": 28.4106,
+      "step": 1561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905784883311514e-05,
+      "loss": 29.452,
+      "step": 1562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905604359484874e-05,
+      "loss": 29.1805,
+      "step": 1563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905423664522667e-05,
+      "loss": 30.6272,
+      "step": 1564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905242798431196e-05,
+      "loss": 28.2476,
+      "step": 1565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905061761216772e-05,
+      "loss": 28.8862,
+      "step": 1566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904880552885707e-05,
+      "loss": 0.7262,
+      "step": 1567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904699173444325e-05,
+      "loss": 28.7632,
+      "step": 1568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904517622898953e-05,
+      "loss": 30.7947,
+      "step": 1569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904335901255926e-05,
+      "loss": 28.1297,
+      "step": 1570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90415400852158e-05,
+      "loss": 0.7558,
+      "step": 1571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903971944702264e-05,
+      "loss": 33.9985,
+      "step": 1572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903789709804329e-05,
+      "loss": 31.1466,
+      "step": 1573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90360730383413e-05,
+      "loss": 0.7342,
+      "step": 1574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903424726798032e-05,
+      "loss": 0.8103,
+      "step": 1575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903241978702404e-05,
+      "loss": 27.3493,
+      "step": 1576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903059059553623e-05,
+      "loss": 30.5532,
+      "step": 1577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902875969358068e-05,
+      "loss": 28.5159,
+      "step": 1578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902692708122126e-05,
+      "loss": 32.014,
+      "step": 1579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902509275852191e-05,
+      "loss": 29.7831,
+      "step": 1580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902325672554664e-05,
+      "loss": 31.1136,
+      "step": 1581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902141898235948e-05,
+      "loss": 0.6879,
+      "step": 1582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901957952902454e-05,
+      "loss": 29.7115,
+      "step": 1583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901773836560602e-05,
+      "loss": 0.8872,
+      "step": 1584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90158954921681e-05,
+      "loss": 0.7618,
+      "step": 1585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90140509087751e-05,
+      "loss": 29.1191,
+      "step": 1586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901220461549136e-05,
+      "loss": 0.7365,
+      "step": 1587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901035661238131e-05,
+      "loss": 29.8478,
+      "step": 1588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90085068995094e-05,
+      "loss": 30.6509,
+      "step": 1589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900665547694018e-05,
+      "loss": 0.8356,
+      "step": 1590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900480234473819e-05,
+      "loss": 29.7857,
+      "step": 1591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900294750296813e-05,
+      "loss": 29.251,
+      "step": 1592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900109095169469e-05,
+      "loss": 29.007,
+      "step": 1593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899923269098262e-05,
+      "loss": 0.7716,
+      "step": 1594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899737272089678e-05,
+      "loss": 0.7662,
+      "step": 1595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899551104150201e-05,
+      "loss": 31.5739,
+      "step": 1596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89936476528633e-05,
+      "loss": 0.7761,
+      "step": 1597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899178255504564e-05,
+      "loss": 0.8099,
+      "step": 1598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89899157481141e-05,
+      "loss": 29.7243,
+      "step": 1599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898804723213377e-05,
+      "loss": 0.7735,
+      "step": 1600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898617700716988e-05,
+      "loss": 0.761,
+      "step": 1601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898430507328766e-05,
+      "loss": 0.7574,
+      "step": 1602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898243143055241e-05,
+      "loss": 30.8957,
+      "step": 1603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898055607902951e-05,
+      "loss": 0.7682,
+      "step": 1604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897867901878434e-05,
+      "loss": 0.8,
+      "step": 1605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897680024988243e-05,
+      "loss": 30.8431,
+      "step": 1606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89749197723893e-05,
+      "loss": 0.7523,
+      "step": 1607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897303758637056e-05,
+      "loss": 33.0162,
+      "step": 1608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897115369189185e-05,
+      "loss": 0.7416,
+      "step": 1609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896926808901891e-05,
+      "loss": 0.7522,
+      "step": 1610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896738077781754e-05,
+      "loss": 0.7303,
+      "step": 1611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896549175835353e-05,
+      "loss": 29.0018,
+      "step": 1612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896360103069282e-05,
+      "loss": 26.4523,
+      "step": 1613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896170859490136e-05,
+      "loss": 30.1608,
+      "step": 1614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895981445104517e-05,
+      "loss": 0.7071,
+      "step": 1615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89579185991903e-05,
+      "loss": 0.7509,
+      "step": 1616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895602103940293e-05,
+      "loss": 0.6958,
+      "step": 1617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895412177174925e-05,
+      "loss": 0.7661,
+      "step": 1618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89522207962955e-05,
+      "loss": 0.744,
+      "step": 1619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895031811310799e-05,
+      "loss": 0.7425,
+      "step": 1620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894841372225314e-05,
+      "loss": 0.8116,
+      "step": 1621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894650762379734e-05,
+      "loss": 29.5279,
+      "step": 1622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894459981780711e-05,
+      "loss": 0.7834,
+      "step": 1623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894269030434899e-05,
+      "loss": 0.7032,
+      "step": 1624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89407790834896e-05,
+      "loss": 0.7566,
+      "step": 1625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893886615529562e-05,
+      "loss": 31.0813,
+      "step": 1626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89369515198338e-05,
+      "loss": 28.5203,
+      "step": 1627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893503517717088e-05,
+      "loss": 30.8477,
+      "step": 1628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893311712737377e-05,
+      "loss": 29.4452,
+      "step": 1629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893119737050936e-05,
+      "loss": 0.7718,
+      "step": 1630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892927590664462e-05,
+      "loss": 0.6912,
+      "step": 1631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892735273584658e-05,
+      "loss": 0.7224,
+      "step": 1632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892542785818233e-05,
+      "loss": 29.9901,
+      "step": 1633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892350127371904e-05,
+      "loss": 31.2375,
+      "step": 1634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89215729825239e-05,
+      "loss": 29.2597,
+      "step": 1635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891964298466419e-05,
+      "loss": 32.7858,
+      "step": 1636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891771128020723e-05,
+      "loss": 31.7775,
+      "step": 1637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891577786922042e-05,
+      "loss": 0.6931,
+      "step": 1638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891384275177121e-05,
+      "loss": 28.1533,
+      "step": 1639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891190592792708e-05,
+      "loss": 27.2613,
+      "step": 1640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890996739775562e-05,
+      "loss": 27.4154,
+      "step": 1641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890802716132445e-05,
+      "loss": 29.0505,
+      "step": 1642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890608521870128e-05,
+      "loss": 27.672,
+      "step": 1643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890414156995384e-05,
+      "loss": 0.7477,
+      "step": 1644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890219621514993e-05,
+      "loss": 30.8382,
+      "step": 1645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890024915435742e-05,
+      "loss": 0.7175,
+      "step": 1646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889830038764423e-05,
+      "loss": 0.7221,
+      "step": 1647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889634991507834e-05,
+      "loss": 0.7836,
+      "step": 1648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889439773672781e-05,
+      "loss": 0.8374,
+      "step": 1649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889244385266072e-05,
+      "loss": 30.1577,
+      "step": 1650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889048826294526e-05,
+      "loss": 30.1941,
+      "step": 1651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888853096764964e-05,
+      "loss": 30.2135,
+      "step": 1652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888657196684214e-05,
+      "loss": 0.7592,
+      "step": 1653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88846112605911e-05,
+      "loss": 0.8338,
+      "step": 1654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888264884896492e-05,
+      "loss": 29.6454,
+      "step": 1655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888068473203206e-05,
+      "loss": 0.7334,
+      "step": 1656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887871890986104e-05,
+      "loss": 0.7469,
+      "step": 1657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887675138252046e-05,
+      "loss": 29.9743,
+      "step": 1658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887478215007893e-05,
+      "loss": 32.7146,
+      "step": 1659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887281121260516e-05,
+      "loss": 30.9053,
+      "step": 1660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88708385701679e-05,
+      "loss": 29.0707,
+      "step": 1661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886886422283597e-05,
+      "loss": 0.735,
+      "step": 1662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886688817067826e-05,
+      "loss": 28.9285,
+      "step": 1663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886491041376369e-05,
+      "loss": 29.8138,
+      "step": 1664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886293095216126e-05,
+      "loss": 0.7396,
+      "step": 1665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886094978594003e-05,
+      "loss": 29.2181,
+      "step": 1666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885896691516909e-05,
+      "loss": 29.7817,
+      "step": 1667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885698233991765e-05,
+      "loss": 0.7399,
+      "step": 1668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885499606025495e-05,
+      "loss": 33.5587,
+      "step": 1669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885300807625022e-05,
+      "loss": 30.0595,
+      "step": 1670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885101838797289e-05,
+      "loss": 29.4672,
+      "step": 1671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88490269954923e-05,
+      "loss": 28.583,
+      "step": 1672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884703389887798e-05,
+      "loss": 29.359,
+      "step": 1673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884503909819942e-05,
+      "loss": 0.6731,
+      "step": 1674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884304259352624e-05,
+      "loss": 0.7299,
+      "step": 1675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884104438492808e-05,
+      "loss": 30.6067,
+      "step": 1676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883904447247462e-05,
+      "loss": 32.4248,
+      "step": 1677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883704285623568e-05,
+      "loss": 28.1324,
+      "step": 1678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883503953628104e-05,
+      "loss": 0.7517,
+      "step": 1679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883303451268062e-05,
+      "loss": 30.3693,
+      "step": 1680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883102778550434e-05,
+      "loss": 28.3433,
+      "step": 1681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882901935482223e-05,
+      "loss": 29.1156,
+      "step": 1682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882700922070435e-05,
+      "loss": 0.7521,
+      "step": 1683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882499738322081e-05,
+      "loss": 0.7503,
+      "step": 1684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882298384244182e-05,
+      "loss": 0.7203,
+      "step": 1685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882096859843761e-05,
+      "loss": 0.7276,
+      "step": 1686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881895165127848e-05,
+      "loss": 30.4493,
+      "step": 1687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881693300103479e-05,
+      "loss": 0.7898,
+      "step": 1688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881491264777697e-05,
+      "loss": 0.7092,
+      "step": 1689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88128905915755e-05,
+      "loss": 27.9568,
+      "step": 1690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881086683250092e-05,
+      "loss": 30.9989,
+      "step": 1691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880884137062383e-05,
+      "loss": 0.6779,
+      "step": 1692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88068142060149e-05,
+      "loss": 0.7124,
+      "step": 1693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880478533874483e-05,
+      "loss": 31.0905,
+      "step": 1694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88027547688844e-05,
+      "loss": 0.6904,
+      "step": 1695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880072249650447e-05,
+      "loss": 31.1756,
+      "step": 1696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879868852167591e-05,
+      "loss": 26.8205,
+      "step": 1697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879665284446971e-05,
+      "loss": 0.7858,
+      "step": 1698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879461546495685e-05,
+      "loss": 28.6405,
+      "step": 1699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879257638320841e-05,
+      "loss": 0.8009,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879053559929555e-05,
+      "loss": 28.5064,
+      "step": 1701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878849311328946e-05,
+      "loss": 31.2432,
+      "step": 1702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878644892526137e-05,
+      "loss": 0.7607,
+      "step": 1703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87844030352826e-05,
+      "loss": 31.1693,
+      "step": 1704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878235544342455e-05,
+      "loss": 29.6,
+      "step": 1705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878030614975862e-05,
+      "loss": 0.765,
+      "step": 1706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87782551543563e-05,
+      "loss": 30.029,
+      "step": 1707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877620245728918e-05,
+      "loss": 27.0574,
+      "step": 1708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877414805862882e-05,
+      "loss": 30.2672,
+      "step": 1709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877209195844692e-05,
+      "loss": 24.9673,
+      "step": 1710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877003415681521e-05,
+      "loss": 27.6334,
+      "step": 1711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876797465380546e-05,
+      "loss": 31.74,
+      "step": 1712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876591344948952e-05,
+      "loss": 30.3672,
+      "step": 1713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876385054393929e-05,
+      "loss": 27.2055,
+      "step": 1714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876178593722676e-05,
+      "loss": 30.1131,
+      "step": 1715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875971962942394e-05,
+      "loss": 0.7897,
+      "step": 1716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875765162060292e-05,
+      "loss": 0.7804,
+      "step": 1717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875558191083585e-05,
+      "loss": 26.9909,
+      "step": 1718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875351050019493e-05,
+      "loss": 29.5611,
+      "step": 1719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87514373887524e-05,
+      "loss": 0.7778,
+      "step": 1720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87493625765806e-05,
+      "loss": 33.3735,
+      "step": 1721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874728606375192e-05,
+      "loss": 27.4468,
+      "step": 1722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874520785033878e-05,
+      "loss": 0.7465,
+      "step": 1723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874312793641371e-05,
+      "loss": 27.6625,
+      "step": 1724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874104632204925e-05,
+      "loss": 0.728,
+      "step": 1725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8738963007318e-05,
+      "loss": 0.7783,
+      "step": 1726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873687799229268e-05,
+      "loss": 29.5597,
+      "step": 1727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8734791277046e-05,
+      "loss": 26.9471,
+      "step": 1728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873270286165075e-05,
+      "loss": 28.639,
+      "step": 1729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873061274617979e-05,
+      "loss": 29.3149,
+      "step": 1730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872852093070606e-05,
+      "loss": 26.7528,
+      "step": 1731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872642741530251e-05,
+      "loss": 26.9341,
+      "step": 1732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872433220004219e-05,
+      "loss": 0.6898,
+      "step": 1733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872223528499817e-05,
+      "loss": 26.2868,
+      "step": 1734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872013667024362e-05,
+      "loss": 0.7098,
+      "step": 1735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871803635585175e-05,
+      "loss": 28.3043,
+      "step": 1736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871593434189584e-05,
+      "loss": 0.8345,
+      "step": 1737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871383062844919e-05,
+      "loss": 30.1251,
+      "step": 1738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871172521558523e-05,
+      "loss": 29.5142,
+      "step": 1739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870961810337737e-05,
+      "loss": 0.6985,
+      "step": 1740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870750929189914e-05,
+      "loss": 0.7582,
+      "step": 1741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87053987812241e-05,
+      "loss": 0.8021,
+      "step": 1742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870328657142588e-05,
+      "loss": 0.7518,
+      "step": 1743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870117266257817e-05,
+      "loss": 0.7597,
+      "step": 1744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869905705475472e-05,
+      "loss": 28.1542,
+      "step": 1745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869693974802932e-05,
+      "loss": 33.7884,
+      "step": 1746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869482074247583e-05,
+      "loss": 31.6928,
+      "step": 1747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869270003816819e-05,
+      "loss": 0.761,
+      "step": 1748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869057763518038e-05,
+      "loss": 28.4414,
+      "step": 1749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868845353358642e-05,
+      "loss": 0.7199,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868632773346044e-05,
+      "loss": 0.775,
+      "step": 1751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868420023487659e-05,
+      "loss": 32.7979,
+      "step": 1752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868207103790908e-05,
+      "loss": 32.2308,
+      "step": 1753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867994014263218e-05,
+      "loss": 29.1825,
+      "step": 1754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867780754912027e-05,
+      "loss": 0.7541,
+      "step": 1755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86756732574477e-05,
+      "loss": 28.3271,
+      "step": 1756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867353726768897e-05,
+      "loss": 0.7358,
+      "step": 1757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867139957991856e-05,
+      "loss": 33.4925,
+      "step": 1758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866926019421104e-05,
+      "loss": 0.7649,
+      "step": 1759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866711911064108e-05,
+      "loss": 0.7464,
+      "step": 1760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866497632928336e-05,
+      "loss": 30.5541,
+      "step": 1761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866283185021261e-05,
+      "loss": 31.9764,
+      "step": 1762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866068567350366e-05,
+      "loss": 0.7659,
+      "step": 1763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865853779923138e-05,
+      "loss": 0.752,
+      "step": 1764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865638822747069e-05,
+      "loss": 29.4262,
+      "step": 1765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86542369582966e-05,
+      "loss": 28.7145,
+      "step": 1766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865208399178417e-05,
+      "loss": 0.7611,
+      "step": 1767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864992932800845e-05,
+      "loss": 29.6247,
+      "step": 1768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864777296704467e-05,
+      "loss": 30.4711,
+      "step": 1769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864561490896801e-05,
+      "loss": 27.1131,
+      "step": 1770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864345515385379e-05,
+      "loss": 30.9593,
+      "step": 1771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86412937017773e-05,
+      "loss": 31.8246,
+      "step": 1772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863913055281402e-05,
+      "loss": 0.7524,
+      "step": 1773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863696570703936e-05,
+      "loss": 29.1986,
+      "step": 1774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863479916452887e-05,
+      "loss": 27.4256,
+      "step": 1775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86326309253581e-05,
+      "loss": 26.2047,
+      "step": 1776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863046098960271e-05,
+      "loss": 26.7391,
+      "step": 1777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862828935733842e-05,
+      "loss": 27.797,
+      "step": 1778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862611602864096e-05,
+      "loss": 28.9357,
+      "step": 1779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862394100358614e-05,
+      "loss": 29.5674,
+      "step": 1780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862176428224988e-05,
+      "loss": 0.741,
+      "step": 1781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861958586470806e-05,
+      "loss": 29.9254,
+      "step": 1782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861740575103673e-05,
+      "loss": 0.7461,
+      "step": 1783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861522394131191e-05,
+      "loss": 0.7331,
+      "step": 1784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861304043560972e-05,
+      "loss": 0.7343,
+      "step": 1785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861085523400636e-05,
+      "loss": 26.0098,
+      "step": 1786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860866833657802e-05,
+      "loss": 27.5351,
+      "step": 1787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860647974340101e-05,
+      "loss": 30.1184,
+      "step": 1788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86042894545517e-05,
+      "loss": 29.7824,
+      "step": 1789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860209747010646e-05,
+      "loss": 0.827,
+      "step": 1790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859990379014179e-05,
+      "loss": 0.7512,
+      "step": 1791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859770841473421e-05,
+      "loss": 0.7923,
+      "step": 1792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859551134396031e-05,
+      "loss": 28.6091,
+      "step": 1793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859331257789674e-05,
+      "loss": 25.1843,
+      "step": 1794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859111211662018e-05,
+      "loss": 0.7951,
+      "step": 1795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858890996020742e-05,
+      "loss": 0.7815,
+      "step": 1796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858670610873528e-05,
+      "loss": 0.7504,
+      "step": 1797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858450056228063e-05,
+      "loss": 28.9453,
+      "step": 1798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858229332092044e-05,
+      "loss": 0.6964,
+      "step": 1799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85800843847317e-05,
+      "loss": 0.813,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857787375379143e-05,
+      "loss": 31.8133,
+      "step": 1801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857566142817681e-05,
+      "loss": 0.6595,
+      "step": 1802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857344740796499e-05,
+      "loss": 30.2691,
+      "step": 1803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857123169323319e-05,
+      "loss": 0.7592,
+      "step": 1804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856901428405874e-05,
+      "loss": 27.9611,
+      "step": 1805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856679518051898e-05,
+      "loss": 30.7156,
+      "step": 1806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856457438269134e-05,
+      "loss": 0.7945,
+      "step": 1807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856235189065326e-05,
+      "loss": 0.8684,
+      "step": 1808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85601277044823e-05,
+      "loss": 30.4027,
+      "step": 1809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855790182425604e-05,
+      "loss": 25.9323,
+      "step": 1810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855567425005216e-05,
+      "loss": 0.7061,
+      "step": 1811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855344498194835e-05,
+      "loss": 0.713,
+      "step": 1812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855121402002235e-05,
+      "loss": 29.6228,
+      "step": 1813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854898136435204e-05,
+      "loss": 0.7817,
+      "step": 1814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854674701501528e-05,
+      "loss": 30.051,
+      "step": 1815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854451097209001e-05,
+      "loss": 0.7525,
+      "step": 1816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854227323565426e-05,
+      "loss": 0.7967,
+      "step": 1817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854003380578608e-05,
+      "loss": 0.7038,
+      "step": 1818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85377926825636e-05,
+      "loss": 0.7485,
+      "step": 1819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8535549866065e-05,
+      "loss": 28.9784,
+      "step": 1820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853330535636851e-05,
+      "loss": 25.9889,
+      "step": 1821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853105915355247e-05,
+      "loss": 0.7801,
+      "step": 1822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852881125769519e-05,
+      "loss": 26.4484,
+      "step": 1823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852656166887513e-05,
+      "loss": 26.8106,
+      "step": 1824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852431038717074e-05,
+      "loss": 28.3856,
+      "step": 1825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852205741266058e-05,
+      "loss": 29.0452,
+      "step": 1826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851980274542323e-05,
+      "loss": 27.2218,
+      "step": 1827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851754638553735e-05,
+      "loss": 26.8167,
+      "step": 1828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851528833308165e-05,
+      "loss": 0.7391,
+      "step": 1829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851302858813494e-05,
+      "loss": 0.815,
+      "step": 1830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8510767150776e-05,
+      "loss": 28.8091,
+      "step": 1831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850850402108375e-05,
+      "loss": 29.7107,
+      "step": 1832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850623919913714e-05,
+      "loss": 29.495,
+      "step": 1833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850397268501516e-05,
+      "loss": 25.5525,
+      "step": 1834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850170447879692e-05,
+      "loss": 0.716,
+      "step": 1835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849943458056153e-05,
+      "loss": 29.8247,
+      "step": 1836
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849716299038814e-05,
+      "loss": 28.2331,
+      "step": 1837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849488970835604e-05,
+      "loss": 27.3275,
+      "step": 1838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849261473454452e-05,
+      "loss": 27.7711,
+      "step": 1839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849033806903295e-05,
+      "loss": 0.7718,
+      "step": 1840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848805971190074e-05,
+      "loss": 0.7603,
+      "step": 1841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84857796632274e-05,
+      "loss": 0.7696,
+      "step": 1842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848349792309243e-05,
+      "loss": 0.7796,
+      "step": 1843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848121449157547e-05,
+      "loss": 30.2262,
+      "step": 1844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847892936875616e-05,
+      "loss": 30.8667,
+      "step": 1845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84766425547142e-05,
+      "loss": 29.113,
+      "step": 1846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84743540495294e-05,
+      "loss": 0.7511,
+      "step": 1847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847206385328158e-05,
+      "loss": 0.7356,
+      "step": 1848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846977196605065e-05,
+      "loss": 31.6697,
+      "step": 1849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846747838791655e-05,
+      "loss": 33.6727,
+      "step": 1850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846518311895929e-05,
+      "loss": 30.9422,
+      "step": 1851
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846288615925894e-05,
+      "loss": 29.8053,
+      "step": 1852
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846058750889565e-05,
+      "loss": 27.8034,
+      "step": 1853
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845828716794959e-05,
+      "loss": 0.8002,
+      "step": 1854
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845598513650103e-05,
+      "loss": 0.7267,
+      "step": 1855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845368141463025e-05,
+      "loss": 32.0409,
+      "step": 1856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845137600241765e-05,
+      "loss": 28.5853,
+      "step": 1857
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844906889994363e-05,
+      "loss": 0.7122,
+      "step": 1858
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844676010728869e-05,
+      "loss": 0.7218,
+      "step": 1859
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844444962453335e-05,
+      "loss": 30.0848,
+      "step": 1860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844213745175826e-05,
+      "loss": 30.9489,
+      "step": 1861
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843982358904402e-05,
+      "loss": 0.7034,
+      "step": 1862
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84375080364714e-05,
+      "loss": 30.1626,
+      "step": 1863
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843519079412116e-05,
+      "loss": 26.3913,
+      "step": 1864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843287186207415e-05,
+      "loss": 0.7642,
+      "step": 1865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843055124041125e-05,
+      "loss": 27.3091,
+      "step": 1866
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842822892921342e-05,
+      "loss": 0.8004,
+      "step": 1867
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84259049285617e-05,
+      "loss": 0.7243,
+      "step": 1868
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842357923853713e-05,
+      "loss": 27.2991,
+      "step": 1869
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842125185922087e-05,
+      "loss": 28.2227,
+      "step": 1870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84189227906941e-05,
+      "loss": 27.0824,
+      "step": 1871
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841659203303806e-05,
+      "loss": 30.0379,
+      "step": 1872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841425958633407e-05,
+      "loss": 25.0389,
+      "step": 1873
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841192545066352e-05,
+      "loss": 0.7758,
+      "step": 1874
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840958962610783e-05,
+      "loss": 24.8474,
+      "step": 1875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840725211274845e-05,
+      "loss": 0.7392,
+      "step": 1876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840491291066696e-05,
+      "loss": 0.7323,
+      "step": 1877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840257201994496e-05,
+      "loss": 0.7041,
+      "step": 1878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840022944066412e-05,
+      "loss": 0.7607,
+      "step": 1879
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839788517290614e-05,
+      "loss": 28.214,
+      "step": 1880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839553921675281e-05,
+      "loss": 0.8235,
+      "step": 1881
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8393191572286e-05,
+      "loss": 26.167,
+      "step": 1882
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839084223958756e-05,
+      "loss": 28.0599,
+      "step": 1883
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838849121873949e-05,
+      "loss": 27.8017,
+      "step": 1884
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838613850982378e-05,
+      "loss": 27.5575,
+      "step": 1885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838378411292252e-05,
+      "loss": 0.7355,
+      "step": 1886
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838142802811785e-05,
+      "loss": 0.7473,
+      "step": 1887
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837907025549195e-05,
+      "loss": 27.9979,
+      "step": 1888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837671079512707e-05,
+      "loss": 30.92,
+      "step": 1889
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837434964710554e-05,
+      "loss": 0.787,
+      "step": 1890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837198681150971e-05,
+      "loss": 28.4239,
+      "step": 1891
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836962228842201e-05,
+      "loss": 26.897,
+      "step": 1892
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836725607792497e-05,
+      "loss": 0.7132,
+      "step": 1893
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836488818010108e-05,
+      "loss": 0.7514,
+      "step": 1894
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836251859503296e-05,
+      "loss": 0.7915,
+      "step": 1895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83601473228033e-05,
+      "loss": 30.3923,
+      "step": 1896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83577743634948e-05,
+      "loss": 29.0209,
+      "step": 1897
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835539971719026e-05,
+      "loss": 28.7744,
+      "step": 1898
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83530233839725e-05,
+      "loss": 0.8175,
+      "step": 1899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835064536392442e-05,
+      "loss": 0.7313,
+      "step": 1900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834826565712901e-05,
+      "loss": 0.7684,
+      "step": 1901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834588426366927e-05,
+      "loss": 27.1998,
+      "step": 1902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834350118362827e-05,
+      "loss": 26.9475,
+      "step": 1903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834111641708914e-05,
+      "loss": 0.7757,
+      "step": 1904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83387299641351e-05,
+      "loss": 0.7607,
+      "step": 1905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833634182484938e-05,
+      "loss": 0.7013,
+      "step": 1906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83339519993153e-05,
+      "loss": 0.6969,
+      "step": 1907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833156048761623e-05,
+      "loss": 0.7193,
+      "step": 1908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832916728983559e-05,
+      "loss": 28.1509,
+      "step": 1909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83267724060569e-05,
+      "loss": 27.8099,
+      "step": 1910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832437583636367e-05,
+      "loss": 27.4065,
+      "step": 1911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832197758083952e-05,
+      "loss": 26.4464,
+      "step": 1912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831957763956813e-05,
+      "loss": 29.9634,
+      "step": 1913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83171760126332e-05,
+      "loss": 30.4624,
+      "step": 1914
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831477270011852e-05,
+      "loss": 26.5861,
+      "step": 1915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831236770210794e-05,
+      "loss": 27.0127,
+      "step": 1916
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830996101868538e-05,
+      "loss": 27.3806,
+      "step": 1917
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830755264993473e-05,
+      "loss": 0.8003,
+      "step": 1918
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830514259594008e-05,
+      "loss": 28.0011,
+      "step": 1919
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830273085678545e-05,
+      "loss": 28.9003,
+      "step": 1920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830031743255503e-05,
+      "loss": 0.7936,
+      "step": 1921
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829790232333298e-05,
+      "loss": 0.7765,
+      "step": 1922
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829548552920356e-05,
+      "loss": 0.697,
+      "step": 1923
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829306705025107e-05,
+      "loss": 28.2979,
+      "step": 1924
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829064688655989e-05,
+      "loss": 29.5346,
+      "step": 1925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828822503821447e-05,
+      "loss": 28.0661,
+      "step": 1926
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828580150529926e-05,
+      "loss": 0.7648,
+      "step": 1927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828337628789884e-05,
+      "loss": 0.7957,
+      "step": 1928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82809493860978e-05,
+      "loss": 25.7153,
+      "step": 1929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827852079998079e-05,
+      "loss": 27.5785,
+      "step": 1930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827609052963256e-05,
+      "loss": 27.1613,
+      "step": 1931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827365857513788e-05,
+      "loss": 0.8035,
+      "step": 1932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827122493658159e-05,
+      "loss": 29.5555,
+      "step": 1933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826878961404857e-05,
+      "loss": 26.0139,
+      "step": 1934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826635260762383e-05,
+      "loss": 29.4961,
+      "step": 1935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826391391739233e-05,
+      "loss": 27.916,
+      "step": 1936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826147354343918e-05,
+      "loss": 29.175,
+      "step": 1937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82590314858495e-05,
+      "loss": 28.1661,
+      "step": 1938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82565877447085e-05,
+      "loss": 28.8923,
+      "step": 1939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82541423201014e-05,
+      "loss": 25.4961,
+      "step": 1940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825169521211354e-05,
+      "loss": 0.75,
+      "step": 1941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824924642083026e-05,
+      "loss": 0.7568,
+      "step": 1942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824679594633703e-05,
+      "loss": 28.9638,
+      "step": 1943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824434378871931e-05,
+      "loss": 27.6494,
+      "step": 1944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824188994806265e-05,
+      "loss": 28.4495,
+      "step": 1945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823943442445263e-05,
+      "loss": 0.7371,
+      "step": 1946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823697721797496e-05,
+      "loss": 28.1449,
+      "step": 1947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823451832871533e-05,
+      "loss": 26.0393,
+      "step": 1948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823205775675952e-05,
+      "loss": 26.9894,
+      "step": 1949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822959550219337e-05,
+      "loss": 27.5667,
+      "step": 1950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822713156510279e-05,
+      "loss": 0.7501,
+      "step": 1951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822466594557371e-05,
+      "loss": 30.8902,
+      "step": 1952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82221986436922e-05,
+      "loss": 28.9833,
+      "step": 1953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821972965954427e-05,
+      "loss": 27.0081,
+      "step": 1954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821725899321609e-05,
+      "loss": 26.1783,
+      "step": 1955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821478664479384e-05,
+      "loss": 0.7403,
+      "step": 1956
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821231261436377e-05,
+      "loss": 26.5745,
+      "step": 1957
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820983690201218e-05,
+      "loss": 29.7362,
+      "step": 1958
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820735950782545e-05,
+      "loss": 28.6297,
+      "step": 1959
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820488043189002e-05,
+      "loss": 0.6838,
+      "step": 1960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820239967429233e-05,
+      "loss": 0.7696,
+      "step": 1961
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819991723511897e-05,
+      "loss": 28.5528,
+      "step": 1962
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81974331144565e-05,
+      "loss": 0.7237,
+      "step": 1963
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81949473123916e-05,
+      "loss": 28.2741,
+      "step": 1964
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8192459829011e-05,
+      "loss": 28.9813,
+      "step": 1965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818997066440148e-05,
+      "loss": 0.7932,
+      "step": 1966
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818747981864985e-05,
+      "loss": 0.7535,
+      "step": 1967
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818498729184303e-05,
+      "loss": 0.7474,
+      "step": 1968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818249308406795e-05,
+      "loss": 26.5938,
+      "step": 1969
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817999719541162e-05,
+      "loss": 28.2397,
+      "step": 1970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817749962596115e-05,
+      "loss": 0.7613,
+      "step": 1971
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817500037580364e-05,
+      "loss": 26.9778,
+      "step": 1972
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817249944502629e-05,
+      "loss": 0.7134,
+      "step": 1973
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816999683371632e-05,
+      "loss": 28.4884,
+      "step": 1974
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816749254196107e-05,
+      "loss": 28.9088,
+      "step": 1975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816498656984789e-05,
+      "loss": 27.844,
+      "step": 1976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81624789174642e-05,
+      "loss": 0.7057,
+      "step": 1977
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815996958489748e-05,
+      "loss": 0.7866,
+      "step": 1978
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815745857223527e-05,
+      "loss": 27.0607,
+      "step": 1979
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815494587956518e-05,
+      "loss": 28.0321,
+      "step": 1980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815243150697485e-05,
+      "loss": 0.7577,
+      "step": 1981
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8149915454552e-05,
+      "loss": 0.7602,
+      "step": 1982
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814739772238443e-05,
+      "loss": 29.3944,
+      "step": 1983
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814487831055994e-05,
+      "loss": 28.0431,
+      "step": 1984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814235721916642e-05,
+      "loss": 0.7471,
+      "step": 1985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813983444829187e-05,
+      "loss": 0.723,
+      "step": 1986
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813730999802422e-05,
+      "loss": 0.7403,
+      "step": 1987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813478386845161e-05,
+      "loss": 26.1215,
+      "step": 1988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813225605966212e-05,
+      "loss": 27.5825,
+      "step": 1989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812972657174395e-05,
+      "loss": 0.7959,
+      "step": 1990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812719540478535e-05,
+      "loss": 0.7859,
+      "step": 1991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81246625588746e-05,
+      "loss": 0.7455,
+      "step": 1992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81221280341001e-05,
+      "loss": 0.7442,
+      "step": 1993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.811959183055021e-05,
+      "loss": 31.7368,
+      "step": 1994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811705394831348e-05,
+      "loss": 29.4874,
+      "step": 1995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811451438747839e-05,
+      "loss": 28.3152,
+      "step": 1996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811197314813354e-05,
+      "loss": 0.7826,
+      "step": 1997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.81094302303676e-05,
+      "loss": 27.4483,
+      "step": 1998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810688563426926e-05,
+      "loss": 26.7344,
+      "step": 1999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810433935992733e-05,
+      "loss": 26.8635,
+      "step": 2000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.81017914074306e-05,
+      "loss": 0.778,
+      "step": 2001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809924177686797e-05,
+      "loss": 26.0788,
+      "step": 2002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80966904683284e-05,
+      "loss": 29.6381,
+      "step": 2003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809413748190088e-05,
+      "loss": 0.7223,
+      "step": 2004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809158281767446e-05,
+      "loss": 25.6803,
+      "step": 2005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808902647573827e-05,
+      "loss": 0.7345,
+      "step": 2006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808646845618153e-05,
+      "loss": 26.0615,
+      "step": 2007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808390875909341e-05,
+      "loss": 0.7345,
+      "step": 2008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808134738456325e-05,
+      "loss": 28.257,
+      "step": 2009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807878433268039e-05,
+      "loss": 26.0003,
+      "step": 2010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807621960353426e-05,
+      "loss": 27.1529,
+      "step": 2011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807365319721432e-05,
+      "loss": 0.7804,
+      "step": 2012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807108511381011e-05,
+      "loss": 26.3943,
+      "step": 2013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80685153534112e-05,
+      "loss": 24.8241,
+      "step": 2014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806594391610727e-05,
+      "loss": 26.1956,
+      "step": 2015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806337080198798e-05,
+      "loss": 26.9707,
+      "step": 2016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806079601114313e-05,
+      "loss": 26.236,
+      "step": 2017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805821954366255e-05,
+      "loss": 28.0223,
+      "step": 2018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805564139963608e-05,
+      "loss": 27.9733,
+      "step": 2019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80530615791537e-05,
+      "loss": 0.7828,
+      "step": 2020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805048008230542e-05,
+      "loss": 25.3625,
+      "step": 2021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804789690918125e-05,
+      "loss": 0.6898,
+      "step": 2022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804531205987133e-05,
+      "loss": 27.8145,
+      "step": 2023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804272553446583e-05,
+      "loss": 0.744,
+      "step": 2024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.8040137333055e-05,
+      "loss": 29.5717,
+      "step": 2025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803754745572912e-05,
+      "loss": 28.6724,
+      "step": 2026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803495590257853e-05,
+      "loss": 27.0216,
+      "step": 2027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803236267369365e-05,
+      "loss": 29.5829,
+      "step": 2028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802976776916494e-05,
+      "loss": 25.4593,
+      "step": 2029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802717118908293e-05,
+      "loss": 27.3959,
+      "step": 2030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80245729335382e-05,
+      "loss": 0.7696,
+      "step": 2031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802197300262139e-05,
+      "loss": 0.7457,
+      "step": 2032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801937139642322e-05,
+      "loss": 26.321,
+      "step": 2033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801676811503442e-05,
+      "loss": 24.6376,
+      "step": 2034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801416315854581e-05,
+      "loss": 27.4148,
+      "step": 2035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80115565270483e-05,
+      "loss": 26.4993,
+      "step": 2036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800894822063278e-05,
+      "loss": 26.7165,
+      "step": 2037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800633823939027e-05,
+      "loss": 27.727,
+      "step": 2038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800372658341182e-05,
+      "loss": 25.5938,
+      "step": 2039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800111325278853e-05,
+      "loss": 0.6939,
+      "step": 2040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799849824761158e-05,
+      "loss": 0.765,
+      "step": 2041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799588156797217e-05,
+      "loss": 0.6564,
+      "step": 2042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799326321396162e-05,
+      "loss": 28.3316,
+      "step": 2043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799064318567124e-05,
+      "loss": 0.7299,
+      "step": 2044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798802148319246e-05,
+      "loss": 25.3149,
+      "step": 2045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798539810661673e-05,
+      "loss": 0.74,
+      "step": 2046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798277305603557e-05,
+      "loss": 26.7755,
+      "step": 2047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798014633154052e-05,
+      "loss": 27.1628,
+      "step": 2048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797751793322327e-05,
+      "loss": 26.9119,
+      "step": 2049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79748878611755e-05,
+      "loss": 23.8693,
+      "step": 2050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797225611548894e-05,
+      "loss": 27.2116,
+      "step": 2051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796962269625542e-05,
+      "loss": 29.2237,
+      "step": 2052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79669876035668e-05,
+      "loss": 0.699,
+      "step": 2053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7964350837515e-05,
+      "loss": 0.7659,
+      "step": 2054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796171239819203e-05,
+      "loss": 28.248,
+      "step": 2055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795907228568993e-05,
+      "loss": 0.7369,
+      "step": 2056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795643050010077e-05,
+      "loss": 28.78,
+      "step": 2057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795378704151675e-05,
+      "loss": 27.3353,
+      "step": 2058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795114191003006e-05,
+      "loss": 27.9446,
+      "step": 2059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794849510573297e-05,
+      "loss": 0.7893,
+      "step": 2060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794584662871787e-05,
+      "loss": 25.8069,
+      "step": 2061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794319647907709e-05,
+      "loss": 28.5988,
+      "step": 2062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794054465690312e-05,
+      "loss": 28.2234,
+      "step": 2063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793789116228846e-05,
+      "loss": 0.6831,
+      "step": 2064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793523599532567e-05,
+      "loss": 27.5371,
+      "step": 2065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79325791561074e-05,
+      "loss": 0.7915,
+      "step": 2066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792992064472632e-05,
+      "loss": 28.603,
+      "step": 2067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792726046127519e-05,
+      "loss": 25.7178,
+      "step": 2068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792459860584678e-05,
+      "loss": 27.6446,
+      "step": 2069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792193507853399e-05,
+      "loss": 0.7455,
+      "step": 2070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79192698794297e-05,
+      "loss": 0.7168,
+      "step": 2071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791660300862693e-05,
+      "loss": 0.7578,
+      "step": 2072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791393446621867e-05,
+      "loss": 26.8052,
+      "step": 2073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791126425229805e-05,
+      "loss": 0.765,
+      "step": 2074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79085923669582e-05,
+      "loss": 26.9768,
+      "step": 2075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790591881029235e-05,
+      "loss": 26.2175,
+      "step": 2076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790324358239376e-05,
+      "loss": 0.723,
+      "step": 2077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790056668335574e-05,
+      "loss": 25.8647,
+      "step": 2078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789788811327171e-05,
+      "loss": 27.386,
+      "step": 2079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789520787223509e-05,
+      "loss": 0.7265,
+      "step": 2080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789252596033937e-05,
+      "loss": 0.7197,
+      "step": 2081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788984237767815e-05,
+      "loss": 25.0043,
+      "step": 2082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788715712434502e-05,
+      "loss": 0.7351,
+      "step": 2083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788447020043367e-05,
+      "loss": 0.7513,
+      "step": 2084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788178160603782e-05,
+      "loss": 26.327,
+      "step": 2085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787909134125128e-05,
+      "loss": 0.7692,
+      "step": 2086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787639940616788e-05,
+      "loss": 25.2502,
+      "step": 2087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787370580088157e-05,
+      "loss": 26.9269,
+      "step": 2088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787101052548628e-05,
+      "loss": 25.0482,
+      "step": 2089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786831358007604e-05,
+      "loss": 0.7062,
+      "step": 2090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786561496474495e-05,
+      "loss": 29.4076,
+      "step": 2091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786291467958716e-05,
+      "loss": 26.6668,
+      "step": 2092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786021272469685e-05,
+      "loss": 27.5909,
+      "step": 2093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78575091001683e-05,
+      "loss": 24.8752,
+      "step": 2094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78548038060958e-05,
+      "loss": 0.7485,
+      "step": 2095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785209684257375e-05,
+      "loss": 0.7824,
+      "step": 2096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784938820969657e-05,
+      "loss": 25.3937,
+      "step": 2097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784667790755877e-05,
+      "loss": 28.3589,
+      "step": 2098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784396593625488e-05,
+      "loss": 26.5322,
+      "step": 2099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784125229587953e-05,
+      "loss": 26.0525,
+      "step": 2100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783853698652737e-05,
+      "loss": 26.5591,
+      "step": 2101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783582000829313e-05,
+      "loss": 25.2425,
+      "step": 2102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78331013612716e-05,
+      "loss": 0.6594,
+      "step": 2103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783038104555763e-05,
+      "loss": 25.8195,
+      "step": 2104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782765906124609e-05,
+      "loss": 28.1077,
+      "step": 2105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782493540843198e-05,
+      "loss": 28.1077,
+      "step": 2106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782221008721027e-05,
+      "loss": 25.0828,
+      "step": 2107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781948309767607e-05,
+      "loss": 25.4143,
+      "step": 2108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781675443992451e-05,
+      "loss": 28.0226,
+      "step": 2109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781402411405076e-05,
+      "loss": 28.8056,
+      "step": 2110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781129212015008e-05,
+      "loss": 0.7777,
+      "step": 2111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780855845831778e-05,
+      "loss": 25.0004,
+      "step": 2112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780582312864923e-05,
+      "loss": 0.7046,
+      "step": 2113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780308613123985e-05,
+      "loss": 0.7383,
+      "step": 2114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780034746618512e-05,
+      "loss": 28.1419,
+      "step": 2115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779760713358059e-05,
+      "loss": 0.7608,
+      "step": 2116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779486513352183e-05,
+      "loss": 0.7176,
+      "step": 2117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779212146610454e-05,
+      "loss": 0.7036,
+      "step": 2118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778937613142441e-05,
+      "loss": 26.2706,
+      "step": 2119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778662912957719e-05,
+      "loss": 0.8214,
+      "step": 2120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778388046065876e-05,
+      "loss": 26.5542,
+      "step": 2121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778113012476497e-05,
+      "loss": 27.8231,
+      "step": 2122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77783781219918e-05,
+      "loss": 27.3785,
+      "step": 2123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777562445243523e-05,
+      "loss": 25.53,
+      "step": 2124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777286911619133e-05,
+      "loss": 0.7202,
+      "step": 2125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777011211335622e-05,
+      "loss": 27.1866,
+      "step": 2126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776735344402608e-05,
+      "loss": 27.4817,
+      "step": 2127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776459310829717e-05,
+      "loss": 0.7015,
+      "step": 2128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776183110626576e-05,
+      "loss": 0.8213,
+      "step": 2129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775906743802821e-05,
+      "loss": 27.3721,
+      "step": 2130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775630210368094e-05,
+      "loss": 25.5974,
+      "step": 2131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775353510332042e-05,
+      "loss": 25.0116,
+      "step": 2132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775076643704315e-05,
+      "loss": 26.5881,
+      "step": 2133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774799610494577e-05,
+      "loss": 0.7724,
+      "step": 2134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774522410712489e-05,
+      "loss": 26.6505,
+      "step": 2135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774245044367722e-05,
+      "loss": 27.1666,
+      "step": 2136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773967511469952e-05,
+      "loss": 0.7042,
+      "step": 2137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773689812028862e-05,
+      "loss": 26.7561,
+      "step": 2138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773411946054136e-05,
+      "loss": 0.6812,
+      "step": 2139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773133913555473e-05,
+      "loss": 0.773,
+      "step": 2140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772855714542568e-05,
+      "loss": 27.3986,
+      "step": 2141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772577349025129e-05,
+      "loss": 30.2619,
+      "step": 2142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772298817012866e-05,
+      "loss": 26.8347,
+      "step": 2143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772020118515495e-05,
+      "loss": 27.7627,
+      "step": 2144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771741253542741e-05,
+      "loss": 26.9747,
+      "step": 2145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771462222104329e-05,
+      "loss": 30.8094,
+      "step": 2146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771183024209995e-05,
+      "loss": 26.2505,
+      "step": 2147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77090365986948e-05,
+      "loss": 0.7447,
+      "step": 2148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770624129092529e-05,
+      "loss": 25.9443,
+      "step": 2149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770344431888891e-05,
+      "loss": 25.1028,
+      "step": 2150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77006456826833e-05,
+      "loss": 0.839,
+      "step": 2151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7697845382406e-05,
+      "loss": 26.3879,
+      "step": 2152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769504341815479e-05,
+      "loss": 24.0153,
+      "step": 2153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769223979002737e-05,
+      "loss": 0.7185,
+      "step": 2154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768943449812154e-05,
+      "loss": 25.3566,
+      "step": 2155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768662754253519e-05,
+      "loss": 0.7658,
+      "step": 2156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768381892336625e-05,
+      "loss": 0.7546,
+      "step": 2157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768100864071267e-05,
+      "loss": 0.7221,
+      "step": 2158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767819669467251e-05,
+      "loss": 25.3915,
+      "step": 2159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767538308534386e-05,
+      "loss": 25.6744,
+      "step": 2160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767256781282486e-05,
+      "loss": 0.7386,
+      "step": 2161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766975087721375e-05,
+      "loss": 0.7244,
+      "step": 2162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766693227860879e-05,
+      "loss": 0.6886,
+      "step": 2163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76641120171083e-05,
+      "loss": 0.776,
+      "step": 2164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766129009281069e-05,
+      "loss": 0.7309,
+      "step": 2165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765846650581437e-05,
+      "loss": 0.6914,
+      "step": 2166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765564125621787e-05,
+      "loss": 26.2345,
+      "step": 2167
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.765281434411974e-05,
+      "loss": 0.7374,
+      "step": 2168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764998576961862e-05,
+      "loss": 25.7411,
+      "step": 2169
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764715553281315e-05,
+      "loss": 25.2987,
+      "step": 2170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764432363380208e-05,
+      "loss": 27.1311,
+      "step": 2171
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764149007268421e-05,
+      "loss": 27.3914,
+      "step": 2172
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76386548495584e-05,
+      "loss": 25.1028,
+      "step": 2173
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763581796452353e-05,
+      "loss": 0.7758,
+      "step": 2174
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76329794176786e-05,
+      "loss": 28.5433,
+      "step": 2175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763013920912261e-05,
+      "loss": 26.9544,
+      "step": 2176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762729733895464e-05,
+      "loss": 25.3808,
+      "step": 2177
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762445380727385e-05,
+      "loss": 27.0357,
+      "step": 2178
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762160861417943e-05,
+      "loss": 0.7821,
+      "step": 2179
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761876175977063e-05,
+      "loss": 23.1992,
+      "step": 2180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761591324414679e-05,
+      "loss": 0.7232,
+      "step": 2181
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761306306740724e-05,
+      "loss": 27.255,
+      "step": 2182
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761021122965144e-05,
+      "loss": 0.7716,
+      "step": 2183
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760735773097888e-05,
+      "loss": 0.7494,
+      "step": 2184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76045025714891e-05,
+      "loss": 0.8039,
+      "step": 2185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760164575128169e-05,
+      "loss": 0.7328,
+      "step": 2186
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759878727045633e-05,
+      "loss": 0.7528,
+      "step": 2187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759592712911274e-05,
+      "loss": 0.8281,
+      "step": 2188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759306532735069e-05,
+      "loss": 28.2296,
+      "step": 2189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759020186527003e-05,
+      "loss": 28.1012,
+      "step": 2190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758733674297062e-05,
+      "loss": 0.7782,
+      "step": 2191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758446996055245e-05,
+      "loss": 30.9882,
+      "step": 2192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758160151811552e-05,
+      "loss": 27.9185,
+      "step": 2193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75787314157599e-05,
+      "loss": 0.7738,
+      "step": 2194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757585965358569e-05,
+      "loss": 28.1458,
+      "step": 2195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757298623169311e-05,
+      "loss": 28.4936,
+      "step": 2196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757011115018238e-05,
+      "loss": 26.0665,
+      "step": 2197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756723440915379e-05,
+      "loss": 0.743,
+      "step": 2198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75643560087077e-05,
+      "loss": 23.7798,
+      "step": 2199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756147594894456e-05,
+      "loss": 24.8464,
+      "step": 2200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755859422996481e-05,
+      "loss": 0.781,
+      "step": 2201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755571085186898e-05,
+      "loss": 24.7214,
+      "step": 2202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755282581475769e-05,
+      "loss": 28.2405,
+      "step": 2203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754993911873153e-05,
+      "loss": 27.1582,
+      "step": 2204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754705076389128e-05,
+      "loss": 27.6957,
+      "step": 2205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754416075033763e-05,
+      "loss": 0.729,
+      "step": 2206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754126907817145e-05,
+      "loss": 27.961,
+      "step": 2207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753837574749359e-05,
+      "loss": 0.79,
+      "step": 2208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753548075840499e-05,
+      "loss": 0.7149,
+      "step": 2209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753258411100665e-05,
+      "loss": 26.682,
+      "step": 2210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752968580539963e-05,
+      "loss": 0.7644,
+      "step": 2211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752678584168502e-05,
+      "loss": 28.3053,
+      "step": 2212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752388421996401e-05,
+      "loss": 26.3584,
+      "step": 2213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75209809403378e-05,
+      "loss": 28.8018,
+      "step": 2214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75180760029077e-05,
+      "loss": 0.7373,
+      "step": 2215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751516940777503e-05,
+      "loss": 0.7704,
+      "step": 2216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751226115504118e-05,
+      "loss": 25.9853,
+      "step": 2217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750935124480764e-05,
+      "loss": 28.5242,
+      "step": 2218
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75064396771759e-05,
+      "loss": 26.4912,
+      "step": 2219
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750352645224754e-05,
+      "loss": 27.5201,
+      "step": 2220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75006115701242e-05,
+      "loss": 26.2617,
+      "step": 2221
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749769503090753e-05,
+      "loss": 0.7181,
+      "step": 2222
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749477683469932e-05,
+      "loss": 23.7744,
+      "step": 2223
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749185698160134e-05,
+      "loss": 29.6763,
+      "step": 2224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74889354717155e-05,
+      "loss": 25.407,
+      "step": 2225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748601230514365e-05,
+      "loss": 25.5514,
+      "step": 2226
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748308748198783e-05,
+      "loss": 0.7161,
+      "step": 2227
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748016100235002e-05,
+      "loss": 25.3781,
+      "step": 2228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747723286633235e-05,
+      "loss": 0.7818,
+      "step": 2229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747430307403696e-05,
+      "loss": 0.7247,
+      "step": 2230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747137162556606e-05,
+      "loss": 0.6994,
+      "step": 2231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74684385210219e-05,
+      "loss": 27.9332,
+      "step": 2232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746550376050682e-05,
+      "loss": 28.9171,
+      "step": 2233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746256734412322e-05,
+      "loss": 0.7191,
+      "step": 2234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74596292719735e-05,
+      "loss": 28.166,
+      "step": 2235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745668954416016e-05,
+      "loss": 27.9605,
+      "step": 2236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745374816078578e-05,
+      "loss": 0.7547,
+      "step": 2237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745080512195298e-05,
+      "loss": 24.9062,
+      "step": 2238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744786042776437e-05,
+      "loss": 0.7571,
+      "step": 2239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744491407832274e-05,
+      "loss": 28.7328,
+      "step": 2240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744196607373086e-05,
+      "loss": 25.426,
+      "step": 2241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743901641409155e-05,
+      "loss": 0.7519,
+      "step": 2242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743606509950773e-05,
+      "loss": 26.3549,
+      "step": 2243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743311213008237e-05,
+      "loss": 23.7943,
+      "step": 2244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743015750591846e-05,
+      "loss": 0.7584,
+      "step": 2245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74272012271191e-05,
+      "loss": 25.9375,
+      "step": 2246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742424329378741e-05,
+      "loss": 27.3243,
+      "step": 2247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742128370602657e-05,
+      "loss": 26.07,
+      "step": 2248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741832246393982e-05,
+      "loss": 26.8453,
+      "step": 2249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74153595676305e-05,
+      "loss": 0.7806,
+      "step": 2250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741239501720197e-05,
+      "loss": 26.6263,
+      "step": 2251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74094288127576e-05,
+      "loss": 0.7578,
+      "step": 2252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740646095440093e-05,
+      "loss": 27.0423,
+      "step": 2253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740349144223546e-05,
+      "loss": 28.6096,
+      "step": 2254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740052027636479e-05,
+      "loss": 27.0648,
+      "step": 2255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739754745689258e-05,
+      "loss": 26.6331,
+      "step": 2256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739457298392251e-05,
+      "loss": 26.2776,
+      "step": 2257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739159685755839e-05,
+      "loss": 26.6455,
+      "step": 2258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738861907790402e-05,
+      "loss": 27.097,
+      "step": 2259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738563964506329e-05,
+      "loss": 25.4007,
+      "step": 2260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738265855914013e-05,
+      "loss": 25.8875,
+      "step": 2261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737967582023853e-05,
+      "loss": 27.5663,
+      "step": 2262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737669142846258e-05,
+      "loss": 26.9974,
+      "step": 2263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737370538391636e-05,
+      "loss": 27.1932,
+      "step": 2264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737071768670405e-05,
+      "loss": 26.2154,
+      "step": 2265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736772833692988e-05,
+      "loss": 24.4426,
+      "step": 2266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736473733469812e-05,
+      "loss": 24.2138,
+      "step": 2267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736174468011314e-05,
+      "loss": 0.8027,
+      "step": 2268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735875037327933e-05,
+      "loss": 0.8182,
+      "step": 2269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735575441430115e-05,
+      "loss": 28.8732,
+      "step": 2270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735275680328311e-05,
+      "loss": 0.8025,
+      "step": 2271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734975754032977e-05,
+      "loss": 24.2546,
+      "step": 2272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73467566255458e-05,
+      "loss": 0.8061,
+      "step": 2273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734375405903587e-05,
+      "loss": 23.6307,
+      "step": 2274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734074984090471e-05,
+      "loss": 0.7582,
+      "step": 2275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733774397125714e-05,
+      "loss": 24.9097,
+      "step": 2276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733473645019803e-05,
+      "loss": 25.9105,
+      "step": 2277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733172727783228e-05,
+      "loss": 27.0609,
+      "step": 2278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732871645426488e-05,
+      "loss": 25.3201,
+      "step": 2279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732570397960086e-05,
+      "loss": 0.7091,
+      "step": 2280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732268985394533e-05,
+      "loss": 26.0103,
+      "step": 2281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73196740774034e-05,
+      "loss": 25.9414,
+      "step": 2282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731665665008031e-05,
+      "loss": 25.1789,
+      "step": 2283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731363757208132e-05,
+      "loss": 0.7835,
+      "step": 2284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731061684351175e-05,
+      "loss": 25.0938,
+      "step": 2285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730759446447697e-05,
+      "loss": 0.7751,
+      "step": 2286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730457043508246e-05,
+      "loss": 0.7635,
+      "step": 2287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730154475543364e-05,
+      "loss": 0.7471,
+      "step": 2288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729851742563614e-05,
+      "loss": 27.5787,
+      "step": 2289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729548844579552e-05,
+      "loss": 0.7552,
+      "step": 2290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729245781601747e-05,
+      "loss": 26.2813,
+      "step": 2291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72894255364077e-05,
+      "loss": 23.6008,
+      "step": 2292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728639160707202e-05,
+      "loss": 0.7164,
+      "step": 2293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728335602811625e-05,
+      "loss": 0.8008,
+      "step": 2294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72803187996463e-05,
+      "loss": 27.7661,
+      "step": 2295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727727992176811e-05,
+      "loss": 25.7029,
+      "step": 2296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72742393945877e-05,
+      "loss": 24.8109,
+      "step": 2297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727119721821115e-05,
+      "loss": 27.1553,
+      "step": 2298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726815339274458e-05,
+      "loss": 0.6874,
+      "step": 2299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726510791829417e-05,
+      "loss": 0.701,
+      "step": 2300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726206079496618e-05,
+      "loss": 0.7465,
+      "step": 2301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725901202286691e-05,
+      "loss": 0.7784,
+      "step": 2302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72559616021027e-05,
+      "loss": 23.8903,
+      "step": 2303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725290953278e-05,
+      "loss": 27.8342,
+      "step": 2304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724985581500522e-05,
+      "loss": 27.2427,
+      "step": 2305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724680044888497e-05,
+      "loss": 0.7538,
+      "step": 2306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724374343452578e-05,
+      "loss": 25.2409,
+      "step": 2307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724068477203432e-05,
+      "loss": 27.1767,
+      "step": 2308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723762446151729e-05,
+      "loss": 23.0586,
+      "step": 2309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723456250308148e-05,
+      "loss": 26.0031,
+      "step": 2310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723149889683365e-05,
+      "loss": 27.0443,
+      "step": 2311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722843364288072e-05,
+      "loss": 0.7481,
+      "step": 2312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72253667413296e-05,
+      "loss": 27.4579,
+      "step": 2313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72222981922873e-05,
+      "loss": 25.3115,
+      "step": 2314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721922799586086e-05,
+      "loss": 26.8676,
+      "step": 2315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721615615215738e-05,
+      "loss": 0.7696,
+      "step": 2316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721308266128403e-05,
+      "loss": 0.7074,
+      "step": 2317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721000752334803e-05,
+      "loss": 25.0382,
+      "step": 2318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720693073845667e-05,
+      "loss": 0.7453,
+      "step": 2319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720385230671727e-05,
+      "loss": 25.0315,
+      "step": 2320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720077222823723e-05,
+      "loss": 0.758,
+      "step": 2321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7197690503124e-05,
+      "loss": 0.7597,
+      "step": 2322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719460713148507e-05,
+      "loss": 0.7209,
+      "step": 2323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719152211342804e-05,
+      "loss": 27.7388,
+      "step": 2324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718843544906051e-05,
+      "loss": 25.8521,
+      "step": 2325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718534713849017e-05,
+      "loss": 0.7371,
+      "step": 2326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718225718182475e-05,
+      "loss": 0.6621,
+      "step": 2327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717916557917206e-05,
+      "loss": 26.6646,
+      "step": 2328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717607233063994e-05,
+      "loss": 23.7741,
+      "step": 2329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71729774363363e-05,
+      "loss": 26.4967,
+      "step": 2330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71698808963691e-05,
+      "loss": 0.6765,
+      "step": 2331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716678271084638e-05,
+      "loss": 23.3232,
+      "step": 2332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716368287987623e-05,
+      "loss": 27.2727,
+      "step": 2333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716058140356677e-05,
+      "loss": 26.7174,
+      "step": 2334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715747828202621e-05,
+      "loss": 0.6815,
+      "step": 2335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715437351536279e-05,
+      "loss": 0.7892,
+      "step": 2336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715126710368484e-05,
+      "loss": 0.757,
+      "step": 2337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714815904710071e-05,
+      "loss": 24.8226,
+      "step": 2338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714504934571887e-05,
+      "loss": 24.2206,
+      "step": 2339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714193799964773e-05,
+      "loss": 25.0937,
+      "step": 2340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71388250089959e-05,
+      "loss": 25.6647,
+      "step": 2341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713571037387195e-05,
+      "loss": 30.3476,
+      "step": 2342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713259409438453e-05,
+      "loss": 23.6328,
+      "step": 2343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712947617064237e-05,
+      "loss": 0.7615,
+      "step": 2344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712635660275423e-05,
+      "loss": 25.7071,
+      "step": 2345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712323539082895e-05,
+      "loss": 0.759,
+      "step": 2346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712011253497539e-05,
+      "loss": 27.3778,
+      "step": 2347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711698803530254e-05,
+      "loss": 0.6563,
+      "step": 2348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711386189191934e-05,
+      "loss": 25.7698,
+      "step": 2349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711073410493489e-05,
+      "loss": 27.7958,
+      "step": 2350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71076046744583e-05,
+      "loss": 26.3421,
+      "step": 2351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710447360059873e-05,
+      "loss": 0.7233,
+      "step": 2352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710134088346543e-05,
+      "loss": 26.8393,
+      "step": 2353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709820652316766e-05,
+      "loss": 26.6431,
+      "step": 2354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709507051981479e-05,
+      "loss": 0.7539,
+      "step": 2355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70919328735162e-05,
+      "loss": 0.7352,
+      "step": 2356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708879358438137e-05,
+      "loss": 24.7958,
+      "step": 2357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708565265251979e-05,
+      "loss": 0.7297,
+      "step": 2358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708251007804106e-05,
+      "loss": 0.7332,
+      "step": 2359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70793658610548e-05,
+      "loss": 0.7708,
+      "step": 2360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70762200016707e-05,
+      "loss": 26.1928,
+      "step": 2361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707307249999849e-05,
+      "loss": 0.7824,
+      "step": 2362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7069923356148e-05,
+      "loss": 23.738,
+      "step": 2363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706677257022906e-05,
+      "loss": 25.3902,
+      "step": 2364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706362014235162e-05,
+      "loss": 26.209,
+      "step": 2365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706046607262564e-05,
+      "loss": 25.4773,
+      "step": 2366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705731036116114e-05,
+      "loss": 0.7482,
+      "step": 2367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705415300806822e-05,
+      "loss": 0.7407,
+      "step": 2368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705099401345704e-05,
+      "loss": 25.8415,
+      "step": 2369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704783337743779e-05,
+      "loss": 24.4923,
+      "step": 2370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704467110012072e-05,
+      "loss": 0.6934,
+      "step": 2371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704150718161617e-05,
+      "loss": 24.9586,
+      "step": 2372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70383416220345e-05,
+      "loss": 25.128,
+      "step": 2373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703517442148615e-05,
+      "loss": 0.7658,
+      "step": 2374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70320055800816e-05,
+      "loss": 27.2921,
+      "step": 2375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702883509793142e-05,
+      "loss": 25.374,
+      "step": 2376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70256629751462e-05,
+      "loss": 27.2952,
+      "step": 2377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70224892118366e-05,
+      "loss": 25.0278,
+      "step": 2378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701931380811334e-05,
+      "loss": 29.2777,
+      "step": 2379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70161367640872e-05,
+      "loss": 0.6805,
+      "step": 2380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7012958079869e-05,
+      "loss": 23.3688,
+      "step": 2381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700977775556967e-05,
+      "loss": 0.8085,
+      "step": 2382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70065957913001e-05,
+      "loss": 22.7425,
+      "step": 2383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700341218717134e-05,
+      "loss": 0.7483,
+      "step": 2384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700022694329444e-05,
+      "loss": 25.8566,
+      "step": 2385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699704005978049e-05,
+      "loss": 0.7648,
+      "step": 2386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699385153674072e-05,
+      "loss": 26.0,
+      "step": 2387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699066137428634e-05,
+      "loss": 0.7521,
+      "step": 2388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698746957252863e-05,
+      "loss": 26.8015,
+      "step": 2389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698427613157894e-05,
+      "loss": 25.7704,
+      "step": 2390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69810810515487e-05,
+      "loss": 27.7591,
+      "step": 2391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697788433254934e-05,
+      "loss": 23.7914,
+      "step": 2392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69746859746924e-05,
+      "loss": 0.7463,
+      "step": 2393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697148597808945e-05,
+      "loss": 28.68,
+      "step": 2394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696828434285213e-05,
+      "loss": 25.1019,
+      "step": 2395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696508106909212e-05,
+      "loss": 0.7579,
+      "step": 2396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69618761569212e-05,
+      "loss": 0.7378,
+      "step": 2397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695866960645113e-05,
+      "loss": 0.7494,
+      "step": 2398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69554614177938e-05,
+      "loss": 26.261,
+      "step": 2399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695225159106111e-05,
+      "loss": 25.2403,
+      "step": 2400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694904012636508e-05,
+      "loss": 23.7765,
+      "step": 2401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694582702381769e-05,
+      "loss": 25.434,
+      "step": 2402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694261228353108e-05,
+      "loss": 0.7866,
+      "step": 2403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693939590561736e-05,
+      "loss": 28.2108,
+      "step": 2404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693617789018877e-05,
+      "loss": 24.8348,
+      "step": 2405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693295823735753e-05,
+      "loss": 26.1203,
+      "step": 2406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692973694723599e-05,
+      "loss": 0.8031,
+      "step": 2407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692651401993653e-05,
+      "loss": 0.7548,
+      "step": 2408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692328945557158e-05,
+      "loss": 25.604,
+      "step": 2409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692006325425363e-05,
+      "loss": 0.7651,
+      "step": 2410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69168354160952e-05,
+      "loss": 25.8304,
+      "step": 2411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691360594120894e-05,
+      "loss": 23.3009,
+      "step": 2412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691037482970748e-05,
+      "loss": 24.5045,
+      "step": 2413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690714208170358e-05,
+      "loss": 0.7545,
+      "step": 2414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690390769730997e-05,
+      "loss": 23.6361,
+      "step": 2415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69006716766395e-05,
+      "loss": 0.7258,
+      "step": 2416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689743401980507e-05,
+      "loss": 25.6763,
+      "step": 2417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689419472691963e-05,
+      "loss": 0.795,
+      "step": 2418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689095379809616e-05,
+      "loss": 0.7157,
+      "step": 2419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688771123344776e-05,
+      "loss": 26.0368,
+      "step": 2420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688446703308751e-05,
+      "loss": 0.7294,
+      "step": 2421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688122119712863e-05,
+      "loss": 25.2604,
+      "step": 2422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68779737256843e-05,
+      "loss": 26.0177,
+      "step": 2423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687472461886786e-05,
+      "loss": 26.0591,
+      "step": 2424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687147387679263e-05,
+      "loss": 24.2912,
+      "step": 2425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686822149957202e-05,
+      "loss": 0.7348,
+      "step": 2426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686496748731949e-05,
+      "loss": 26.3947,
+      "step": 2427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686171184014856e-05,
+      "loss": 27.1248,
+      "step": 2428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685845455817281e-05,
+      "loss": 26.1699,
+      "step": 2429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685519564150588e-05,
+      "loss": 0.7694,
+      "step": 2430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685193509026145e-05,
+      "loss": 26.0887,
+      "step": 2431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684867290455324e-05,
+      "loss": 27.231,
+      "step": 2432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684540908449509e-05,
+      "loss": 23.3026,
+      "step": 2433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684214363020086e-05,
+      "loss": 0.7055,
+      "step": 2434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683887654178445e-05,
+      "loss": 25.805,
+      "step": 2435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683560781935986e-05,
+      "loss": 26.4864,
+      "step": 2436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683233746304108e-05,
+      "loss": 0.7335,
+      "step": 2437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682906547294224e-05,
+      "loss": 24.7891,
+      "step": 2438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682579184917746e-05,
+      "loss": 25.738,
+      "step": 2439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682251659186095e-05,
+      "loss": 24.725,
+      "step": 2440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681923970110698e-05,
+      "loss": 25.7699,
+      "step": 2441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681596117702984e-05,
+      "loss": 23.7379,
+      "step": 2442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681268101974394e-05,
+      "loss": 24.8924,
+      "step": 2443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68093992293637e-05,
+      "loss": 24.2469,
+      "step": 2444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680611580600359e-05,
+      "loss": 0.6925,
+      "step": 2445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680283074977816e-05,
+      "loss": 26.0261,
+      "step": 2446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679954406080202e-05,
+      "loss": 24.7223,
+      "step": 2447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679625573918982e-05,
+      "loss": 24.768,
+      "step": 2448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679296578505628e-05,
+      "loss": 0.7124,
+      "step": 2449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678967419851619e-05,
+      "loss": 26.7583,
+      "step": 2450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678638097968434e-05,
+      "loss": 25.4214,
+      "step": 2451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678308612867566e-05,
+      "loss": 0.7249,
+      "step": 2452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677978964560505e-05,
+      "loss": 0.7006,
+      "step": 2453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677649153058754e-05,
+      "loss": 25.6346,
+      "step": 2454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677319178373819e-05,
+      "loss": 0.7477,
+      "step": 2455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676989040517209e-05,
+      "loss": 24.924,
+      "step": 2456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676658739500442e-05,
+      "loss": 23.9507,
+      "step": 2457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676328275335043e-05,
+      "loss": 0.7581,
+      "step": 2458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675997648032538e-05,
+      "loss": 0.8124,
+      "step": 2459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67566685760446e-05,
+      "loss": 0.8006,
+      "step": 2460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675335904062353e-05,
+      "loss": 0.7159,
+      "step": 2461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67500478741776e-05,
+      "loss": 0.7454,
+      "step": 2462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674673507682231e-05,
+      "loss": 0.7282,
+      "step": 2463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674342064867326e-05,
+      "loss": 27.922,
+      "step": 2464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674010458984606e-05,
+      "loss": 0.7022,
+      "step": 2465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673678690045639e-05,
+      "loss": 0.7684,
+      "step": 2466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673346758061999e-05,
+      "loss": 28.1069,
+      "step": 2467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673014663045267e-05,
+      "loss": 0.834,
+      "step": 2468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672682405007027e-05,
+      "loss": 25.1648,
+      "step": 2469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672349983958871e-05,
+      "loss": 25.6916,
+      "step": 2470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672017399912394e-05,
+      "loss": 27.5486,
+      "step": 2471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671684652879201e-05,
+      "loss": 28.7848,
+      "step": 2472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671351742870898e-05,
+      "loss": 24.7769,
+      "step": 2473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671018669899098e-05,
+      "loss": 0.7634,
+      "step": 2474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670685433975426e-05,
+      "loss": 24.4919,
+      "step": 2475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6703520351115e-05,
+      "loss": 26.6198,
+      "step": 2476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670018473318955e-05,
+      "loss": 0.7781,
+      "step": 2477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669684748609429e-05,
+      "loss": 26.8206,
+      "step": 2478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66935086099456e-05,
+      "loss": 0.7642,
+      "step": 2479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669016810485999e-05,
+      "loss": 0.6836,
+      "step": 2480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668682597095395e-05,
+      "loss": 0.7652,
+      "step": 2481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668348220834413e-05,
+      "loss": 0.7756,
+      "step": 2482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668013681714717e-05,
+      "loss": 0.7611,
+      "step": 2483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667678979747974e-05,
+      "loss": 0.7412,
+      "step": 2484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667344114945864e-05,
+      "loss": 24.7463,
+      "step": 2485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66700908732007e-05,
+      "loss": 0.6551,
+      "step": 2486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666673896882272e-05,
+      "loss": 26.7908,
+      "step": 2487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666338543644173e-05,
+      "loss": 0.7222,
+      "step": 2488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666003027617465e-05,
+      "loss": 0.7468,
+      "step": 2489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665667348813857e-05,
+      "loss": 0.7647,
+      "step": 2490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665331507245058e-05,
+      "loss": 0.7052,
+      "step": 2491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664995502922784e-05,
+      "loss": 26.7133,
+      "step": 2492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664659335858755e-05,
+      "loss": 0.7556,
+      "step": 2493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664323006064702e-05,
+      "loss": 25.2475,
+      "step": 2494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663986513552354e-05,
+      "loss": 25.3451,
+      "step": 2495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663649858333455e-05,
+      "loss": 0.7835,
+      "step": 2496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663313040419743e-05,
+      "loss": 0.7149,
+      "step": 2497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662976059822974e-05,
+      "loss": 0.721,
+      "step": 2498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662638916554901e-05,
+      "loss": 28.8741,
+      "step": 2499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662301610627286e-05,
+      "loss": 29.9015,
+      "step": 2500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661964142051896e-05,
+      "loss": 0.7527,
+      "step": 2501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661626510840503e-05,
+      "loss": 0.7248,
+      "step": 2502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661288717004888e-05,
+      "loss": 26.0446,
+      "step": 2503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660950760556833e-05,
+      "loss": 28.5668,
+      "step": 2504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660612641508127e-05,
+      "loss": 23.5159,
+      "step": 2505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660274359870569e-05,
+      "loss": 0.7802,
+      "step": 2506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659935915655958e-05,
+      "loss": 0.8063,
+      "step": 2507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659597308876101e-05,
+      "loss": 0.6973,
+      "step": 2508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659258539542808e-05,
+      "loss": 27.8553,
+      "step": 2509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658919607667903e-05,
+      "loss": 25.4785,
+      "step": 2510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658580513263204e-05,
+      "loss": 26.0032,
+      "step": 2511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658241256340545e-05,
+      "loss": 27.4914,
+      "step": 2512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65790183691176e-05,
+      "loss": 0.744,
+      "step": 2513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657562254988688e-05,
+      "loss": 0.7771,
+      "step": 2514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.657222510583178e-05,
+      "loss": 25.4725,
+      "step": 2515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65688260370708e-05,
+      "loss": 27.3911,
+      "step": 2516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656542534372255e-05,
+      "loss": 0.7036,
+      "step": 2517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656202302590562e-05,
+      "loss": 25.7038,
+      "step": 2518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655861908373876e-05,
+      "loss": 25.4225,
+      "step": 2519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655521351734066e-05,
+      "loss": 26.9486,
+      "step": 2520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655180632683017e-05,
+      "loss": 0.7588,
+      "step": 2521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654839751232611e-05,
+      "loss": 26.6005,
+      "step": 2522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654498707394746e-05,
+      "loss": 25.6535,
+      "step": 2523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654157501181313e-05,
+      "loss": 0.728,
+      "step": 2524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65381613260422e-05,
+      "loss": 25.4423,
+      "step": 2525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653474601675374e-05,
+      "loss": 0.7284,
+      "step": 2526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653132908406688e-05,
+      "loss": 24.7694,
+      "step": 2527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652791052810087e-05,
+      "loss": 23.9389,
+      "step": 2528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652449034897491e-05,
+      "loss": 25.3277,
+      "step": 2529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652106854680836e-05,
+      "loss": 0.7173,
+      "step": 2530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651764512172055e-05,
+      "loss": 26.4984,
+      "step": 2531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651422007383096e-05,
+      "loss": 0.7538,
+      "step": 2532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651079340325904e-05,
+      "loss": 0.6684,
+      "step": 2533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650736511012435e-05,
+      "loss": 0.7386,
+      "step": 2534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650393519454646e-05,
+      "loss": 26.6737,
+      "step": 2535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650050365664506e-05,
+      "loss": 0.7641,
+      "step": 2536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649707049653985e-05,
+      "loss": 25.8293,
+      "step": 2537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649363571435059e-05,
+      "loss": 0.7337,
+      "step": 2538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64901993101971e-05,
+      "loss": 26.3177,
+      "step": 2539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648676128419928e-05,
+      "loss": 24.7766,
+      "step": 2540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648332163647706e-05,
+      "loss": 0.7621,
+      "step": 2541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647988036715042e-05,
+      "loss": 0.6818,
+      "step": 2542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647643747633943e-05,
+      "loss": 0.7567,
+      "step": 2543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647299296416421e-05,
+      "loss": 24.5946,
+      "step": 2544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64695468307449e-05,
+      "loss": 22.7793,
+      "step": 2545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646609907620172e-05,
+      "loss": 25.1843,
+      "step": 2546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646264970065496e-05,
+      "loss": 0.6997,
+      "step": 2547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645919870422493e-05,
+      "loss": 0.694,
+      "step": 2548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645574608703206e-05,
+      "loss": 24.2618,
+      "step": 2549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645229184919678e-05,
+      "loss": 0.664,
+      "step": 2550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644883599083958e-05,
+      "loss": 24.0345,
+      "step": 2551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644537851208103e-05,
+      "loss": 26.0136,
+      "step": 2552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644191941304175e-05,
+      "loss": 24.6402,
+      "step": 2553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64384586938424e-05,
+      "loss": 0.7057,
+      "step": 2554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643499635460374e-05,
+      "loss": 22.6816,
+      "step": 2555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643153239544653e-05,
+      "loss": 0.6657,
+      "step": 2556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642806681649161e-05,
+      "loss": 23.4375,
+      "step": 2557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64245996178599e-05,
+      "loss": 26.5524,
+      "step": 2558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642113079967233e-05,
+      "loss": 0.7018,
+      "step": 2559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641766036204994e-05,
+      "loss": 26.789,
+      "step": 2560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641418830511377e-05,
+      "loss": 24.3644,
+      "step": 2561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641071462898497e-05,
+      "loss": 23.8914,
+      "step": 2562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640723933378471e-05,
+      "loss": 25.5652,
+      "step": 2563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640376241963423e-05,
+      "loss": 0.7578,
+      "step": 2564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640028388665482e-05,
+      "loss": 25.5007,
+      "step": 2565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639680373496784e-05,
+      "loss": 24.0986,
+      "step": 2566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63933219646947e-05,
+      "loss": 28.1547,
+      "step": 2567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638983857595684e-05,
+      "loss": 0.7222,
+      "step": 2568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638635356887582e-05,
+      "loss": 25.0185,
+      "step": 2569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638286694357319e-05,
+      "loss": 0.7337,
+      "step": 2570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637937870017058e-05,
+      "loss": 21.8088,
+      "step": 2571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637588883878969e-05,
+      "loss": 0.6834,
+      "step": 2572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637239735955226e-05,
+      "loss": 0.7215,
+      "step": 2573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636890426258011e-05,
+      "loss": 0.7,
+      "step": 2574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636540954799506e-05,
+      "loss": 26.5222,
+      "step": 2575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636191321591908e-05,
+      "loss": 0.7747,
+      "step": 2576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635841526647411e-05,
+      "loss": 24.6788,
+      "step": 2577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635491569978219e-05,
+      "loss": 0.7395,
+      "step": 2578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635141451596538e-05,
+      "loss": 0.7512,
+      "step": 2579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634791171514585e-05,
+      "loss": 0.7568,
+      "step": 2580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634440729744578e-05,
+      "loss": 23.1569,
+      "step": 2581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634090126298742e-05,
+      "loss": 26.4955,
+      "step": 2582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633739361189312e-05,
+      "loss": 27.3704,
+      "step": 2583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633388434428521e-05,
+      "loss": 28.1524,
+      "step": 2584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63303734602861e-05,
+      "loss": 25.1967,
+      "step": 2585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632686096001832e-05,
+      "loss": 26.2577,
+      "step": 2586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632334684360437e-05,
+      "loss": 27.0667,
+      "step": 2587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631983111116685e-05,
+      "loss": 0.7619,
+      "step": 2588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631631376282839e-05,
+      "loss": 0.7977,
+      "step": 2589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631279479871173e-05,
+      "loss": 0.7747,
+      "step": 2590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630927421893963e-05,
+      "loss": 0.7732,
+      "step": 2591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630575202363487e-05,
+      "loss": 0.7163,
+      "step": 2592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630222821292036e-05,
+      "loss": 0.7618,
+      "step": 2593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629870278691901e-05,
+      "loss": 0.7986,
+      "step": 2594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629517574575385e-05,
+      "loss": 30.2345,
+      "step": 2595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629164708954786e-05,
+      "loss": 29.8168,
+      "step": 2596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628811681842415e-05,
+      "loss": 0.8144,
+      "step": 2597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628458493250593e-05,
+      "loss": 29.525,
+      "step": 2598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628105143191637e-05,
+      "loss": 27.5063,
+      "step": 2599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627751631677873e-05,
+      "loss": 28.54,
+      "step": 2600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627397958721637e-05,
+      "loss": 27.5353,
+      "step": 2601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627044124335264e-05,
+      "loss": 0.7165,
+      "step": 2602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626690128531101e-05,
+      "loss": 0.7509,
+      "step": 2603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626335971321494e-05,
+      "loss": 0.7054,
+      "step": 2604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6259816527188e-05,
+      "loss": 28.3803,
+      "step": 2605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625627172735379e-05,
+      "loss": 0.7603,
+      "step": 2606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625272531383597e-05,
+      "loss": 0.7169,
+      "step": 2607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624917728675826e-05,
+      "loss": 0.7927,
+      "step": 2608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624562764624445e-05,
+      "loss": 0.7203,
+      "step": 2609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624207639241836e-05,
+      "loss": 25.3186,
+      "step": 2610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623852352540386e-05,
+      "loss": 25.1926,
+      "step": 2611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623496904532494e-05,
+      "loss": 24.1828,
+      "step": 2612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623141295230555e-05,
+      "loss": 0.7482,
+      "step": 2613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62278552464698e-05,
+      "loss": 26.9633,
+      "step": 2614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622429592794176e-05,
+      "loss": 0.783,
+      "step": 2615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622073499684561e-05,
+      "loss": 0.7329,
+      "step": 2616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621717245330557e-05,
+      "loss": 0.7355,
+      "step": 2617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621360829744595e-05,
+      "loss": 27.4801,
+      "step": 2618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621004252939105e-05,
+      "loss": 25.7342,
+      "step": 2619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62064751492653e-05,
+      "loss": 0.7542,
+      "step": 2620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620290615719313e-05,
+      "loss": 25.372,
+      "step": 2621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619933555329905e-05,
+      "loss": 25.0911,
+      "step": 2622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619576333770761e-05,
+      "loss": 22.6756,
+      "step": 2623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619218951054348e-05,
+      "loss": 25.737,
+      "step": 2624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618861407193127e-05,
+      "loss": 0.7984,
+      "step": 2625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618503702199574e-05,
+      "loss": 25.4285,
+      "step": 2626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618145836086168e-05,
+      "loss": 23.8479,
+      "step": 2627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617787808865395e-05,
+      "loss": 26.4619,
+      "step": 2628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617429620549742e-05,
+      "loss": 25.7957,
+      "step": 2629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617071271151705e-05,
+      "loss": 22.4936,
+      "step": 2630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616712760683789e-05,
+      "loss": 26.9272,
+      "step": 2631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616354089158496e-05,
+      "loss": 0.7367,
+      "step": 2632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615995256588342e-05,
+      "loss": 23.8771,
+      "step": 2633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615636262985843e-05,
+      "loss": 27.2665,
+      "step": 2634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615277108363523e-05,
+      "loss": 0.7734,
+      "step": 2635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614917792733913e-05,
+      "loss": 27.3319,
+      "step": 2636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614558316109547e-05,
+      "loss": 25.7758,
+      "step": 2637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614198678502965e-05,
+      "loss": 0.6806,
+      "step": 2638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613838879926713e-05,
+      "loss": 0.7077,
+      "step": 2639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613478920393344e-05,
+      "loss": 0.726,
+      "step": 2640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613118799915417e-05,
+      "loss": 26.9984,
+      "step": 2641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612758518505491e-05,
+      "loss": 0.7377,
+      "step": 2642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612398076176138e-05,
+      "loss": 24.7322,
+      "step": 2643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612037472939929e-05,
+      "loss": 0.6762,
+      "step": 2644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61167670880945e-05,
+      "loss": 25.8914,
+      "step": 2645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611315783797279e-05,
+      "loss": 23.8271,
+      "step": 2646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610954697916011e-05,
+      "loss": 0.7173,
+      "step": 2647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610593451178244e-05,
+      "loss": 26.4709,
+      "step": 2648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610232043596576e-05,
+      "loss": 0.7128,
+      "step": 2649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60987047518362e-05,
+      "loss": 23.4127,
+      "step": 2650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609508745951987e-05,
+      "loss": 25.2772,
+      "step": 2651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609146855914297e-05,
+      "loss": 0.7102,
+      "step": 2652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608784805083172e-05,
+      "loss": 0.6737,
+      "step": 2653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608422593471246e-05,
+      "loss": 24.0883,
+      "step": 2654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608060221091155e-05,
+      "loss": 0.7178,
+      "step": 2655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607697687955536e-05,
+      "loss": 0.8026,
+      "step": 2656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607334994077044e-05,
+      "loss": 25.0374,
+      "step": 2657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606972139468324e-05,
+      "loss": 24.6711,
+      "step": 2658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606609124142038e-05,
+      "loss": 23.4986,
+      "step": 2659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606245948110852e-05,
+      "loss": 0.7221,
+      "step": 2660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605882611387432e-05,
+      "loss": 24.06,
+      "step": 2661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605519113984455e-05,
+      "loss": 0.7202,
+      "step": 2662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605155455914602e-05,
+      "loss": 25.4737,
+      "step": 2663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604791637190557e-05,
+      "loss": 24.8438,
+      "step": 2664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604427657825017e-05,
+      "loss": 24.4842,
+      "step": 2665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604063517830677e-05,
+      "loss": 26.6492,
+      "step": 2666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603699217220239e-05,
+      "loss": 26.4767,
+      "step": 2667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603334756006412e-05,
+      "loss": 24.762,
+      "step": 2668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602970134201914e-05,
+      "loss": 24.1286,
+      "step": 2669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60260535181946e-05,
+      "loss": 28.2884,
+      "step": 2670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60224040887178e-05,
+      "loss": 25.7955,
+      "step": 2671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601875305371604e-05,
+      "loss": 25.5088,
+      "step": 2672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601510041331668e-05,
+      "loss": 23.1101,
+      "step": 2673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601144616764714e-05,
+      "loss": 25.3362,
+      "step": 2674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600779031683493e-05,
+      "loss": 25.7675,
+      "step": 2675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600413286100757e-05,
+      "loss": 26.3067,
+      "step": 2676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600047380029265e-05,
+      "loss": 23.8595,
+      "step": 2677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599681313481781e-05,
+      "loss": 0.6853,
+      "step": 2678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599315086471077e-05,
+      "loss": 25.7925,
+      "step": 2679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598948699009928e-05,
+      "loss": 25.9573,
+      "step": 2680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598582151111117e-05,
+      "loss": 24.2635,
+      "step": 2681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59821544278743e-05,
+      "loss": 24.1665,
+      "step": 2682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597848574051662e-05,
+      "loss": 23.2023,
+      "step": 2683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597481544916609e-05,
+      "loss": 23.7387,
+      "step": 2684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597114355395076e-05,
+      "loss": 0.7195,
+      "step": 2685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596747005499873e-05,
+      "loss": 26.3803,
+      "step": 2686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596379495243816e-05,
+      "loss": 0.782,
+      "step": 2687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.596011824639723e-05,
+      "loss": 23.2306,
+      "step": 2688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595643993700425e-05,
+      "loss": 24.2817,
+      "step": 2689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595276002438749e-05,
+      "loss": 0.7183,
+      "step": 2690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594907850867536e-05,
+      "loss": 25.2062,
+      "step": 2691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594539538999627e-05,
+      "loss": 0.7313,
+      "step": 2692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594171066847874e-05,
+      "loss": 0.7553,
+      "step": 2693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593802434425128e-05,
+      "loss": 23.9224,
+      "step": 2694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593433641744252e-05,
+      "loss": 24.6331,
+      "step": 2695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59306468881811e-05,
+      "loss": 0.8016,
+      "step": 2696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592695575659571e-05,
+      "loss": 26.0859,
+      "step": 2697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592326302281517e-05,
+      "loss": 24.5798,
+      "step": 2698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591956868696826e-05,
+      "loss": 0.757,
+      "step": 2699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591587274918388e-05,
+      "loss": 20.9126,
+      "step": 2700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591217520959095e-05,
+      "loss": 0.7307,
+      "step": 2701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590847606831849e-05,
+      "loss": 25.2674,
+      "step": 2702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59047753254955e-05,
+      "loss": 25.7206,
+      "step": 2703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590107298125114e-05,
+      "loss": 24.0035,
+      "step": 2704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589736903571454e-05,
+      "loss": 25.2938,
+      "step": 2705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58936634890149e-05,
+      "loss": 21.8822,
+      "step": 2706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58899563412815e-05,
+      "loss": 26.1576,
+      "step": 2707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588624759264369e-05,
+      "loss": 23.909,
+      "step": 2708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588253724323082e-05,
+      "loss": 22.9329,
+      "step": 2709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587882529317235e-05,
+      "loss": 0.7387,
+      "step": 2710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587511174259777e-05,
+      "loss": 23.4296,
+      "step": 2711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587139659163662e-05,
+      "loss": 25.8491,
+      "step": 2712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586767984041852e-05,
+      "loss": 24.6877,
+      "step": 2713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586396148907313e-05,
+      "loss": 0.7218,
+      "step": 2714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586024153773015e-05,
+      "loss": 24.5856,
+      "step": 2715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585651998651936e-05,
+      "loss": 0.7333,
+      "step": 2716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58527968355706e-05,
+      "loss": 26.1974,
+      "step": 2717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584907208501376e-05,
+      "loss": 23.9249,
+      "step": 2718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584534573497876e-05,
+      "loss": 0.7123,
+      "step": 2719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584161778559561e-05,
+      "loss": 23.5368,
+      "step": 2720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583788823699436e-05,
+      "loss": 22.5169,
+      "step": 2721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58341570893051e-05,
+      "loss": 0.7166,
+      "step": 2722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583042434265803e-05,
+      "loss": 0.7092,
+      "step": 2723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582668999718333e-05,
+      "loss": 24.2053,
+      "step": 2724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582295405301131e-05,
+      "loss": 0.7347,
+      "step": 2725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581921651027227e-05,
+      "loss": 25.5166,
+      "step": 2726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581547736909663e-05,
+      "loss": 0.7359,
+      "step": 2727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581173662961482e-05,
+      "loss": 23.8469,
+      "step": 2728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580799429195732e-05,
+      "loss": 0.7465,
+      "step": 2729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580425035625468e-05,
+      "loss": 0.7514,
+      "step": 2730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580050482263756e-05,
+      "loss": 26.3736,
+      "step": 2731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579675769123659e-05,
+      "loss": 27.8258,
+      "step": 2732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579300896218249e-05,
+      "loss": 0.7099,
+      "step": 2733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578925863560604e-05,
+      "loss": 0.6986,
+      "step": 2734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578550671163807e-05,
+      "loss": 25.4587,
+      "step": 2735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578175319040948e-05,
+      "loss": 26.8425,
+      "step": 2736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577799807205121e-05,
+      "loss": 0.7305,
+      "step": 2737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577424135669426e-05,
+      "loss": 27.1565,
+      "step": 2738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577048304446967e-05,
+      "loss": 22.1865,
+      "step": 2739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576672313550858e-05,
+      "loss": 22.3111,
+      "step": 2740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576296162994212e-05,
+      "loss": 24.4151,
+      "step": 2741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575919852790156e-05,
+      "loss": 0.8148,
+      "step": 2742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575543382951815e-05,
+      "loss": 0.7187,
+      "step": 2743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575166753492322e-05,
+      "loss": 24.7863,
+      "step": 2744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574789964424818e-05,
+      "loss": 0.7373,
+      "step": 2745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574413015762445e-05,
+      "loss": 24.7883,
+      "step": 2746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574035907518354e-05,
+      "loss": 0.7657,
+      "step": 2747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573658639705703e-05,
+      "loss": 24.5392,
+      "step": 2748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57328121233765e-05,
+      "loss": 0.7223,
+      "step": 2749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572903625427364e-05,
+      "loss": 26.6639,
+      "step": 2750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572525878988015e-05,
+      "loss": 0.8277,
+      "step": 2751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572147973032785e-05,
+      "loss": 0.7198,
+      "step": 2752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571769907574853e-05,
+      "loss": 0.746,
+      "step": 2753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571391682627412e-05,
+      "loss": 0.6753,
+      "step": 2754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571013298203655e-05,
+      "loss": 0.8213,
+      "step": 2755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570634754316781e-05,
+      "loss": 25.8691,
+      "step": 2756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570256050979996e-05,
+      "loss": 0.7444,
+      "step": 2757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569877188206515e-05,
+      "loss": 0.7082,
+      "step": 2758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569498166009549e-05,
+      "loss": 23.3543,
+      "step": 2759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569118984402325e-05,
+      "loss": 25.2786,
+      "step": 2760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56873964339807e-05,
+      "loss": 25.98,
+      "step": 2761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568360143010017e-05,
+      "loss": 21.8601,
+      "step": 2762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567980483251405e-05,
+      "loss": 0.7211,
+      "step": 2763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567600664135478e-05,
+      "loss": 0.7534,
+      "step": 2764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56722068567549e-05,
+      "loss": 0.7747,
+      "step": 2765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566840547884693e-05,
+      "loss": 25.2575,
+      "step": 2766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566460250776347e-05,
+      "loss": 0.7155,
+      "step": 2767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566079794363725e-05,
+      "loss": 24.4327,
+      "step": 2768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565699178660095e-05,
+      "loss": 24.2331,
+      "step": 2769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565318403678736e-05,
+      "loss": 0.7144,
+      "step": 2770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564937469432931e-05,
+      "loss": 0.7952,
+      "step": 2771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56455637593597e-05,
+      "loss": 24.0087,
+      "step": 2772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564175123201147e-05,
+      "loss": 26.35,
+      "step": 2773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563793711241763e-05,
+      "loss": 25.7388,
+      "step": 2774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563412140071125e-05,
+      "loss": 25.5292,
+      "step": 2775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56303040970254e-05,
+      "loss": 24.3156,
+      "step": 2776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56264852014933e-05,
+      "loss": 0.803,
+      "step": 2777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562266471424816e-05,
+      "loss": 23.7408,
+      "step": 2778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561884263542325e-05,
+      "loss": 24.2499,
+      "step": 2779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56150189651519e-05,
+      "loss": 24.7022,
+      "step": 2780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561119370356753e-05,
+      "loss": 22.63,
+      "step": 2781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560736685080355e-05,
+      "loss": 0.7657,
+      "step": 2782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56035384069935e-05,
+      "loss": 24.3331,
+      "step": 2783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55997083722709e-05,
+      "loss": 22.7514,
+      "step": 2784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559587674676939e-05,
+      "loss": 24.2785,
+      "step": 2785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559204353062266e-05,
+      "loss": 23.8933,
+      "step": 2786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558820872396438e-05,
+      "loss": 24.5194,
+      "step": 2787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558437232692836e-05,
+      "loss": 0.7128,
+      "step": 2788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558053433964845e-05,
+      "loss": 25.7721,
+      "step": 2789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55766947622585e-05,
+      "loss": 26.3272,
+      "step": 2790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557285359489248e-05,
+      "loss": 23.7261,
+      "step": 2791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556901083768441e-05,
+      "loss": 22.3702,
+      "step": 2792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556516649076831e-05,
+      "loss": 26.0979,
+      "step": 2793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556132055427833e-05,
+      "loss": 22.6495,
+      "step": 2794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555747302834861e-05,
+      "loss": 20.2171,
+      "step": 2795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555362391311338e-05,
+      "loss": 0.7711,
+      "step": 2796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554977320870693e-05,
+      "loss": 22.2334,
+      "step": 2797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554592091526358e-05,
+      "loss": 25.7042,
+      "step": 2798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554206703291775e-05,
+      "loss": 0.7849,
+      "step": 2799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553821156180385e-05,
+      "loss": 0.7442,
+      "step": 2800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553435450205638e-05,
+      "loss": 0.7121,
+      "step": 2801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553049585380993e-05,
+      "loss": 23.3754,
+      "step": 2802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552663561719907e-05,
+      "loss": 24.0977,
+      "step": 2803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552277379235853e-05,
+      "loss": 0.746,
+      "step": 2804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551891037942296e-05,
+      "loss": 20.9255,
+      "step": 2805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551504537852719e-05,
+      "loss": 24.4124,
+      "step": 2806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551117878980601e-05,
+      "loss": 23.5739,
+      "step": 2807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550731061339435e-05,
+      "loss": 23.282,
+      "step": 2808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550344084942712e-05,
+      "loss": 25.7712,
+      "step": 2809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549956949803936e-05,
+      "loss": 0.769,
+      "step": 2810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549569655936607e-05,
+      "loss": 0.6985,
+      "step": 2811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549182203354242e-05,
+      "loss": 0.7352,
+      "step": 2812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548794592070353e-05,
+      "loss": 0.7464,
+      "step": 2813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548406822098464e-05,
+      "loss": 0.6957,
+      "step": 2814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548018893452104e-05,
+      "loss": 0.7324,
+      "step": 2815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547630806144803e-05,
+      "loss": 0.7366,
+      "step": 2816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547242560190102e-05,
+      "loss": 0.7163,
+      "step": 2817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546854155601543e-05,
+      "loss": 24.4015,
+      "step": 2818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546465592392678e-05,
+      "loss": 0.703,
+      "step": 2819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546076870577062e-05,
+      "loss": 0.6664,
+      "step": 2820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545687990168256e-05,
+      "loss": 22.7517,
+      "step": 2821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545298951179826e-05,
+      "loss": 25.4348,
+      "step": 2822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544909753625342e-05,
+      "loss": 24.2207,
+      "step": 2823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544520397518385e-05,
+      "loss": 22.3743,
+      "step": 2824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544130882872535e-05,
+      "loss": 0.7896,
+      "step": 2825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543741209701382e-05,
+      "loss": 23.829,
+      "step": 2826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543351378018519e-05,
+      "loss": 0.7776,
+      "step": 2827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542961387837548e-05,
+      "loss": 0.7177,
+      "step": 2828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542571239172072e-05,
+      "loss": 0.6716,
+      "step": 2829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542180932035701e-05,
+      "loss": 0.7709,
+      "step": 2830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541790466442052e-05,
+      "loss": 0.6693,
+      "step": 2831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541399842404747e-05,
+      "loss": 25.9608,
+      "step": 2832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541009059937414e-05,
+      "loss": 25.9562,
+      "step": 2833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540618119053683e-05,
+      "loss": 26.6795,
+      "step": 2834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540227019767195e-05,
+      "loss": 22.768,
+      "step": 2835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539835762091592e-05,
+      "loss": 24.0472,
+      "step": 2836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539444346040525e-05,
+      "loss": 22.4681,
+      "step": 2837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539052771627647e-05,
+      "loss": 0.6781,
+      "step": 2838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53866103886662e-05,
+      "loss": 0.7045,
+      "step": 2839
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53826914777111e-05,
+      "loss": 0.7895,
+      "step": 2840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537877098354786e-05,
+      "loss": 25.4428,
+      "step": 2841
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537484890631327e-05,
+      "loss": 23.4021,
+      "step": 2842
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537092524614415e-05,
+      "loss": 0.6884,
+      "step": 2843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536700000317737e-05,
+      "loss": 0.7554,
+      "step": 2844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536307317754989e-05,
+      "loss": 25.6622,
+      "step": 2845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535914476939866e-05,
+      "loss": 0.7871,
+      "step": 2846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535521477886078e-05,
+      "loss": 26.3811,
+      "step": 2847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53512832060733e-05,
+      "loss": 24.6711,
+      "step": 2848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534735005117338e-05,
+      "loss": 0.6914,
+      "step": 2849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534341531429829e-05,
+      "loss": 25.0788,
+      "step": 2850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533947899558522e-05,
+      "loss": 26.5063,
+      "step": 2851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533554109517154e-05,
+      "loss": 22.9636,
+      "step": 2852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53316016131946e-05,
+      "loss": 26.0876,
+      "step": 2853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532766054979186e-05,
+      "loss": 23.5354,
+      "step": 2854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532371790510076e-05,
+      "loss": 0.7547,
+      "step": 2855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531977367925888e-05,
+      "loss": 0.742,
+      "step": 2856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531582787240383e-05,
+      "loss": 0.7387,
+      "step": 2857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531188048467322e-05,
+      "loss": 25.3796,
+      "step": 2858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530793151620479e-05,
+      "loss": 0.8044,
+      "step": 2859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530398096713629e-05,
+      "loss": 27.359,
+      "step": 2860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530002883760552e-05,
+      "loss": 0.7259,
+      "step": 2861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529607512775039e-05,
+      "loss": 23.4817,
+      "step": 2862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52921198377088e-05,
+      "loss": 0.7114,
+      "step": 2863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528816296761876e-05,
+      "loss": 0.7378,
+      "step": 2864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528420451761829e-05,
+      "loss": 24.4504,
+      "step": 2865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528024448784546e-05,
+      "loss": 24.8878,
+      "step": 2866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527628287843846e-05,
+      "loss": 0.7907,
+      "step": 2867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527231968953547e-05,
+      "loss": 25.5903,
+      "step": 2868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526835492127475e-05,
+      "loss": 24.8788,
+      "step": 2869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526438857379463e-05,
+      "loss": 24.0816,
+      "step": 2870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526042064723347e-05,
+      "loss": 23.1898,
+      "step": 2871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525645114172968e-05,
+      "loss": 0.7724,
+      "step": 2872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525248005742177e-05,
+      "loss": 0.8243,
+      "step": 2873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524850739444824e-05,
+      "loss": 26.2722,
+      "step": 2874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52445331529477e-05,
+      "loss": 25.539,
+      "step": 2875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524055733305879e-05,
+      "loss": 0.7187,
+      "step": 2876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52365799349202e-05,
+      "loss": 0.7546,
+      "step": 2877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52326009586707e-05,
+      "loss": 21.0384,
+      "step": 2878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522862040444909e-05,
+      "loss": 24.9433,
+      "step": 2879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522463827239423e-05,
+      "loss": 0.7524,
+      "step": 2880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522065456264507e-05,
+      "loss": 0.7532,
+      "step": 2881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521666927534053e-05,
+      "loss": 0.7442,
+      "step": 2882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52126824106197e-05,
+      "loss": 0.7022,
+      "step": 2883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52086939686216e-05,
+      "loss": 25.3052,
+      "step": 2884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520470394948542e-05,
+      "loss": 24.3654,
+      "step": 2885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520071235335035e-05,
+      "loss": 24.7258,
+      "step": 2886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519671918035562e-05,
+      "loss": 23.5546,
+      "step": 2887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519272443064055e-05,
+      "loss": 22.7493,
+      "step": 2888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518872810434449e-05,
+      "loss": 25.5381,
+      "step": 2889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518473020160685e-05,
+      "loss": 23.5921,
+      "step": 2890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518073072256712e-05,
+      "loss": 23.9504,
+      "step": 2891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517672966736482e-05,
+      "loss": 23.3468,
+      "step": 2892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51727270361395e-05,
+      "loss": 25.0157,
+      "step": 2893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516872282903084e-05,
+      "loss": 23.5671,
+      "step": 2894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516471704617848e-05,
+      "loss": 23.133,
+      "step": 2895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516070968772222e-05,
+      "loss": 0.7264,
+      "step": 2896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515670075380181e-05,
+      "loss": 0.7137,
+      "step": 2897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515269024455714e-05,
+      "loss": 25.7491,
+      "step": 2898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514867816012809e-05,
+      "loss": 21.3966,
+      "step": 2899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514466450065465e-05,
+      "loss": 22.2932,
+      "step": 2900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514064926627684e-05,
+      "loss": 23.5673,
+      "step": 2901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513663245713471e-05,
+      "loss": 0.7411,
+      "step": 2902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51326140733684e-05,
+      "loss": 0.7916,
+      "step": 2903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512859411511809e-05,
+      "loss": 24.0165,
+      "step": 2904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512457258252406e-05,
+      "loss": 22.8667,
+      "step": 2905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512054947572654e-05,
+      "loss": 22.2907,
+      "step": 2906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511652479486592e-05,
+      "loss": 23.867,
+      "step": 2907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51124985400826e-05,
+      "loss": 25.1198,
+      "step": 2908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5108470711517e-05,
+      "loss": 25.8084,
+      "step": 2909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510444130930971e-05,
+      "loss": 0.731,
+      "step": 2910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510041033360123e-05,
+      "loss": 0.755,
+      "step": 2911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509637778453221e-05,
+      "loss": 24.5663,
+      "step": 2912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509234366224333e-05,
+      "loss": 0.7949,
+      "step": 2913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508830796687532e-05,
+      "loss": 26.1552,
+      "step": 2914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508427069856896e-05,
+      "loss": 0.8358,
+      "step": 2915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50802318574651e-05,
+      "loss": 0.7821,
+      "step": 2916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507619144370465e-05,
+      "loss": 0.7514,
+      "step": 2917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507214945742854e-05,
+      "loss": 23.699,
+      "step": 2918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506810589877779e-05,
+      "loss": 24.3017,
+      "step": 2919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506406076789347e-05,
+      "loss": 26.1304,
+      "step": 2920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506001406491669e-05,
+      "loss": 0.759,
+      "step": 2921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505596578998861e-05,
+      "loss": 23.3752,
+      "step": 2922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50519159432505e-05,
+      "loss": 23.1714,
+      "step": 2923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504786452484357e-05,
+      "loss": 24.0949,
+      "step": 2924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504381153490921e-05,
+      "loss": 23.0273,
+      "step": 2925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50397569735888e-05,
+      "loss": 26.5694,
+      "step": 2926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503570084102378e-05,
+      "loss": 24.2401,
+      "step": 2927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503164313735566e-05,
+      "loss": 0.8089,
+      "step": 2928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502758386272599e-05,
+      "loss": 23.7164,
+      "step": 2929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502352301727638e-05,
+      "loss": 23.6778,
+      "step": 2930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50194606011485e-05,
+      "loss": 24.967,
+      "step": 2931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501539661448408e-05,
+      "loss": 23.3393,
+      "step": 2932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501133105742487e-05,
+      "loss": 24.7431,
+      "step": 2933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500726393011272e-05,
+      "loss": 24.2836,
+      "step": 2934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50031952326895e-05,
+      "loss": 22.1688,
+      "step": 2935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499912496529715e-05,
+      "loss": 23.9087,
+      "step": 2936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499505312807768e-05,
+      "loss": 22.8195,
+      "step": 2937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499097972117315e-05,
+      "loss": 0.665,
+      "step": 2938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498690474472562e-05,
+      "loss": 25.2295,
+      "step": 2939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498282819887727e-05,
+      "loss": 0.7111,
+      "step": 2940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497875008377033e-05,
+      "loss": 24.177,
+      "step": 2941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497467039954705e-05,
+      "loss": 0.7552,
+      "step": 2942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497058914634975e-05,
+      "loss": 0.7682,
+      "step": 2943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496650632432082e-05,
+      "loss": 26.895,
+      "step": 2944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49624219336027e-05,
+      "loss": 0.6942,
+      "step": 2945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495833597433784e-05,
+      "loss": 0.7424,
+      "step": 2946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495424844666881e-05,
+      "loss": 25.2511,
+      "step": 2947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495015935073822e-05,
+      "loss": 21.8389,
+      "step": 2948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494606868668868e-05,
+      "loss": 0.7477,
+      "step": 2949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494197645466293e-05,
+      "loss": 23.7225,
+      "step": 2950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493788265480371e-05,
+      "loss": 25.8648,
+      "step": 2951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493378728725386e-05,
+      "loss": 0.7198,
+      "step": 2952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492969035215622e-05,
+      "loss": 25.9928,
+      "step": 2953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492559184965372e-05,
+      "loss": 26.5683,
+      "step": 2954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492149177988938e-05,
+      "loss": 0.6913,
+      "step": 2955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491739014300617e-05,
+      "loss": 0.6524,
+      "step": 2956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491328693914722e-05,
+      "loss": 23.5301,
+      "step": 2957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490918216845566e-05,
+      "loss": 23.7545,
+      "step": 2958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490507583107471e-05,
+      "loss": 0.8069,
+      "step": 2959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490096792714757e-05,
+      "loss": 26.333,
+      "step": 2960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489685845681762e-05,
+      "loss": 0.7769,
+      "step": 2961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489274742022815e-05,
+      "loss": 24.8967,
+      "step": 2962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488863481752262e-05,
+      "loss": 24.6267,
+      "step": 2963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48845206488445e-05,
+      "loss": 22.9237,
+      "step": 2964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488040491433729e-05,
+      "loss": 23.3399,
+      "step": 2965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48762876141446e-05,
+      "loss": 0.7755,
+      "step": 2966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487216874841006e-05,
+      "loss": 26.0463,
+      "step": 2967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486804831727733e-05,
+      "loss": 25.2018,
+      "step": 2968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486392632089018e-05,
+      "loss": 0.7148,
+      "step": 2969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485980275939241e-05,
+      "loss": 0.7183,
+      "step": 2970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485567763292786e-05,
+      "loss": 22.934,
+      "step": 2971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485155094164047e-05,
+      "loss": 22.5947,
+      "step": 2972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484742268567416e-05,
+      "loss": 24.8338,
+      "step": 2973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484329286517296e-05,
+      "loss": 24.2912,
+      "step": 2974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483916148028096e-05,
+      "loss": 0.6761,
+      "step": 2975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483502853114226e-05,
+      "loss": 22.8985,
+      "step": 2976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483089401790106e-05,
+      "loss": 23.4508,
+      "step": 2977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482675794070159e-05,
+      "loss": 0.7629,
+      "step": 2978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482262029968814e-05,
+      "loss": 23.0532,
+      "step": 2979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481848109500506e-05,
+      "loss": 22.2141,
+      "step": 2980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481434032679675e-05,
+      "loss": 23.8582,
+      "step": 2981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481019799520765e-05,
+      "loss": 23.1422,
+      "step": 2982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480605410038227e-05,
+      "loss": 0.769,
+      "step": 2983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480190864246519e-05,
+      "loss": 27.4952,
+      "step": 2984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479776162160101e-05,
+      "loss": 0.7899,
+      "step": 2985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47936130379344e-05,
+      "loss": 0.7082,
+      "step": 2986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478946289161011e-05,
+      "loss": 0.7137,
+      "step": 2987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478531118277289e-05,
+      "loss": 23.1836,
+      "step": 2988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47811579115676e-05,
+      "loss": 23.5088,
+      "step": 2989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477700307813912e-05,
+      "loss": 21.3502,
+      "step": 2990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477284668263239e-05,
+      "loss": 25.6828,
+      "step": 2991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476868872519242e-05,
+      "loss": 25.4293,
+      "step": 2992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476452920596425e-05,
+      "loss": 0.6739,
+      "step": 2993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476036812509298e-05,
+      "loss": 0.7747,
+      "step": 2994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475620548272381e-05,
+      "loss": 26.8485,
+      "step": 2995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475204127900194e-05,
+      "loss": 24.2253,
+      "step": 2996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474787551407262e-05,
+      "loss": 0.7281,
+      "step": 2997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474370818808119e-05,
+      "loss": 24.6764,
+      "step": 2998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473953930117304e-05,
+      "loss": 23.5343,
+      "step": 2999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473536885349359e-05,
+      "loss": 23.0915,
+      "step": 3000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473119684518834e-05,
+      "loss": 22.3897,
+      "step": 3001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472702327640282e-05,
+      "loss": 0.7037,
+      "step": 3002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472284814728265e-05,
+      "loss": 0.7322,
+      "step": 3003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471867145797346e-05,
+      "loss": 26.8061,
+      "step": 3004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471449320862096e-05,
+      "loss": 0.7329,
+      "step": 3005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471031339937093e-05,
+      "loss": 25.2199,
+      "step": 3006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470613203036915e-05,
+      "loss": 0.7214,
+      "step": 3007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470194910176154e-05,
+      "loss": 24.8739,
+      "step": 3008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469776461369398e-05,
+      "loss": 0.7155,
+      "step": 3009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469357856631246e-05,
+      "loss": 22.6753,
+      "step": 3010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468939095976303e-05,
+      "loss": 23.8156,
+      "step": 3011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468520179419178e-05,
+      "loss": 24.9634,
+      "step": 3012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468101106974484e-05,
+      "loss": 0.6658,
+      "step": 3013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467681878656837e-05,
+      "loss": 23.2893,
+      "step": 3014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467262494480869e-05,
+      "loss": 23.4049,
+      "step": 3015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466842954461206e-05,
+      "loss": 23.1229,
+      "step": 3016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466423258612484e-05,
+      "loss": 0.7625,
+      "step": 3017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466003406949347e-05,
+      "loss": 24.8128,
+      "step": 3018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46558339948644e-05,
+      "loss": 23.0606,
+      "step": 3019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465163236238416e-05,
+      "loss": 24.6506,
+      "step": 3020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464742917219932e-05,
+      "loss": 0.7358,
+      "step": 3021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464322442445651e-05,
+      "loss": 23.1782,
+      "step": 3022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463901811930243e-05,
+      "loss": 23.4655,
+      "step": 3023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463481025688379e-05,
+      "loss": 22.6972,
+      "step": 3024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463060083734743e-05,
+      "loss": 22.6069,
+      "step": 3025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462638986084016e-05,
+      "loss": 0.7157,
+      "step": 3026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46221773275089e-05,
+      "loss": 25.6395,
+      "step": 3027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461796323750059e-05,
+      "loss": 22.0348,
+      "step": 3028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461374759096226e-05,
+      "loss": 23.4377,
+      "step": 3029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460953038804098e-05,
+      "loss": 0.7278,
+      "step": 3030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460531162888386e-05,
+      "loss": 22.5408,
+      "step": 3031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460109131363809e-05,
+      "loss": 0.7946,
+      "step": 3032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459686944245086e-05,
+      "loss": 27.1165,
+      "step": 3033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459264601546949e-05,
+      "loss": 25.6277,
+      "step": 3034
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458842103284129e-05,
+      "loss": 23.6728,
+      "step": 3035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458419449471367e-05,
+      "loss": 23.516,
+      "step": 3036
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457996640123408e-05,
+      "loss": 24.34,
+      "step": 3037
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457573675255001e-05,
+      "loss": 0.7186,
+      "step": 3038
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457150554880902e-05,
+      "loss": 23.7514,
+      "step": 3039
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456727279015871e-05,
+      "loss": 0.7807,
+      "step": 3040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456303847674674e-05,
+      "loss": 22.7471,
+      "step": 3041
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455880260872086e-05,
+      "loss": 25.7109,
+      "step": 3042
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45545651862288e-05,
+      "loss": 22.7357,
+      "step": 3043
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45503262094184e-05,
+      "loss": 26.227,
+      "step": 3044
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454608567843754e-05,
+      "loss": 23.4269,
+      "step": 3045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454184359343416e-05,
+      "loss": 0.7713,
+      "step": 3046
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453759995455624e-05,
+      "loss": 21.0257,
+      "step": 3047
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453335476195184e-05,
+      "loss": 0.7479,
+      "step": 3048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452910801576903e-05,
+      "loss": 27.7282,
+      "step": 3049
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452485971615598e-05,
+      "loss": 22.9358,
+      "step": 3050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452060986326088e-05,
+      "loss": 23.9402,
+      "step": 3051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451635845723201e-05,
+      "loss": 24.8921,
+      "step": 3052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451210549821767e-05,
+      "loss": 22.3269,
+      "step": 3053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450785098636624e-05,
+      "loss": 26.3432,
+      "step": 3054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45035949218261e-05,
+      "loss": 22.9033,
+      "step": 3055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449933730474577e-05,
+      "loss": 0.7132,
+      "step": 3056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449507813527376e-05,
+      "loss": 23.3047,
+      "step": 3057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449081741355867e-05,
+      "loss": 21.8739,
+      "step": 3058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448655513974913e-05,
+      "loss": 0.6979,
+      "step": 3059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448229131399384e-05,
+      "loss": 0.7662,
+      "step": 3060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447802593644152e-05,
+      "loss": 0.7507,
+      "step": 3061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4473759007241e-05,
+      "loss": 0.7244,
+      "step": 3062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446949052654113e-05,
+      "loss": 25.0521,
+      "step": 3063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44652204944908e-05,
+      "loss": 24.3273,
+      "step": 3064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4460948911239e-05,
+      "loss": 0.7602,
+      "step": 3065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445667577693473e-05,
+      "loss": 24.8376,
+      "step": 3066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445240109172706e-05,
+      "loss": 24.8055,
+      "step": 3067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444812485576512e-05,
+      "loss": 23.8015,
+      "step": 3068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444384706919809e-05,
+      "loss": 24.6648,
+      "step": 3069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443956773217522e-05,
+      "loss": 0.7368,
+      "step": 3070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443528684484575e-05,
+      "loss": 24.2659,
+      "step": 3071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443100440735908e-05,
+      "loss": 22.7409,
+      "step": 3072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442672041986457e-05,
+      "loss": 0.7481,
+      "step": 3073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442243488251167e-05,
+      "loss": 23.7721,
+      "step": 3074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441814779544991e-05,
+      "loss": 24.6204,
+      "step": 3075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441385915882883e-05,
+      "loss": 0.7869,
+      "step": 3076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440956897279803e-05,
+      "loss": 0.7361,
+      "step": 3077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44052772375072e-05,
+      "loss": 21.9365,
+      "step": 3078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440098395310606e-05,
+      "loss": 24.0697,
+      "step": 3079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439668911974435e-05,
+      "loss": 23.642,
+      "step": 3080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439239273757194e-05,
+      "loss": 22.6905,
+      "step": 3081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438809480673869e-05,
+      "loss": 23.4681,
+      "step": 3082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438379532739455e-05,
+      "loss": 0.7234,
+      "step": 3083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43794942996895e-05,
+      "loss": 0.7153,
+      "step": 3084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437519172377358e-05,
+      "loss": 0.7011,
+      "step": 3085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43708875997969e-05,
+      "loss": 22.2116,
+      "step": 3086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436658192790961e-05,
+      "loss": 23.4718,
+      "step": 3087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43622747082619e-05,
+      "loss": 0.7577,
+      "step": 3088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435796594100407e-05,
+      "loss": 0.8082,
+      "step": 3089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43536556262864e-05,
+      "loss": 23.0573,
+      "step": 3090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434934376425926e-05,
+      "loss": 0.7439,
+      "step": 3091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434503035507309e-05,
+      "loss": 24.8891,
+      "step": 3092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434071539887834e-05,
+      "loss": 0.7205,
+      "step": 3093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433639889582558e-05,
+      "loss": 0.7884,
+      "step": 3094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433208084606535e-05,
+      "loss": 24.129,
+      "step": 3095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432776124974832e-05,
+      "loss": 0.7811,
+      "step": 3096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432344010702516e-05,
+      "loss": 24.1185,
+      "step": 3097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431911741804664e-05,
+      "loss": 22.5936,
+      "step": 3098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431479318296354e-05,
+      "loss": 0.7337,
+      "step": 3099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431046740192674e-05,
+      "loss": 20.3174,
+      "step": 3100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430614007508712e-05,
+      "loss": 0.712,
+      "step": 3101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430181120259565e-05,
+      "loss": 25.9932,
+      "step": 3102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429748078460335e-05,
+      "loss": 0.7193,
+      "step": 3103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429314882126129e-05,
+      "loss": 0.7375,
+      "step": 3104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428881531272059e-05,
+      "loss": 0.7322,
+      "step": 3105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428448025913244e-05,
+      "loss": 0.738,
+      "step": 3106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428014366064803e-05,
+      "loss": 0.7518,
+      "step": 3107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42758055174187e-05,
+      "loss": 0.7248,
+      "step": 3108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427146582959576e-05,
+      "loss": 25.0227,
+      "step": 3109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426712459733062e-05,
+      "loss": 24.2775,
+      "step": 3110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42627818207747e-05,
+      "loss": 0.7301,
+      "step": 3111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425843750007952e-05,
+      "loss": 23.7229,
+      "step": 3112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425409163539663e-05,
+      "loss": 0.6472,
+      "step": 3113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424974422687765e-05,
+      "loss": 24.1224,
+      "step": 3114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424539527467421e-05,
+      "loss": 23.5783,
+      "step": 3115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424104477893807e-05,
+      "loss": 20.7245,
+      "step": 3116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423669273982096e-05,
+      "loss": 0.7843,
+      "step": 3117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423233915747473e-05,
+      "loss": 23.0215,
+      "step": 3118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422798403205127e-05,
+      "loss": 25.0931,
+      "step": 3119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422362736370247e-05,
+      "loss": 0.7295,
+      "step": 3120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421926915258035e-05,
+      "loss": 21.3664,
+      "step": 3121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421490939883693e-05,
+      "loss": 0.8136,
+      "step": 3122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421054810262432e-05,
+      "loss": 0.733,
+      "step": 3123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420618526409466e-05,
+      "loss": 0.729,
+      "step": 3124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420182088340013e-05,
+      "loss": 0.7569,
+      "step": 3125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419745496069301e-05,
+      "loss": 25.7001,
+      "step": 3126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41930874961256e-05,
+      "loss": 23.4657,
+      "step": 3127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418871848985027e-05,
+      "loss": 25.0789,
+      "step": 3128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418434794201944e-05,
+      "loss": 0.7452,
+      "step": 3129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417997585278555e-05,
+      "loss": 22.0675,
+      "step": 3130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417560222230115e-05,
+      "loss": 21.8764,
+      "step": 3131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417122705071882e-05,
+      "loss": 22.4027,
+      "step": 3132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416685033819116e-05,
+      "loss": 22.9254,
+      "step": 3133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416247208487088e-05,
+      "loss": 0.754,
+      "step": 3134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415809229091073e-05,
+      "loss": 0.6999,
+      "step": 3135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415371095646348e-05,
+      "loss": 21.9559,
+      "step": 3136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414932808168198e-05,
+      "loss": 22.7087,
+      "step": 3137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414494366671913e-05,
+      "loss": 21.9945,
+      "step": 3138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414055771172788e-05,
+      "loss": 26.4262,
+      "step": 3139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413617021686126e-05,
+      "loss": 24.0235,
+      "step": 3140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413178118227231e-05,
+      "loss": 22.6625,
+      "step": 3141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412739060811413e-05,
+      "loss": 0.7291,
+      "step": 3142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412299849453993e-05,
+      "loss": 0.7256,
+      "step": 3143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41186048417029e-05,
+      "loss": 22.7801,
+      "step": 3144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411420964975632e-05,
+      "loss": 0.7257,
+      "step": 3145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410981291885352e-05,
+      "loss": 0.7055,
+      "step": 3146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410541464914791e-05,
+      "loss": 26.3406,
+      "step": 3147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410101484079289e-05,
+      "loss": 0.7032,
+      "step": 3148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409661349394197e-05,
+      "loss": 20.4124,
+      "step": 3149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409221060874868e-05,
+      "loss": 0.7467,
+      "step": 3150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408780618536663e-05,
+      "loss": 20.206,
+      "step": 3151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408340022394948e-05,
+      "loss": 21.1306,
+      "step": 3152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407899272465091e-05,
+      "loss": 0.7092,
+      "step": 3153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407458368762469e-05,
+      "loss": 21.8663,
+      "step": 3154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407017311302465e-05,
+      "loss": 22.5544,
+      "step": 3155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406576100100463e-05,
+      "loss": 0.7757,
+      "step": 3156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406134735171856e-05,
+      "loss": 0.683,
+      "step": 3157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405693216532043e-05,
+      "loss": 23.8219,
+      "step": 3158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405251544196423e-05,
+      "loss": 24.9424,
+      "step": 3159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404809718180407e-05,
+      "loss": 0.7019,
+      "step": 3160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404367738499409e-05,
+      "loss": 24.0206,
+      "step": 3161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403925605168844e-05,
+      "loss": 22.7374,
+      "step": 3162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40348331820414e-05,
+      "loss": 23.4802,
+      "step": 3163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403040877620725e-05,
+      "loss": 21.9972,
+      "step": 3164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402598283434032e-05,
+      "loss": 22.9386,
+      "step": 3165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402155535659505e-05,
+      "loss": 0.7268,
+      "step": 3166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401712634312587e-05,
+      "loss": 20.916,
+      "step": 3167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40126957940873e-05,
+      "loss": 22.2575,
+      "step": 3168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40082637096339e-05,
+      "loss": 23.4374,
+      "step": 3169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400383008992029e-05,
+      "loss": 0.6661,
+      "step": 3170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399939493510112e-05,
+      "loss": 24.5944,
+      "step": 3171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399495824533113e-05,
+      "loss": 23.7663,
+      "step": 3172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399052002076511e-05,
+      "loss": 0.7541,
+      "step": 3173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398608026155786e-05,
+      "loss": 0.7391,
+      "step": 3174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398163896786429e-05,
+      "loss": 25.4176,
+      "step": 3175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397719613983933e-05,
+      "loss": 0.7565,
+      "step": 3176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397275177763799e-05,
+      "loss": 22.633,
+      "step": 3177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396830588141525e-05,
+      "loss": 24.1554,
+      "step": 3178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39638584513263e-05,
+      "loss": 0.7321,
+      "step": 3179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39594094875262e-05,
+      "loss": 23.2548,
+      "step": 3180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395495899017024e-05,
+      "loss": 24.714,
+      "step": 3181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395050695941361e-05,
+      "loss": 22.2013,
+      "step": 3182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394605339541167e-05,
+      "loss": 23.9854,
+      "step": 3183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394159829831977e-05,
+      "loss": 0.7348,
+      "step": 3184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39371416682933e-05,
+      "loss": 21.9132,
+      "step": 3185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393268350548779e-05,
+      "loss": 22.4938,
+      "step": 3186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392822381005873e-05,
+      "loss": 0.7339,
+      "step": 3187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392376258216169e-05,
+      "loss": 21.1809,
+      "step": 3188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391929982195232e-05,
+      "loss": 0.6779,
+      "step": 3189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391483552958631e-05,
+      "loss": 0.7247,
+      "step": 3190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39103697052194e-05,
+      "loss": 0.819,
+      "step": 3191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390590234900738e-05,
+      "loss": 0.7425,
+      "step": 3192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390143346110609e-05,
+      "loss": 0.7287,
+      "step": 3193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389696304167144e-05,
+      "loss": 0.8199,
+      "step": 3194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389249109085939e-05,
+      "loss": 25.0664,
+      "step": 3195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388801760882592e-05,
+      "loss": 23.633,
+      "step": 3196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388354259572712e-05,
+      "loss": 22.3215,
+      "step": 3197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387906605171908e-05,
+      "loss": 0.7364,
+      "step": 3198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387458797695798e-05,
+      "loss": 21.3577,
+      "step": 3199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387010837160005e-05,
+      "loss": 22.9597,
+      "step": 3200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386562723580154e-05,
+      "loss": 24.1131,
+      "step": 3201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38611445697188e-05,
+      "loss": 25.2633,
+      "step": 3202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385666037350821e-05,
+      "loss": 23.8666,
+      "step": 3203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385217464732619e-05,
+      "loss": 25.2995,
+      "step": 3204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384768739132925e-05,
+      "loss": 0.7837,
+      "step": 3205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38431986056739e-05,
+      "loss": 24.6664,
+      "step": 3206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383870829051675e-05,
+      "loss": 22.3606,
+      "step": 3207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383421644601447e-05,
+      "loss": 21.8652,
+      "step": 3208
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382972307232372e-05,
+      "loss": 22.341,
+      "step": 3209
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382522816960129e-05,
+      "loss": 23.8809,
+      "step": 3210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382073173800396e-05,
+      "loss": 21.3307,
+      "step": 3211
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381623377768861e-05,
+      "loss": 0.7543,
+      "step": 3212
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381173428881214e-05,
+      "loss": 23.5816,
+      "step": 3213
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380723327153153e-05,
+      "loss": 0.7375,
+      "step": 3214
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380273072600382e-05,
+      "loss": 0.714,
+      "step": 3215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379822665238604e-05,
+      "loss": 22.8921,
+      "step": 3216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379372105083535e-05,
+      "loss": 24.1144,
+      "step": 3217
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378921392150892e-05,
+      "loss": 0.7197,
+      "step": 3218
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378470526456398e-05,
+      "loss": 23.6015,
+      "step": 3219
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378019508015782e-05,
+      "loss": 22.9463,
+      "step": 3220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377568336844781e-05,
+      "loss": 0.7594,
+      "step": 3221
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37711701295913e-05,
+      "loss": 0.6731,
+      "step": 3222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376665536374576e-05,
+      "loss": 22.1723,
+      "step": 3223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37621390710687e-05,
+      "loss": 23.9265,
+      "step": 3224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375762125171765e-05,
+      "loss": 24.268,
+      "step": 3225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375310190585022e-05,
+      "loss": 23.521,
+      "step": 3226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37485810336241e-05,
+      "loss": 22.121,
+      "step": 3227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374405863519696e-05,
+      "loss": 0.7395,
+      "step": 3228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37395347107266e-05,
+      "loss": 0.7442,
+      "step": 3229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373500926037084e-05,
+      "loss": 0.7091,
+      "step": 3230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373048228428755e-05,
+      "loss": 21.3277,
+      "step": 3231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372595378263465e-05,
+      "loss": 19.8319,
+      "step": 3232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372142375557011e-05,
+      "loss": 21.225,
+      "step": 3233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371689220325197e-05,
+      "loss": 23.4301,
+      "step": 3234
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371235912583833e-05,
+      "loss": 0.7334,
+      "step": 3235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370782452348734e-05,
+      "loss": 22.978,
+      "step": 3236
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370328839635716e-05,
+      "loss": 21.4842,
+      "step": 3237
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369875074460605e-05,
+      "loss": 23.5818,
+      "step": 3238
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369421156839231e-05,
+      "loss": 23.0035,
+      "step": 3239
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368967086787428e-05,
+      "loss": 0.7245,
+      "step": 3240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368512864321039e-05,
+      "loss": 23.0186,
+      "step": 3241
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36805848945591e-05,
+      "loss": 25.409,
+      "step": 3242
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36760396220789e-05,
+      "loss": 22.7818,
+      "step": 3243
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367149282592835e-05,
+      "loss": 23.6721,
+      "step": 3244
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366694450626608e-05,
+      "loss": 20.8318,
+      "step": 3245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366239466325077e-05,
+      "loss": 24.1887,
+      "step": 3246
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365784329704115e-05,
+      "loss": 22.7442,
+      "step": 3247
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365329040779598e-05,
+      "loss": 0.7709,
+      "step": 3248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36487359956741e-05,
+      "loss": 21.9651,
+      "step": 3249
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364418006083438e-05,
+      "loss": 0.7391,
+      "step": 3250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363962260343577e-05,
+      "loss": 0.7913,
+      "step": 3251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363506362363728e-05,
+      "loss": 0.7267,
+      "step": 3252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36305031215979e-05,
+      "loss": 0.6996,
+      "step": 3253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362594109747677e-05,
+      "loss": 23.945,
+      "step": 3254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362137755143303e-05,
+      "loss": 25.5258,
+      "step": 3255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361681248362588e-05,
+      "loss": 0.6995,
+      "step": 3256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361224589421457e-05,
+      "loss": 22.8238,
+      "step": 3257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360767778335841e-05,
+      "loss": 24.9455,
+      "step": 3258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360310815121677e-05,
+      "loss": 0.7893,
+      "step": 3259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359853699794907e-05,
+      "loss": 23.225,
+      "step": 3260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359396432371476e-05,
+      "loss": 0.6586,
+      "step": 3261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358939012867339e-05,
+      "loss": 0.7561,
+      "step": 3262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358481441298448e-05,
+      "loss": 24.4619,
+      "step": 3263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358023717680772e-05,
+      "loss": 21.7707,
+      "step": 3264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357565842030273e-05,
+      "loss": 0.6967,
+      "step": 3265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35710781436293e-05,
+      "loss": 20.7799,
+      "step": 3266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356649634694719e-05,
+      "loss": 0.7296,
+      "step": 3267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356191303041624e-05,
+      "loss": 0.7429,
+      "step": 3268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355732819419633e-05,
+      "loss": 24.714,
+      "step": 3269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355274183844744e-05,
+      "loss": 21.513,
+      "step": 3270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354815396332953e-05,
+      "loss": 21.5944,
+      "step": 3271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354356456900268e-05,
+      "loss": 23.499,
+      "step": 3272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353897365562699e-05,
+      "loss": 0.6706,
+      "step": 3273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353438122336257e-05,
+      "loss": 0.7374,
+      "step": 3274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35297872723697e-05,
+      "loss": 24.7551,
+      "step": 3275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35251918028086e-05,
+      "loss": 0.6908,
+      "step": 3276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352059481483961e-05,
+      "loss": 0.7516,
+      "step": 3277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351599630862309e-05,
+      "loss": 0.7195,
+      "step": 3278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351139628431947e-05,
+      "loss": 22.9554,
+      "step": 3279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350679474208919e-05,
+      "loss": 21.3087,
+      "step": 3280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35021916820928e-05,
+      "loss": 0.7614,
+      "step": 3281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349758710449091e-05,
+      "loss": 0.7948,
+      "step": 3282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349298100944412e-05,
+      "loss": 24.421,
+      "step": 3283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348837339711314e-05,
+      "loss": 23.5311,
+      "step": 3284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348376426765868e-05,
+      "loss": 0.7463,
+      "step": 3285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347915362124156e-05,
+      "loss": 23.0255,
+      "step": 3286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347454145802259e-05,
+      "loss": 24.626,
+      "step": 3287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346992777816272e-05,
+      "loss": 22.1248,
+      "step": 3288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346531258182287e-05,
+      "loss": 0.6654,
+      "step": 3289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346069586916404e-05,
+      "loss": 0.7328,
+      "step": 3290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34560776403473e-05,
+      "loss": 0.7806,
+      "step": 3291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345145789553376e-05,
+      "loss": 22.7199,
+      "step": 3292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344683663488458e-05,
+      "loss": 22.5735,
+      "step": 3293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344221385856098e-05,
+      "loss": 21.9754,
+      "step": 3294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343758956672421e-05,
+      "loss": 0.8083,
+      "step": 3295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343296375953563e-05,
+      "loss": 0.7425,
+      "step": 3296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342833643715658e-05,
+      "loss": 23.3284,
+      "step": 3297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34237075997485e-05,
+      "loss": 0.716,
+      "step": 3298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341907724747286e-05,
+      "loss": 24.1049,
+      "step": 3299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341444538049122e-05,
+      "loss": 21.5333,
+      "step": 3300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340981199896515e-05,
+      "loss": 0.7885,
+      "step": 3301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340517710305628e-05,
+      "loss": 23.4765,
+      "step": 3302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34005406929263e-05,
+      "loss": 0.7795,
+      "step": 3303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339590276873696e-05,
+      "loss": 0.7291,
+      "step": 3304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339126333065007e-05,
+      "loss": 0.7806,
+      "step": 3305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338662237882749e-05,
+      "loss": 0.7209,
+      "step": 3306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338197991343107e-05,
+      "loss": 0.6833,
+      "step": 3307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337733593462282e-05,
+      "loss": 0.7082,
+      "step": 3308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337269044256473e-05,
+      "loss": 22.6386,
+      "step": 3309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336804343741887e-05,
+      "loss": 0.7303,
+      "step": 3310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336339491934731e-05,
+      "loss": 22.1399,
+      "step": 3311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335874488851228e-05,
+      "loss": 25.0569,
+      "step": 3312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335409334507596e-05,
+      "loss": 24.3045,
+      "step": 3313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334944028920064e-05,
+      "loss": 24.228,
+      "step": 3314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334478572104863e-05,
+      "loss": 22.3399,
+      "step": 3315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334012964078233e-05,
+      "loss": 0.7126,
+      "step": 3316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333547204856414e-05,
+      "loss": 0.7647,
+      "step": 3317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333081294455658e-05,
+      "loss": 0.7384,
+      "step": 3318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332615232892215e-05,
+      "loss": 0.702,
+      "step": 3319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332149020182347e-05,
+      "loss": 25.384,
+      "step": 3320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331682656342316e-05,
+      "loss": 0.7225,
+      "step": 3321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331216141388393e-05,
+      "loss": 24.5909,
+      "step": 3322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33074947533685e-05,
+      "loss": 0.7807,
+      "step": 3323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330282658203971e-05,
+      "loss": 22.2684,
+      "step": 3324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329815690006037e-05,
+      "loss": 0.8025,
+      "step": 3325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329348570759342e-05,
+      "loss": 0.8395,
+      "step": 3326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32888130048018e-05,
+      "loss": 23.9316,
+      "step": 3327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328413879184853e-05,
+      "loss": 23.3244,
+      "step": 3328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327946306889666e-05,
+      "loss": 0.6762,
+      "step": 3329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327478583610934e-05,
+      "loss": 22.6074,
+      "step": 3330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327010709364968e-05,
+      "loss": 0.7193,
+      "step": 3331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326542684168096e-05,
+      "loss": 22.3326,
+      "step": 3332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326074508036642e-05,
+      "loss": 0.7311,
+      "step": 3333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325606180986939e-05,
+      "loss": 22.3874,
+      "step": 3334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325137703035325e-05,
+      "loss": 24.0223,
+      "step": 3335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324669074198144e-05,
+      "loss": 23.8434,
+      "step": 3336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324200294491745e-05,
+      "loss": 24.3605,
+      "step": 3337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323731363932481e-05,
+      "loss": 0.8118,
+      "step": 3338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323262282536709e-05,
+      "loss": 25.5388,
+      "step": 3339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322793050320797e-05,
+      "loss": 0.6649,
+      "step": 3340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322323667301112e-05,
+      "loss": 0.7614,
+      "step": 3341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32185413349403e-05,
+      "loss": 22.9315,
+      "step": 3342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321384448915928e-05,
+      "loss": 24.519,
+      "step": 3343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320914613583196e-05,
+      "loss": 22.0224,
+      "step": 3344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320444627512221e-05,
+      "loss": 22.2845,
+      "step": 3345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319974490719399e-05,
+      "loss": 0.6742,
+      "step": 3346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319504203221135e-05,
+      "loss": 23.0781,
+      "step": 3347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31903376503383e-05,
+      "loss": 23.0328,
+      "step": 3348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318563176173899e-05,
+      "loss": 22.5307,
+      "step": 3349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318092436657757e-05,
+      "loss": 0.7355,
+      "step": 3350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317621546501827e-05,
+      "loss": 0.6983,
+      "step": 3351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317150505722538e-05,
+      "loss": 0.7504,
+      "step": 3352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316679314336318e-05,
+      "loss": 0.7636,
+      "step": 3353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316207972359609e-05,
+      "loss": 24.3775,
+      "step": 3354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315736479808854e-05,
+      "loss": 22.4199,
+      "step": 3355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315264836700499e-05,
+      "loss": 0.7041,
+      "step": 3356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314793043050998e-05,
+      "loss": 24.3016,
+      "step": 3357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314321098876812e-05,
+      "loss": 23.6291,
+      "step": 3358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313849004194402e-05,
+      "loss": 21.2277,
+      "step": 3359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313376759020241e-05,
+      "loss": 0.7219,
+      "step": 3360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3129043633708e-05,
+      "loss": 24.3389,
+      "step": 3361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312431817262563e-05,
+      "loss": 23.791,
+      "step": 3362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31195912071201e-05,
+      "loss": 0.6559,
+      "step": 3363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311486273735636e-05,
+      "loss": 23.4743,
+      "step": 3364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311013276349934e-05,
+      "loss": 25.2672,
+      "step": 3365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310540128571406e-05,
+      "loss": 0.7345,
+      "step": 3366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310066830416556e-05,
+      "loss": 21.9825,
+      "step": 3367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309593381901898e-05,
+      "loss": 21.7945,
+      "step": 3368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309119783043948e-05,
+      "loss": 0.7224,
+      "step": 3369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308646033859227e-05,
+      "loss": 23.6255,
+      "step": 3370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308172134364261e-05,
+      "loss": 0.6775,
+      "step": 3371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307698084575587e-05,
+      "loss": 25.1717,
+      "step": 3372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307223884509737e-05,
+      "loss": 23.3865,
+      "step": 3373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306749534183257e-05,
+      "loss": 0.7436,
+      "step": 3374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306275033612694e-05,
+      "loss": 24.5516,
+      "step": 3375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305800382814603e-05,
+      "loss": 0.801,
+      "step": 3376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305325581805538e-05,
+      "loss": 0.7193,
+      "step": 3377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304850630602069e-05,
+      "loss": 0.7665,
+      "step": 3378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30437552922076e-05,
+      "loss": 0.7333,
+      "step": 3379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303900277678188e-05,
+      "loss": 21.4026,
+      "step": 3380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303424875990931e-05,
+      "loss": 21.6031,
+      "step": 3381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302949324175575e-05,
+      "loss": 0.6644,
+      "step": 3382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302473622248709e-05,
+      "loss": 23.4867,
+      "step": 3383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.30199777022693e-05,
+      "loss": 21.3865,
+      "step": 3384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301521768126835e-05,
+      "loss": 0.7106,
+      "step": 3385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301045615965031e-05,
+      "loss": 22.242,
+      "step": 3386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300569313758132e-05,
+      "loss": 24.794,
+      "step": 3387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.30009286152275e-05,
+      "loss": 0.7352,
+      "step": 3388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29961625927551e-05,
+      "loss": 23.3322,
+      "step": 3389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.299139507033035e-05,
+      "loss": 0.7783,
+      "step": 3390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298662604811959e-05,
+      "loss": 22.4129,
+      "step": 3391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298185552628917e-05,
+      "loss": 21.5821,
+      "step": 3392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297708350500556e-05,
+      "loss": 22.346,
+      "step": 3393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297230998443519e-05,
+      "loss": 0.7069,
+      "step": 3394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29675349647446e-05,
+      "loss": 22.8609,
+      "step": 3395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296275844610038e-05,
+      "loss": 0.7061,
+      "step": 3396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295798042866917e-05,
+      "loss": 24.9488,
+      "step": 3397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295320091261761e-05,
+      "loss": 23.3452,
+      "step": 3398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294841989811249e-05,
+      "loss": 23.0876,
+      "step": 3399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294363738532056e-05,
+      "loss": 24.8092,
+      "step": 3400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293885337440869e-05,
+      "loss": 0.7591,
+      "step": 3401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293406786554375e-05,
+      "loss": 22.8888,
+      "step": 3402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292928085889269e-05,
+      "loss": 0.7569,
+      "step": 3403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292449235462253e-05,
+      "loss": 0.7462,
+      "step": 3404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291970235290031e-05,
+      "loss": 0.7524,
+      "step": 3405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291491085389311e-05,
+      "loss": 0.7793,
+      "step": 3406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29101178577681e-05,
+      "loss": 0.7669,
+      "step": 3407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29053233646925e-05,
+      "loss": 0.7109,
+      "step": 3408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290052737483359e-05,
+      "loss": 0.7763,
+      "step": 3409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289572988835861e-05,
+      "loss": 0.7682,
+      "step": 3410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289093090543499e-05,
+      "loss": 24.2155,
+      "step": 3411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288613042623011e-05,
+      "loss": 24.8813,
+      "step": 3412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288132845091144e-05,
+      "loss": 0.6985,
+      "step": 3413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287652497964653e-05,
+      "loss": 23.8561,
+      "step": 3414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287172001260291e-05,
+      "loss": 0.7466,
+      "step": 3415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286691354994826e-05,
+      "loss": 0.7048,
+      "step": 3416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286210559185018e-05,
+      "loss": 0.7461,
+      "step": 3417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285729613847646e-05,
+      "loss": 0.6636,
+      "step": 3418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285248518999488e-05,
+      "loss": 23.0496,
+      "step": 3419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284767274657325e-05,
+      "loss": 21.9241,
+      "step": 3420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284285880837946e-05,
+      "loss": 22.7018,
+      "step": 3421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283804337558147e-05,
+      "loss": 0.7331,
+      "step": 3422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283322644834723e-05,
+      "loss": 24.2973,
+      "step": 3423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28284080268448e-05,
+      "loss": 22.6819,
+      "step": 3424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282358811124227e-05,
+      "loss": 24.2355,
+      "step": 3425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281876670170781e-05,
+      "loss": 0.6952,
+      "step": 3426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28139437984096e-05,
+      "loss": 21.4962,
+      "step": 3427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28091194015159e-05,
+      "loss": 0.756,
+      "step": 3428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280429351119499e-05,
+      "loss": 0.7395,
+      "step": 3429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279946612761526e-05,
+      "loss": 22.1884,
+      "step": 3430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279463725094507e-05,
+      "loss": 25.2644,
+      "step": 3431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278980688135292e-05,
+      "loss": 0.6749,
+      "step": 3432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27849750190073e-05,
+      "loss": 24.196,
+      "step": 3433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278014166407678e-05,
+      "loss": 22.6571,
+      "step": 3434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277530681672997e-05,
+      "loss": 24.294,
+      "step": 3435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277047047713556e-05,
+      "loss": 23.4738,
+      "step": 3436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276563264546223e-05,
+      "loss": 0.6995,
+      "step": 3437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27607933218788e-05,
+      "loss": 25.2702,
+      "step": 3438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275595250655402e-05,
+      "loss": 22.5336,
+      "step": 3439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275111019965685e-05,
+      "loss": 22.0191,
+      "step": 3440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274626640135616e-05,
+      "loss": 0.7753,
+      "step": 3441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274142111182095e-05,
+      "loss": 0.7346,
+      "step": 3442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273657433122025e-05,
+      "loss": 23.0668,
+      "step": 3443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273172605972313e-05,
+      "loss": 24.0317,
+      "step": 3444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272687629749875e-05,
+      "loss": 21.9374,
+      "step": 3445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27220250447163e-05,
+      "loss": 24.1382,
+      "step": 3446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271717230154498e-05,
+      "loss": 24.0356,
+      "step": 3447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271231806815412e-05,
+      "loss": 0.6796,
+      "step": 3448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270746234471303e-05,
+      "loss": 23.3915,
+      "step": 3449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270260513139116e-05,
+      "loss": 22.8073,
+      "step": 3450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26977464283579e-05,
+      "loss": 24.9949,
+      "step": 3451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269288623578276e-05,
+      "loss": 0.7828,
+      "step": 3452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268802455383531e-05,
+      "loss": 24.8134,
+      "step": 3453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268316138268516e-05,
+      "loss": 21.5701,
+      "step": 3454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267829672250194e-05,
+      "loss": 0.7357,
+      "step": 3455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267343057345539e-05,
+      "loss": 0.7489,
+      "step": 3456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266856293571521e-05,
+      "loss": 22.6471,
+      "step": 3457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266369380945127e-05,
+      "loss": 25.6472,
+      "step": 3458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265882319483343e-05,
+      "loss": 25.2611,
+      "step": 3459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265395109203157e-05,
+      "loss": 0.748,
+      "step": 3460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264907750121568e-05,
+      "loss": 22.6001,
+      "step": 3461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264420242255576e-05,
+      "loss": 20.7043,
+      "step": 3462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263932585622191e-05,
+      "loss": 22.1168,
+      "step": 3463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263444780238421e-05,
+      "loss": 0.7138,
+      "step": 3464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262956826121288e-05,
+      "loss": 0.7388,
+      "step": 3465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262468723287812e-05,
+      "loss": 22.4858,
+      "step": 3466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261980471755022e-05,
+      "loss": 24.758,
+      "step": 3467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26149207153995e-05,
+      "loss": 0.757,
+      "step": 3468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261003522659635e-05,
+      "loss": 22.8848,
+      "step": 3469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26051482513112e-05,
+      "loss": 0.723,
+      "step": 3470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260025978971454e-05,
+      "loss": 0.6899,
+      "step": 3471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25953698419769e-05,
+      "loss": 0.7139,
+      "step": 3472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259047840826887e-05,
+      "loss": 22.2983,
+      "step": 3473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25855854887611e-05,
+      "loss": 23.5902,
+      "step": 3474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258069108362427e-05,
+      "loss": 22.3782,
+      "step": 3475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257579519302914e-05,
+      "loss": 23.4004,
+      "step": 3476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257089781714649e-05,
+      "loss": 0.7057,
+      "step": 3477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256599895614719e-05,
+      "loss": 22.8654,
+      "step": 3478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256109861020213e-05,
+      "loss": 22.1849,
+      "step": 3479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255619677948223e-05,
+      "loss": 0.6656,
+      "step": 3480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255129346415856e-05,
+      "loss": 0.667,
+      "step": 3481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254638866440213e-05,
+      "loss": 0.7283,
+      "step": 3482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254148238038404e-05,
+      "loss": 0.7544,
+      "step": 3483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25365746122755e-05,
+      "loss": 0.7529,
+      "step": 3484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253166536024766e-05,
+      "loss": 0.7176,
+      "step": 3485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252675462447183e-05,
+      "loss": 24.1582,
+      "step": 3486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252184240511927e-05,
+      "loss": 23.1398,
+      "step": 3487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251692870236141e-05,
+      "loss": 22.1189,
+      "step": 3488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251201351636965e-05,
+      "loss": 25.7342,
+      "step": 3489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250709684731542e-05,
+      "loss": 24.9714,
+      "step": 3490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250217869537029e-05,
+      "loss": 22.2276,
+      "step": 3491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249725906070581e-05,
+      "loss": 0.7332,
+      "step": 3492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249233794349359e-05,
+      "loss": 0.8154,
+      "step": 3493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248741534390534e-05,
+      "loss": 0.6904,
+      "step": 3494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248249126211279e-05,
+      "loss": 23.4353,
+      "step": 3495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24775656982877e-05,
+      "loss": 0.728,
+      "step": 3496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247263865260188e-05,
+      "loss": 23.3321,
+      "step": 3497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246771012522726e-05,
+      "loss": 21.6292,
+      "step": 3498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246278011633576e-05,
+      "loss": 22.3662,
+      "step": 3499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.245784862609934e-05,
+      "loss": 21.7978,
+      "step": 3500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.245291565469008e-05,
+      "loss": 0.7619,
+      "step": 3501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244798120228005e-05,
+      "loss": 0.734,
+      "step": 3502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244304526904139e-05,
+      "loss": 22.2721,
+      "step": 3503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243810785514628e-05,
+      "loss": 22.7482,
+      "step": 3504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243316896076701e-05,
+      "loss": 0.747,
+      "step": 3505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242822858607584e-05,
+      "loss": 0.7555,
+      "step": 3506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242328673124512e-05,
+      "loss": 0.7374,
+      "step": 3507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241834339644726e-05,
+      "loss": 21.3897,
+      "step": 3508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241339858185472e-05,
+      "loss": 0.7785,
+      "step": 3509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240845228763997e-05,
+      "loss": 0.7173,
+      "step": 3510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24035045139756e-05,
+      "loss": 23.9539,
+      "step": 3511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23985552610342e-05,
+      "loss": 23.5387,
+      "step": 3512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239360452898844e-05,
+      "loss": 0.7221,
+      "step": 3513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238865231801103e-05,
+      "loss": 0.7602,
+      "step": 3514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238369862827472e-05,
+      "loss": 20.3659,
+      "step": 3515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237874345995231e-05,
+      "loss": 23.5439,
+      "step": 3516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237378681321669e-05,
+      "loss": 24.5436,
+      "step": 3517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236882868824077e-05,
+      "loss": 23.2869,
+      "step": 3518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236386908519752e-05,
+      "loss": 24.287,
+      "step": 3519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235890800425996e-05,
+      "loss": 0.7628,
+      "step": 3520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235394544560114e-05,
+      "loss": 0.7169,
+      "step": 3521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234898140939421e-05,
+      "loss": 21.0958,
+      "step": 3522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234401589581234e-05,
+      "loss": 21.9971,
+      "step": 3523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233904890502873e-05,
+      "loss": 19.7146,
+      "step": 3524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233408043721668e-05,
+      "loss": 0.7132,
+      "step": 3525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232911049254952e-05,
+      "loss": 22.3086,
+      "step": 3526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232413907120062e-05,
+      "loss": 23.947,
+      "step": 3527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231916617334342e-05,
+      "loss": 20.256,
+      "step": 3528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23141917991514e-05,
+      "loss": 0.6792,
+      "step": 3529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230921594879808e-05,
+      "loss": 0.7411,
+      "step": 3530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230423862245707e-05,
+      "loss": 0.7533,
+      "step": 3531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229925982030198e-05,
+      "loss": 23.3093,
+      "step": 3532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229427954250654e-05,
+      "loss": 0.7229,
+      "step": 3533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228929778924447e-05,
+      "loss": 0.8094,
+      "step": 3534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228431456068953e-05,
+      "loss": 23.3878,
+      "step": 3535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227932985701562e-05,
+      "loss": 0.7614,
+      "step": 3536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22743436783966e-05,
+      "loss": 0.7208,
+      "step": 3537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226935602500641e-05,
+      "loss": 21.0771,
+      "step": 3538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226436689701907e-05,
+      "loss": 0.8045,
+      "step": 3539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225937629460863e-05,
+      "loss": 20.718,
+      "step": 3540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225438421794918e-05,
+      "loss": 0.77,
+      "step": 3541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224939066721488e-05,
+      "loss": 24.1084,
+      "step": 3542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22443956425799e-05,
+      "loss": 22.6829,
+      "step": 3543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223939914421855e-05,
+      "loss": 24.1503,
+      "step": 3544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223440117230509e-05,
+      "loss": 21.4113,
+      "step": 3545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22294017270139e-05,
+      "loss": 20.2775,
+      "step": 3546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222440080851941e-05,
+      "loss": 21.2274,
+      "step": 3547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221939841699604e-05,
+      "loss": 22.5552,
+      "step": 3548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221439455261833e-05,
+      "loss": 0.7019,
+      "step": 3549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220938921556082e-05,
+      "loss": 0.799,
+      "step": 3550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220438240599813e-05,
+      "loss": 23.1024,
+      "step": 3551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219937412410495e-05,
+      "loss": 0.6941,
+      "step": 3552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219436437005597e-05,
+      "loss": 0.7026,
+      "step": 3553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218935314402598e-05,
+      "loss": 0.7485,
+      "step": 3554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.218434044618979e-05,
+      "loss": 0.7473,
+      "step": 3555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217932627672226e-05,
+      "loss": 0.7262,
+      "step": 3556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217431063579834e-05,
+      "loss": 25.3378,
+      "step": 3557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216929352359298e-05,
+      "loss": 24.9002,
+      "step": 3558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216427494028122e-05,
+      "loss": 0.7279,
+      "step": 3559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215925488603812e-05,
+      "loss": 0.7088,
+      "step": 3560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215423336103884e-05,
+      "loss": 23.1729,
+      "step": 3561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214921036545852e-05,
+      "loss": 21.419,
+      "step": 3562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214418589947243e-05,
+      "loss": 23.2397,
+      "step": 3563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213915996325581e-05,
+      "loss": 22.8159,
+      "step": 3564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213413255698403e-05,
+      "loss": 0.7353,
+      "step": 3565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212910368083245e-05,
+      "loss": 0.7894,
+      "step": 3566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212407333497653e-05,
+      "loss": 22.09,
+      "step": 3567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211904151959174e-05,
+      "loss": 23.2292,
+      "step": 3568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211400823485363e-05,
+      "loss": 0.6879,
+      "step": 3569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210897348093777e-05,
+      "loss": 22.6338,
+      "step": 3570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210393725801981e-05,
+      "loss": 23.2432,
+      "step": 3571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209889956627546e-05,
+      "loss": 24.1805,
+      "step": 3572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209386040588044e-05,
+      "loss": 23.5398,
+      "step": 3573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208881977701056e-05,
+      "loss": 21.393,
+      "step": 3574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208377767984165e-05,
+      "loss": 24.1437,
+      "step": 3575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207873411454961e-05,
+      "loss": 0.743,
+      "step": 3576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207368908131042e-05,
+      "loss": 21.8584,
+      "step": 3577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206864258030003e-05,
+      "loss": 0.7013,
+      "step": 3578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206359461169453e-05,
+      "loss": 0.7064,
+      "step": 3579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205854517567002e-05,
+      "loss": 21.9697,
+      "step": 3580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205349427240262e-05,
+      "loss": 22.6202,
+      "step": 3581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.204844190206858e-05,
+      "loss": 0.7003,
+      "step": 3582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20433880648441e-05,
+      "loss": 0.688,
+      "step": 3583
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203833276090554e-05,
+      "loss": 0.6998,
+      "step": 3584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203327599042923e-05,
+      "loss": 21.8469,
+      "step": 3585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202821775359159e-05,
+      "loss": 0.7417,
+      "step": 3586
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202315805056908e-05,
+      "loss": 22.9371,
+      "step": 3587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201809688153818e-05,
+      "loss": 22.9815,
+      "step": 3588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201303424667548e-05,
+      "loss": 23.277,
+      "step": 3589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200797014615762e-05,
+      "loss": 18.526,
+      "step": 3590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200290458016123e-05,
+      "loss": 0.702,
+      "step": 3591
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199783754886305e-05,
+      "loss": 22.0509,
+      "step": 3592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19927690524398e-05,
+      "loss": 0.7465,
+      "step": 3593
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198769909106834e-05,
+      "loss": 22.4301,
+      "step": 3594
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198262766492554e-05,
+      "loss": 21.5288,
+      "step": 3595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197755477418829e-05,
+      "loss": 0.7096,
+      "step": 3596
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19724804190336e-05,
+      "loss": 22.4428,
+      "step": 3597
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196740459963844e-05,
+      "loss": 21.4463,
+      "step": 3598
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196232731617995e-05,
+      "loss": 0.7704,
+      "step": 3599
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195724856883521e-05,
+      "loss": 21.3524,
+      "step": 3600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19521683577814e-05,
+      "loss": 21.705,
+      "step": 3601
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194708668319576e-05,
+      "loss": 21.5748,
+      "step": 3602
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194200354525557e-05,
+      "loss": 21.7661,
+      "step": 3603
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193691894413815e-05,
+      "loss": 0.7539,
+      "step": 3604
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193183288002086e-05,
+      "loss": 0.7146,
+      "step": 3605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192674535308116e-05,
+      "loss": 0.7548,
+      "step": 3606
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192165636349653e-05,
+      "loss": 0.7538,
+      "step": 3607
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19165659114445e-05,
+      "loss": 21.5873,
+      "step": 3608
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191147399710264e-05,
+      "loss": 24.9743,
+      "step": 3609
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19063806206486e-05,
+      "loss": 23.2392,
+      "step": 3610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190128578226007e-05,
+      "loss": 25.5067,
+      "step": 3611
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189618948211477e-05,
+      "loss": 21.4087,
+      "step": 3612
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18910917203905e-05,
+      "loss": 21.7062,
+      "step": 3613
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188599249726511e-05,
+      "loss": 20.4478,
+      "step": 3614
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188089181291646e-05,
+      "loss": 22.3359,
+      "step": 3615
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187578966752252e-05,
+      "loss": 20.093,
+      "step": 3616
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187068606126125e-05,
+      "loss": 0.6952,
+      "step": 3617
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186558099431073e-05,
+      "loss": 20.4446,
+      "step": 3618
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186047446684903e-05,
+      "loss": 0.7914,
+      "step": 3619
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18553664790543e-05,
+      "loss": 25.1433,
+      "step": 3620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185025703110474e-05,
+      "loss": 22.9255,
+      "step": 3621
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18451461231786e-05,
+      "loss": 0.6989,
+      "step": 3622
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184003375545418e-05,
+      "loss": 21.4938,
+      "step": 3623
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183491992810979e-05,
+      "loss": 21.5445,
+      "step": 3624
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182980464132388e-05,
+      "loss": 0.7212,
+      "step": 3625
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18246878952749e-05,
+      "loss": 0.7176,
+      "step": 3626
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18195696901413e-05,
+      "loss": 20.5881,
+      "step": 3627
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181445002610168e-05,
+      "loss": 23.4073,
+      "step": 3628
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180932890333462e-05,
+      "loss": 22.5955,
+      "step": 3629
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180420632201879e-05,
+      "loss": 0.6918,
+      "step": 3630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179908228233288e-05,
+      "loss": 0.7505,
+      "step": 3631
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179395678445565e-05,
+      "loss": 0.6644,
+      "step": 3632
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178882982856593e-05,
+      "loss": 22.6532,
+      "step": 3633
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178370141484253e-05,
+      "loss": 0.7513,
+      "step": 3634
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177857154346441e-05,
+      "loss": 21.8207,
+      "step": 3635
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177344021461048e-05,
+      "loss": 22.9059,
+      "step": 3636
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176830742845978e-05,
+      "loss": 0.6769,
+      "step": 3637
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176317318519139e-05,
+      "loss": 21.9958,
+      "step": 3638
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175803748498437e-05,
+      "loss": 22.7045,
+      "step": 3639
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175290032801793e-05,
+      "loss": 0.8252,
+      "step": 3640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174776171447124e-05,
+      "loss": 20.9558,
+      "step": 3641
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174262164452361e-05,
+      "loss": 20.6495,
+      "step": 3642
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173748011835433e-05,
+      "loss": 0.7509,
+      "step": 3643
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173233713614276e-05,
+      "loss": 0.7276,
+      "step": 3644
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172719269806833e-05,
+      "loss": 0.7734,
+      "step": 3645
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17220468043105e-05,
+      "loss": 22.0012,
+      "step": 3646
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17168994550488e-05,
+      "loss": 22.0754,
+      "step": 3647
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171175065046279e-05,
+      "loss": 21.4644,
+      "step": 3648
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170660039073206e-05,
+      "loss": 0.7867,
+      "step": 3649
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170144867603634e-05,
+      "loss": 22.8917,
+      "step": 3650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169629550655531e-05,
+      "loss": 22.3668,
+      "step": 3651
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169114088246878e-05,
+      "loss": 0.7134,
+      "step": 3652
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168598480395651e-05,
+      "loss": 0.821,
+      "step": 3653
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168082727119843e-05,
+      "loss": 20.0803,
+      "step": 3654
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167566828437443e-05,
+      "loss": 0.7104,
+      "step": 3655
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167050784366452e-05,
+      "loss": 0.6982,
+      "step": 3656
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16653459492487e-05,
+      "loss": 0.7662,
+      "step": 3657
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166018260130704e-05,
+      "loss": 0.7594,
+      "step": 3658
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16550178000197e-05,
+      "loss": 24.0058,
+      "step": 3659
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164985154556682e-05,
+      "loss": 20.6003,
+      "step": 3660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164468383812864e-05,
+      "loss": 0.6951,
+      "step": 3661
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163951467788546e-05,
+      "loss": 0.688,
+      "step": 3662
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163434406501759e-05,
+      "loss": 0.7279,
+      "step": 3663
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16291719997054e-05,
+      "loss": 0.7305,
+      "step": 3664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162399848212935e-05,
+      "loss": 0.7329,
+      "step": 3665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16188235124699e-05,
+      "loss": 0.7392,
+      "step": 3666
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161364709090761e-05,
+      "loss": 0.6509,
+      "step": 3667
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.160846921762303e-05,
+      "loss": 23.3396,
+      "step": 3668
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16032898927968e-05,
+      "loss": 0.7032,
+      "step": 3669
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159810911660961e-05,
+      "loss": 23.4841,
+      "step": 3670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159292688924221e-05,
+      "loss": 22.0044,
+      "step": 3671
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158774321087537e-05,
+      "loss": 20.3573,
+      "step": 3672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158255808168994e-05,
+      "loss": 22.0077,
+      "step": 3673
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157737150186678e-05,
+      "loss": 19.5817,
+      "step": 3674
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157218347158686e-05,
+      "loss": 0.7583,
+      "step": 3675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156699399103115e-05,
+      "loss": 0.7704,
+      "step": 3676
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156180306038067e-05,
+      "loss": 22.746,
+      "step": 3677
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155661067981656e-05,
+      "loss": 22.3226,
+      "step": 3678
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155141684951991e-05,
+      "loss": 23.0644,
+      "step": 3679
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154622156967193e-05,
+      "loss": 0.7402,
+      "step": 3680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154102484045387e-05,
+      "loss": 23.4863,
+      "step": 3681
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153582666204701e-05,
+      "loss": 24.5228,
+      "step": 3682
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15306270346327e-05,
+      "loss": 23.1403,
+      "step": 3683
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152542595839233e-05,
+      "loss": 23.5416,
+      "step": 3684
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152022343350734e-05,
+      "loss": 23.6652,
+      "step": 3685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151501946015921e-05,
+      "loss": 22.0784,
+      "step": 3686
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.150981403852952e-05,
+      "loss": 22.5968,
+      "step": 3687
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.150460716879985e-05,
+      "loss": 21.0615,
+      "step": 3688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149939885115181e-05,
+      "loss": 23.7662,
+      "step": 3689
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149418908576717e-05,
+      "loss": 21.7016,
+      "step": 3690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14889778728276e-05,
+      "loss": 23.7523,
+      "step": 3691
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148376521251494e-05,
+      "loss": 20.8264,
+      "step": 3692
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147855110501103e-05,
+      "loss": 24.6719,
+      "step": 3693
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147333555049776e-05,
+      "loss": 21.4396,
+      "step": 3694
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146811854915709e-05,
+      "loss": 0.7667,
+      "step": 3695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1462900101171e-05,
+      "loss": 0.6995,
+      "step": 3696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145768020672157e-05,
+      "loss": 20.3923,
+      "step": 3697
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145245886599088e-05,
+      "loss": 0.752,
+      "step": 3698
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144723607916107e-05,
+      "loss": 0.6922,
+      "step": 3699
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144201184641436e-05,
+      "loss": 22.2582,
+      "step": 3700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143678616793298e-05,
+      "loss": 19.7422,
+      "step": 3701
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143155904389927e-05,
+      "loss": 0.6989,
+      "step": 3702
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142633047449554e-05,
+      "loss": 0.6575,
+      "step": 3703
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142110045990422e-05,
+      "loss": 23.2261,
+      "step": 3704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141586900030774e-05,
+      "loss": 23.1909,
+      "step": 3705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141063609588863e-05,
+      "loss": 0.7613,
+      "step": 3706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140540174682942e-05,
+      "loss": 22.8887,
+      "step": 3707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140016595331274e-05,
+      "loss": 0.7462,
+      "step": 3708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13949287155212e-05,
+      "loss": 20.6362,
+      "step": 3709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138969003363756e-05,
+      "loss": 22.4532,
+      "step": 3710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138444990784453e-05,
+      "loss": 21.3968,
+      "step": 3711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137920833832494e-05,
+      "loss": 21.0674,
+      "step": 3712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137396532526164e-05,
+      "loss": 21.8917,
+      "step": 3713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136872086883755e-05,
+      "loss": 22.5912,
+      "step": 3714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136347496923559e-05,
+      "loss": 21.6983,
+      "step": 3715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13582276266388e-05,
+      "loss": 0.7298,
+      "step": 3716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135297884123023e-05,
+      "loss": 23.5124,
+      "step": 3717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.134772861319298e-05,
+      "loss": 0.6976,
+      "step": 3718
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.134247694271021e-05,
+      "loss": 22.7739,
+      "step": 3719
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133722382996514e-05,
+      "loss": 0.7216,
+      "step": 3720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133196927514102e-05,
+      "loss": 0.68,
+      "step": 3721
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132671327842115e-05,
+      "loss": 23.4229,
+      "step": 3722
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13214558399889e-05,
+      "loss": 0.7322,
+      "step": 3723
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131619696002769e-05,
+      "loss": 21.7981,
+      "step": 3724
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131093663872095e-05,
+      "loss": 0.7543,
+      "step": 3725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130567487625221e-05,
+      "loss": 21.0707,
+      "step": 3726
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130041167280504e-05,
+      "loss": 21.0009,
+      "step": 3727
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129514702856303e-05,
+      "loss": 0.6935,
+      "step": 3728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128988094370985e-05,
+      "loss": 22.8899,
+      "step": 3729
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128461341842923e-05,
+      "loss": 0.7165,
+      "step": 3730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127934445290488e-05,
+      "loss": 23.0426,
+      "step": 3731
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127407404732066e-05,
+      "loss": 22.3249,
+      "step": 3732
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126880220186042e-05,
+      "loss": 0.7221,
+      "step": 3733
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126352891670806e-05,
+      "loss": 23.3632,
+      "step": 3734
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125825419204754e-05,
+      "loss": 22.018,
+      "step": 3735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12529780280629e-05,
+      "loss": 22.0065,
+      "step": 3736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12477004249382e-05,
+      "loss": 22.6511,
+      "step": 3737
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124242138285751e-05,
+      "loss": 0.756,
+      "step": 3738
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123714090200504e-05,
+      "loss": 22.2014,
+      "step": 3739
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123185898256496e-05,
+      "loss": 21.5872,
+      "step": 3740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122657562472159e-05,
+      "loss": 22.3139,
+      "step": 3741
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12212908286592e-05,
+      "loss": 24.0947,
+      "step": 3742
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121600459456216e-05,
+      "loss": 23.7106,
+      "step": 3743
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121071692261492e-05,
+      "loss": 0.7471,
+      "step": 3744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120542781300188e-05,
+      "loss": 21.6789,
+      "step": 3745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12001372659076e-05,
+      "loss": 23.426,
+      "step": 3746
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119484528151664e-05,
+      "loss": 24.4157,
+      "step": 3747
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118955186001362e-05,
+      "loss": 24.1529,
+      "step": 3748
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118425700158317e-05,
+      "loss": 21.1539,
+      "step": 3749
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117896070641003e-05,
+      "loss": 0.738,
+      "step": 3750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117366297467897e-05,
+      "loss": 22.3363,
+      "step": 3751
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116836380657481e-05,
+      "loss": 0.7921,
+      "step": 3752
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116306320228239e-05,
+      "loss": 22.0565,
+      "step": 3753
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115776116198664e-05,
+      "loss": 0.7439,
+      "step": 3754
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115245768587254e-05,
+      "loss": 0.7748,
+      "step": 3755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114715277412508e-05,
+      "loss": 0.7324,
+      "step": 3756
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114184642692933e-05,
+      "loss": 22.8037,
+      "step": 3757
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113653864447042e-05,
+      "loss": 22.2957,
+      "step": 3758
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113122942693351e-05,
+      "loss": 22.2921,
+      "step": 3759
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112591877450383e-05,
+      "loss": 23.464,
+      "step": 3760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11206066873666e-05,
+      "loss": 24.6517,
+      "step": 3761
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111529316570718e-05,
+      "loss": 22.8915,
+      "step": 3762
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110997820971092e-05,
+      "loss": 0.6856,
+      "step": 3763
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110466181956324e-05,
+      "loss": 21.9503,
+      "step": 3764
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109934399544961e-05,
+      "loss": 0.7203,
+      "step": 3765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109402473755553e-05,
+      "loss": 0.7631,
+      "step": 3766
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108870404606657e-05,
+      "loss": 0.7044,
+      "step": 3767
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108338192116838e-05,
+      "loss": 0.6468,
+      "step": 3768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107805836304658e-05,
+      "loss": 23.1711,
+      "step": 3769
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107273337188691e-05,
+      "loss": 0.8065,
+      "step": 3770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106740694787512e-05,
+      "loss": 0.7357,
+      "step": 3771
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106207909119706e-05,
+      "loss": 0.7619,
+      "step": 3772
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105674980203855e-05,
+      "loss": 20.1563,
+      "step": 3773
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105141908058555e-05,
+      "loss": 23.4593,
+      "step": 3774
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.104608692702399e-05,
+      "loss": 0.78,
+      "step": 3775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10407533415399e-05,
+      "loss": 20.6973,
+      "step": 3776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103541832431937e-05,
+      "loss": 19.996,
+      "step": 3777
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103008187554848e-05,
+      "loss": 22.6798,
+      "step": 3778
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102474399541341e-05,
+      "loss": 21.674,
+      "step": 3779
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101940468410038e-05,
+      "loss": 22.9581,
+      "step": 3780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101406394179564e-05,
+      "loss": 0.7425,
+      "step": 3781
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100872176868553e-05,
+      "loss": 0.6846,
+      "step": 3782
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10033781649564e-05,
+      "loss": 24.6604,
+      "step": 3783
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099803313079467e-05,
+      "loss": 0.7619,
+      "step": 3784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09926866663868e-05,
+      "loss": 22.5101,
+      "step": 3785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098733877191929e-05,
+      "loss": 0.8182,
+      "step": 3786
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098198944757875e-05,
+      "loss": 23.3364,
+      "step": 3787
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097663869355176e-05,
+      "loss": 19.7107,
+      "step": 3788
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0971286510025e-05,
+      "loss": 21.3358,
+      "step": 3789
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096593289718518e-05,
+      "loss": 0.7544,
+      "step": 3790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096057785521905e-05,
+      "loss": 0.7526,
+      "step": 3791
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095522138431345e-05,
+      "loss": 0.7559,
+      "step": 3792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094986348465522e-05,
+      "loss": 23.3842,
+      "step": 3793
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09445041564313e-05,
+      "loss": 22.0036,
+      "step": 3794
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093914339982864e-05,
+      "loss": 0.7071,
+      "step": 3795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093378121503425e-05,
+      "loss": 0.7415,
+      "step": 3796
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09284176022352e-05,
+      "loss": 0.7356,
+      "step": 3797
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092305256161859e-05,
+      "loss": 23.1143,
+      "step": 3798
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09176860933716e-05,
+      "loss": 22.0957,
+      "step": 3799
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091231819768146e-05,
+      "loss": 0.704,
+      "step": 3800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090694887473539e-05,
+      "loss": 23.195,
+      "step": 3801
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090157812472073e-05,
+      "loss": 0.7764,
+      "step": 3802
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089620594782484e-05,
+      "loss": 21.9402,
+      "step": 3803
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089083234423513e-05,
+      "loss": 24.0075,
+      "step": 3804
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088545731413904e-05,
+      "loss": 21.6732,
+      "step": 3805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088008085772413e-05,
+      "loss": 21.2613,
+      "step": 3806
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087470297517792e-05,
+      "loss": 21.2178,
+      "step": 3807
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086932366668803e-05,
+      "loss": 21.5314,
+      "step": 3808
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086394293244214e-05,
+      "loss": 21.6341,
+      "step": 3809
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085856077262794e-05,
+      "loss": 19.9796,
+      "step": 3810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08531771874332e-05,
+      "loss": 0.7253,
+      "step": 3811
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084779217704573e-05,
+      "loss": 0.751,
+      "step": 3812
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084240574165338e-05,
+      "loss": 0.6635,
+      "step": 3813
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083701788144407e-05,
+      "loss": 21.2145,
+      "step": 3814
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083162859660577e-05,
+      "loss": 22.0315,
+      "step": 3815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082623788732647e-05,
+      "loss": 22.9397,
+      "step": 3816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082084575379423e-05,
+      "loss": 18.8113,
+      "step": 3817
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081545219619716e-05,
+      "loss": 0.7488,
+      "step": 3818
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081005721472343e-05,
+      "loss": 0.7334,
+      "step": 3819
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080466080956125e-05,
+      "loss": 0.6972,
+      "step": 3820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079926298089885e-05,
+      "loss": 0.7315,
+      "step": 3821
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079386372892456e-05,
+      "loss": 0.7347,
+      "step": 3822
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078846305382673e-05,
+      "loss": 0.7337,
+      "step": 3823
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078306095579376e-05,
+      "loss": 0.6895,
+      "step": 3824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077765743501413e-05,
+      "loss": 0.7252,
+      "step": 3825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077225249167632e-05,
+      "loss": 29.0341,
+      "step": 3826
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076684612596891e-05,
+      "loss": 0.7449,
+      "step": 3827
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076143833808047e-05,
+      "loss": 0.7355,
+      "step": 3828
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075602912819969e-05,
+      "loss": 24.1552,
+      "step": 3829
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075061849651524e-05,
+      "loss": 22.6684,
+      "step": 3830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074520644321592e-05,
+      "loss": 0.7397,
+      "step": 3831
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073979296849047e-05,
+      "loss": 0.681,
+      "step": 3832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073437807252781e-05,
+      "loss": 21.9835,
+      "step": 3833
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07289617555168e-05,
+      "loss": 25.6183,
+      "step": 3834
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072354401764641e-05,
+      "loss": 22.327,
+      "step": 3835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071812485910562e-05,
+      "loss": 0.8,
+      "step": 3836
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071270428008352e-05,
+      "loss": 23.4974,
+      "step": 3837
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070728228076916e-05,
+      "loss": 0.7472,
+      "step": 3838
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070185886135175e-05,
+      "loss": 23.6459,
+      "step": 3839
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069643402202043e-05,
+      "loss": 23.3989,
+      "step": 3840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069100776296448e-05,
+      "loss": 25.5823,
+      "step": 3841
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068558008437321e-05,
+      "loss": 22.4957,
+      "step": 3842
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068015098643595e-05,
+      "loss": 22.898,
+      "step": 3843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06747204693421e-05,
+      "loss": 22.1394,
+      "step": 3844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066928853328111e-05,
+      "loss": 24.6483,
+      "step": 3845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066385517844249e-05,
+      "loss": 0.7363,
+      "step": 3846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065842040501576e-05,
+      "loss": 0.7109,
+      "step": 3847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065298421319053e-05,
+      "loss": 20.7317,
+      "step": 3848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064754660315644e-05,
+      "loss": 21.6313,
+      "step": 3849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064210757510321e-05,
+      "loss": 22.1227,
+      "step": 3850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063666712922054e-05,
+      "loss": 0.7588,
+      "step": 3851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063122526569827e-05,
+      "loss": 0.7397,
+      "step": 3852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062578198472622e-05,
+      "loss": 0.7356,
+      "step": 3853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062033728649429e-05,
+      "loss": 0.7335,
+      "step": 3854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061489117119239e-05,
+      "loss": 0.7426,
+      "step": 3855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060944363901056e-05,
+      "loss": 21.3611,
+      "step": 3856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060399469013882e-05,
+      "loss": 23.4648,
+      "step": 3857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059854432476726e-05,
+      "loss": 0.7267,
+      "step": 3858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0593092543086e-05,
+      "loss": 24.488,
+      "step": 3859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058763934528528e-05,
+      "loss": 24.2974,
+      "step": 3860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058218473155528e-05,
+      "loss": 0.7202,
+      "step": 3861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057672870208634e-05,
+      "loss": 22.4163,
+      "step": 3862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057127125706877e-05,
+      "loss": 22.2187,
+      "step": 3863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056581239669297e-05,
+      "loss": 23.7229,
+      "step": 3864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056035212114935e-05,
+      "loss": 0.7058,
+      "step": 3865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.055489043062841e-05,
+      "loss": 21.0093,
+      "step": 3866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054942732532071e-05,
+      "loss": 22.2909,
+      "step": 3867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054396280541679e-05,
+      "loss": 21.8171,
+      "step": 3868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053849687110732e-05,
+      "loss": 23.2693,
+      "step": 3869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053302952258295e-05,
+      "loss": 19.8757,
+      "step": 3870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.052756076003446e-05,
+      "loss": 0.717,
+      "step": 3871
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.052209058365258e-05,
+      "loss": 22.9297,
+      "step": 3872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051661899362818e-05,
+      "loss": 0.7135,
+      "step": 3873
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051114599015211e-05,
+      "loss": 21.3485,
+      "step": 3874
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050567157341534e-05,
+      "loss": 25.5672,
+      "step": 3875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050019574360881e-05,
+      "loss": 20.4905,
+      "step": 3876
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049471850092356e-05,
+      "loss": 23.7802,
+      "step": 3877
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048923984555067e-05,
+      "loss": 0.7825,
+      "step": 3878
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048375977768126e-05,
+      "loss": 20.8052,
+      "step": 3879
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047827829750652e-05,
+      "loss": 21.7405,
+      "step": 3880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047279540521767e-05,
+      "loss": 0.7354,
+      "step": 3881
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046731110100599e-05,
+      "loss": 21.9275,
+      "step": 3882
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046182538506279e-05,
+      "loss": 23.0704,
+      "step": 3883
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045633825757943e-05,
+      "loss": 21.911,
+      "step": 3884
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045084971874738e-05,
+      "loss": 0.6675,
+      "step": 3885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044535976875807e-05,
+      "loss": 0.7148,
+      "step": 3886
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043986840780304e-05,
+      "loss": 0.7726,
+      "step": 3887
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043437563607385e-05,
+      "loss": 21.8838,
+      "step": 3888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.042888145376212e-05,
+      "loss": 19.6801,
+      "step": 3889
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.042338586105953e-05,
+      "loss": 24.6846,
+      "step": 3890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041788885815781e-05,
+      "loss": 24.2179,
+      "step": 3891
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041239044524867e-05,
+      "loss": 22.8282,
+      "step": 3892
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040689062252399e-05,
+      "loss": 21.9521,
+      "step": 3893
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04013893901756e-05,
+      "loss": 23.5619,
+      "step": 3894
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039588674839543e-05,
+      "loss": 0.725,
+      "step": 3895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039038269737542e-05,
+      "loss": 0.7545,
+      "step": 3896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038487723730762e-05,
+      "loss": 21.4244,
+      "step": 3897
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.037937036838407e-05,
+      "loss": 21.8475,
+      "step": 3898
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.037386209079687e-05,
+      "loss": 0.7082,
+      "step": 3899
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03683524047382e-05,
+      "loss": 23.1075,
+      "step": 3900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036284131040027e-05,
+      "loss": 23.1396,
+      "step": 3901
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035732880797532e-05,
+      "loss": 0.7478,
+      "step": 3902
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035181489765568e-05,
+      "loss": 22.9042,
+      "step": 3903
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034629957963368e-05,
+      "loss": 20.7272,
+      "step": 3904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034078285410174e-05,
+      "loss": 21.8558,
+      "step": 3905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033526472125233e-05,
+      "loss": 21.1506,
+      "step": 3906
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032974518127794e-05,
+      "loss": 22.1779,
+      "step": 3907
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032422423437113e-05,
+      "loss": 0.7335,
+      "step": 3908
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031870188072446e-05,
+      "loss": 0.7258,
+      "step": 3909
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031317812053065e-05,
+      "loss": 0.6906,
+      "step": 3910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030765295398237e-05,
+      "loss": 21.4503,
+      "step": 3911
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030212638127235e-05,
+      "loss": 20.671,
+      "step": 3912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029659840259342e-05,
+      "loss": 24.2532,
+      "step": 3913
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029106901813839e-05,
+      "loss": 20.4303,
+      "step": 3914
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028553822810021e-05,
+      "loss": 23.1218,
+      "step": 3915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028000603267178e-05,
+      "loss": 0.7297,
+      "step": 3916
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027447243204612e-05,
+      "loss": 23.5324,
+      "step": 3917
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026893742641625e-05,
+      "loss": 23.7863,
+      "step": 3918
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026340101597528e-05,
+      "loss": 21.8286,
+      "step": 3919
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025786320091633e-05,
+      "loss": 23.572,
+      "step": 3920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025232398143264e-05,
+      "loss": 0.6854,
+      "step": 3921
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024678335771741e-05,
+      "loss": 22.1489,
+      "step": 3922
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024124132996394e-05,
+      "loss": 23.4375,
+      "step": 3923
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023569789836554e-05,
+      "loss": 22.1272,
+      "step": 3924
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023015306311564e-05,
+      "loss": 24.6182,
+      "step": 3925
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022460682440767e-05,
+      "loss": 22.0402,
+      "step": 3926
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021905918243509e-05,
+      "loss": 0.7169,
+      "step": 3927
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021351013739143e-05,
+      "loss": 0.6774,
+      "step": 3928
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02079596894703e-05,
+      "loss": 22.4038,
+      "step": 3929
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020240783886533e-05,
+      "loss": 0.7213,
+      "step": 3930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019685458577018e-05,
+      "loss": 0.7773,
+      "step": 3931
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019129993037858e-05,
+      "loss": 0.7348,
+      "step": 3932
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018574387288432e-05,
+      "loss": 22.0636,
+      "step": 3933
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018018641348123e-05,
+      "loss": 21.874,
+      "step": 3934
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017462755236318e-05,
+      "loss": 19.9792,
+      "step": 3935
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016906728972408e-05,
+      "loss": 0.788,
+      "step": 3936
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016350562575793e-05,
+      "loss": 23.1076,
+      "step": 3937
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015794256065875e-05,
+      "loss": 0.6977,
+      "step": 3938
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01523780946206e-05,
+      "loss": 23.1739,
+      "step": 3939
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014681222783759e-05,
+      "loss": 21.4427,
+      "step": 3940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014124496050391e-05,
+      "loss": 23.6435,
+      "step": 3941
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013567629281376e-05,
+      "loss": 0.7561,
+      "step": 3942
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013010622496144e-05,
+      "loss": 20.5414,
+      "step": 3943
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012453475714123e-05,
+      "loss": 22.2365,
+      "step": 3944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011896188954751e-05,
+      "loss": 21.251,
+      "step": 3945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011338762237469e-05,
+      "loss": 23.1535,
+      "step": 3946
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010781195581723e-05,
+      "loss": 23.6766,
+      "step": 3947
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010223489006966e-05,
+      "loss": 19.4974,
+      "step": 3948
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009665642532652e-05,
+      "loss": 24.3379,
+      "step": 3949
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009107656178241e-05,
+      "loss": 21.0019,
+      "step": 3950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.008549529963201e-05,
+      "loss": 22.4736,
+      "step": 3951
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007991263907003e-05,
+      "loss": 20.9853,
+      "step": 3952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00743285802912e-05,
+      "loss": 0.7283,
+      "step": 3953
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006874312349033e-05,
+      "loss": 0.7491,
+      "step": 3954
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00631562688623e-05,
+      "loss": 0.7027,
+      "step": 3955
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005756801660198e-05,
+      "loss": 22.3278,
+      "step": 3956
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005197836690433e-05,
+      "loss": 23.0211,
+      "step": 3957
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004638731996435e-05,
+      "loss": 0.6931,
+      "step": 3958
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004079487597709e-05,
+      "loss": 19.4949,
+      "step": 3959
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003520103513765e-05,
+      "loss": 0.7595,
+      "step": 3960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002960579764116e-05,
+      "loss": 20.1401,
+      "step": 3961
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002400916368284e-05,
+      "loss": 21.6696,
+      "step": 3962
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00184111334579e-05,
+      "loss": 20.9528,
+      "step": 3963
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.001281170716165e-05,
+      "loss": 0.7433,
+      "step": 3964
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000721088498945e-05,
+      "loss": 0.6562,
+      "step": 3965
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000160866713664e-05,
+      "loss": 22.1911,
+      "step": 3966
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.999600505379869e-05,
+      "loss": 22.8934,
+      "step": 3967
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.999040004517107e-05,
+      "loss": 0.7232,
+      "step": 3968
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998479364144935e-05,
+      "loss": 23.2247,
+      "step": 3969
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997918584282908e-05,
+      "loss": 21.812,
+      "step": 3970
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99735766495059e-05,
+      "loss": 24.0778,
+      "step": 3971
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996796606167548e-05,
+      "loss": 0.7191,
+      "step": 3972
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996235407953357e-05,
+      "loss": 22.8209,
+      "step": 3973
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995674070327593e-05,
+      "loss": 0.7039,
+      "step": 3974
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995112593309842e-05,
+      "loss": 0.7358,
+      "step": 3975
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994550976919687e-05,
+      "loss": 20.92,
+      "step": 3976
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993989221176722e-05,
+      "loss": 23.2372,
+      "step": 3977
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993427326100547e-05,
+      "loss": 21.0406,
+      "step": 3978
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99286529171076e-05,
+      "loss": 0.6719,
+      "step": 3979
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99230311802697e-05,
+      "loss": 20.1597,
+      "step": 3980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991740805068788e-05,
+      "loss": 23.7636,
+      "step": 3981
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991178352855833e-05,
+      "loss": 0.672,
+      "step": 3982
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.990615761407724e-05,
+      "loss": 0.7027,
+      "step": 3983
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99005303074409e-05,
+      "loss": 20.4722,
+      "step": 3984
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989490160884558e-05,
+      "loss": 0.7062,
+      "step": 3985
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98892715184877e-05,
+      "loss": 20.8671,
+      "step": 3986
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98836400365636e-05,
+      "loss": 0.7237,
+      "step": 3987
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98780071632698e-05,
+      "loss": 21.9213,
+      "step": 3988
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987237289880277e-05,
+      "loss": 0.719,
+      "step": 3989
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986673724335908e-05,
+      "loss": 21.042,
+      "step": 3990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986110019713534e-05,
+      "loss": 22.9199,
+      "step": 3991
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985546176032818e-05,
+      "loss": 23.7906,
+      "step": 3992
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984982193313432e-05,
+      "loss": 20.4194,
+      "step": 3993
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98441807157505e-05,
+      "loss": 22.207,
+      "step": 3994
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983853810837354e-05,
+      "loss": 0.7021,
+      "step": 3995
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983289411120026e-05,
+      "loss": 20.7595,
+      "step": 3996
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982724872442756e-05,
+      "loss": 0.7216,
+      "step": 3997
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982160194825237e-05,
+      "loss": 0.7562,
+      "step": 3998
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981595378287173e-05,
+      "loss": 0.7128,
+      "step": 3999
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981030422848264e-05,
+      "loss": 0.7403,
+      "step": 4000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980465328528219e-05,
+      "loss": 22.8873,
+      "step": 4001
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979900095346754e-05,
+      "loss": 21.657,
+      "step": 4002
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979334723323584e-05,
+      "loss": 0.7096,
+      "step": 4003
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978769212478436e-05,
+      "loss": 21.1041,
+      "step": 4004
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978203562831036e-05,
+      "loss": 0.681,
+      "step": 4005
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97763777440112e-05,
+      "loss": 21.8485,
+      "step": 4006
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977071847208421e-05,
+      "loss": 0.7463,
+      "step": 4007
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976505781272686e-05,
+      "loss": 20.5064,
+      "step": 4008
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975939576613661e-05,
+      "loss": 21.7354,
+      "step": 4009
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975373233251099e-05,
+      "loss": 20.8047,
+      "step": 4010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974806751204756e-05,
+      "loss": 21.4077,
+      "step": 4011
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974240130494397e-05,
+      "loss": 21.0452,
+      "step": 4012
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973673371139785e-05,
+      "loss": 21.9996,
+      "step": 4013
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973106473160695e-05,
+      "loss": 21.3087,
+      "step": 4014
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972539436576903e-05,
+      "loss": 0.697,
+      "step": 4015
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971972261408189e-05,
+      "loss": 22.5271,
+      "step": 4016
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971404947674341e-05,
+      "loss": 20.3196,
+      "step": 4017
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970837495395149e-05,
+      "loss": 0.706,
+      "step": 4018
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970269904590409e-05,
+      "loss": 22.1006,
+      "step": 4019
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969702175279923e-05,
+      "loss": 0.7729,
+      "step": 4020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969134307483497e-05,
+      "loss": 20.8155,
+      "step": 4021
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96856630122094e-05,
+      "loss": 23.7686,
+      "step": 4022
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967998156512068e-05,
+      "loss": 0.7423,
+      "step": 4023
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9674298733767e-05,
+      "loss": 0.7991,
+      "step": 4024
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96686145183466e-05,
+      "loss": 0.7528,
+      "step": 4025
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966292891905783e-05,
+      "loss": 21.241,
+      "step": 4026
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965724193609899e-05,
+      "loss": 21.9619,
+      "step": 4027
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965155356966848e-05,
+      "loss": 21.5307,
+      "step": 4028
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.964586381996476e-05,
+      "loss": 0.7389,
+      "step": 4029
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96401726871863e-05,
+      "loss": 24.3269,
+      "step": 4030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.963448017153166e-05,
+      "loss": 0.7414,
+      "step": 4031
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962878627319941e-05,
+      "loss": 22.8299,
+      "step": 4032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962309099238822e-05,
+      "loss": 20.3287,
+      "step": 4033
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961739432929672e-05,
+      "loss": 21.7827,
+      "step": 4034
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961169628412368e-05,
+      "loss": 20.0378,
+      "step": 4035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960599685706787e-05,
+      "loss": 0.7689,
+      "step": 4036
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960029604832811e-05,
+      "loss": 0.6913,
+      "step": 4037
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95945938581033e-05,
+      "loss": 21.3588,
+      "step": 4038
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958889028659234e-05,
+      "loss": 0.7276,
+      "step": 4039
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958318533399423e-05,
+      "loss": 22.9491,
+      "step": 4040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957747900050797e-05,
+      "loss": 0.7109,
+      "step": 4041
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957177128633263e-05,
+      "loss": 21.6156,
+      "step": 4042
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956606219166732e-05,
+      "loss": 0.7545,
+      "step": 4043
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956035171671125e-05,
+      "loss": 22.2619,
+      "step": 4044
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955463986166358e-05,
+      "loss": 0.7321,
+      "step": 4045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954892662672358e-05,
+      "loss": 0.7238,
+      "step": 4046
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95432120120906e-05,
+      "loss": 0.6912,
+      "step": 4047
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953749601796397e-05,
+      "loss": 24.1697,
+      "step": 4048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953177864454311e-05,
+      "loss": 23.5698,
+      "step": 4049
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952605989202745e-05,
+      "loss": 0.7835,
+      "step": 4050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952033976061652e-05,
+      "loss": 23.357,
+      "step": 4051
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951461825050985e-05,
+      "loss": 24.1288,
+      "step": 4052
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950889536190705e-05,
+      "loss": 0.7299,
+      "step": 4053
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950317109500776e-05,
+      "loss": 22.7647,
+      "step": 4054
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94974454500117e-05,
+      "loss": 0.7851,
+      "step": 4055
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.949171842711858e-05,
+      "loss": 0.7593,
+      "step": 4056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94859900265282e-05,
+      "loss": 0.7568,
+      "step": 4057
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948026024844041e-05,
+      "loss": 20.4812,
+      "step": 4058
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94745290930551e-05,
+      "loss": 0.7701,
+      "step": 4059
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946879656057219e-05,
+      "loss": 0.6907,
+      "step": 4060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946306265119167e-05,
+      "loss": 22.6706,
+      "step": 4061
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945732736511357e-05,
+      "loss": 20.12,
+      "step": 4062
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945159070253797e-05,
+      "loss": 0.7151,
+      "step": 4063
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944585266366501e-05,
+      "loss": 0.7337,
+      "step": 4064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944011324869486e-05,
+      "loss": 21.3647,
+      "step": 4065
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943437245782772e-05,
+      "loss": 20.7246,
+      "step": 4066
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94286302912639e-05,
+      "loss": 22.6552,
+      "step": 4067
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942288674920367e-05,
+      "loss": 21.7924,
+      "step": 4068
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941714183184746e-05,
+      "loss": 23.2812,
+      "step": 4069
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941139553939563e-05,
+      "loss": 0.745,
+      "step": 4070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94056478720487e-05,
+      "loss": 20.1495,
+      "step": 4071
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939989883000712e-05,
+      "loss": 20.9833,
+      "step": 4072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939414841347149e-05,
+      "loss": 20.6844,
+      "step": 4073
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.93883966226424e-05,
+      "loss": 0.7487,
+      "step": 4074
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938264345772052e-05,
+      "loss": 0.7502,
+      "step": 4075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937688891890655e-05,
+      "loss": 0.7128,
+      "step": 4076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93711330064012e-05,
+      "loss": 0.7058,
+      "step": 4077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936537572040535e-05,
+      "loss": 24.4519,
+      "step": 4078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935961706111977e-05,
+      "loss": 0.75,
+      "step": 4079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93538570287454e-05,
+      "loss": 0.6729,
+      "step": 4080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934809562348317e-05,
+      "loss": 21.1869,
+      "step": 4081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934233284553407e-05,
+      "loss": 23.6115,
+      "step": 4082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933656869509914e-05,
+      "loss": 0.7618,
+      "step": 4083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933080317237948e-05,
+      "loss": 0.7009,
+      "step": 4084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932503627757617e-05,
+      "loss": 0.7511,
+      "step": 4085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931926801089047e-05,
+      "loss": 20.708,
+      "step": 4086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931349837252356e-05,
+      "loss": 22.7102,
+      "step": 4087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930772736267674e-05,
+      "loss": 22.2965,
+      "step": 4088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930195498155132e-05,
+      "loss": 0.6716,
+      "step": 4089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929618122934869e-05,
+      "loss": 0.7442,
+      "step": 4090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929040610627024e-05,
+      "loss": 23.1285,
+      "step": 4091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928462961251747e-05,
+      "loss": 20.5296,
+      "step": 4092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927885174829189e-05,
+      "loss": 21.201,
+      "step": 4093
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927307251379507e-05,
+      "loss": 22.4288,
+      "step": 4094
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926729190922861e-05,
+      "loss": 22.0438,
+      "step": 4095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926150993479419e-05,
+      "loss": 22.2112,
+      "step": 4096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92557265906935e-05,
+      "loss": 0.6919,
+      "step": 4097
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924994187712828e-05,
+      "loss": 21.6443,
+      "step": 4098
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924415579430038e-05,
+      "loss": 22.2446,
+      "step": 4099
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923836834241161e-05,
+      "loss": 21.0382,
+      "step": 4100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923257952166391e-05,
+      "loss": 22.3268,
+      "step": 4101
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922678933225918e-05,
+      "loss": 0.6877,
+      "step": 4102
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922099777439946e-05,
+      "loss": 0.6826,
+      "step": 4103
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921520484828675e-05,
+      "loss": 23.8714,
+      "step": 4104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920941055412317e-05,
+      "loss": 23.1962,
+      "step": 4105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920361489211087e-05,
+      "loss": 0.7381,
+      "step": 4106
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919781786245198e-05,
+      "loss": 19.9078,
+      "step": 4107
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919201946534879e-05,
+      "loss": 0.7523,
+      "step": 4108
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918621970100356e-05,
+      "loss": 0.6776,
+      "step": 4109
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918041856961862e-05,
+      "loss": 21.9456,
+      "step": 4110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917461607139634e-05,
+      "loss": 21.5532,
+      "step": 4111
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916881220653916e-05,
+      "loss": 20.2381,
+      "step": 4112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916300697524955e-05,
+      "loss": 0.7822,
+      "step": 4113
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915720037773e-05,
+      "loss": 0.693,
+      "step": 4114
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915139241418311e-05,
+      "loss": 24.3318,
+      "step": 4115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914558308481149e-05,
+      "loss": 0.7337,
+      "step": 4116
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913977238981778e-05,
+      "loss": 21.4031,
+      "step": 4117
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913396032940473e-05,
+      "loss": 23.5487,
+      "step": 4118
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912814690377506e-05,
+      "loss": 0.7142,
+      "step": 4119
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912233211313161e-05,
+      "loss": 21.8493,
+      "step": 4120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911651595767721e-05,
+      "loss": 0.7487,
+      "step": 4121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911069843761474e-05,
+      "loss": 0.6878,
+      "step": 4122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910487955314721e-05,
+      "loss": 0.6839,
+      "step": 4123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909905930447756e-05,
+      "loss": 22.4016,
+      "step": 4124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909323769180886e-05,
+      "loss": 21.5925,
+      "step": 4125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.908741471534419e-05,
+      "loss": 21.1729,
+      "step": 4126
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90815903752867e-05,
+      "loss": 0.6878,
+      "step": 4127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.907576467183956e-05,
+      "loss": 0.8139,
+      "step": 4128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906993760520601e-05,
+      "loss": 0.6988,
+      "step": 4129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906410917558935e-05,
+      "loss": 20.9717,
+      "step": 4130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905827938319289e-05,
+      "loss": 0.7216,
+      "step": 4131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905244822822e-05,
+      "loss": 22.6152,
+      "step": 4132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.904661571087413e-05,
+      "loss": 21.8292,
+      "step": 4133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.904078183135873e-05,
+      "loss": 20.5587,
+      "step": 4134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.903494658987733e-05,
+      "loss": 21.8286,
+      "step": 4135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902910998663348e-05,
+      "loss": 0.7293,
+      "step": 4136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902327202183082e-05,
+      "loss": 20.4005,
+      "step": 4137
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9017432695673e-05,
+      "loss": 22.2274,
+      "step": 4138
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901159200836371e-05,
+      "loss": 0.6827,
+      "step": 4139
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900574996010673e-05,
+      "loss": 23.5555,
+      "step": 4140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899990655110586e-05,
+      "loss": 21.4212,
+      "step": 4141
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899406178156494e-05,
+      "loss": 23.1637,
+      "step": 4142
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89882156516879e-05,
+      "loss": 0.7111,
+      "step": 4143
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898236816167864e-05,
+      "loss": 22.1175,
+      "step": 4144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897651931174119e-05,
+      "loss": 0.7553,
+      "step": 4145
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897066910207958e-05,
+      "loss": 24.0603,
+      "step": 4146
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89648175328979e-05,
+      "loss": 23.6286,
+      "step": 4147
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89589646044003e-05,
+      "loss": 0.7497,
+      "step": 4148
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895311031679093e-05,
+      "loss": 21.9517,
+      "step": 4149
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894725467027403e-05,
+      "loss": 22.9425,
+      "step": 4150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894139766505392e-05,
+      "loss": 22.8592,
+      "step": 4151
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893553930133487e-05,
+      "loss": 23.5935,
+      "step": 4152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892967957932129e-05,
+      "loss": 22.1394,
+      "step": 4153
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892381849921757e-05,
+      "loss": 23.7838,
+      "step": 4154
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891795606122821e-05,
+      "loss": 21.2337,
+      "step": 4155
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89120922655577e-05,
+      "loss": 23.5396,
+      "step": 4156
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890622711241063e-05,
+      "loss": 20.6497,
+      "step": 4157
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890036060199158e-05,
+      "loss": 0.6671,
+      "step": 4158
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889449273450524e-05,
+      "loss": 22.2459,
+      "step": 4159
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888862351015629e-05,
+      "loss": 23.3069,
+      "step": 4160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888275292914948e-05,
+      "loss": 0.8087,
+      "step": 4161
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887688099168963e-05,
+      "loss": 0.6999,
+      "step": 4162
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887100769798156e-05,
+      "loss": 20.5196,
+      "step": 4163
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886513304823019e-05,
+      "loss": 0.7261,
+      "step": 4164
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885925704264044e-05,
+      "loss": 20.4977,
+      "step": 4165
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885337968141733e-05,
+      "loss": 23.607,
+      "step": 4166
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884750096476586e-05,
+      "loss": 0.6916,
+      "step": 4167
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884162089289114e-05,
+      "loss": 23.1268,
+      "step": 4168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883573946599828e-05,
+      "loss": 23.3284,
+      "step": 4169
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882985668429246e-05,
+      "loss": 21.978,
+      "step": 4170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882397254797893e-05,
+      "loss": 22.9275,
+      "step": 4171
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.881808705726293e-05,
+      "loss": 0.6657,
+      "step": 4172
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88122002123498e-05,
+      "loss": 21.3171,
+      "step": 4173
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880631201344491e-05,
+      "loss": 0.699,
+      "step": 4174
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880042246075365e-05,
+      "loss": 0.716,
+      "step": 4175
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87945315544815e-05,
+      "loss": 0.7214,
+      "step": 4176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878863929483398e-05,
+      "loss": 0.7249,
+      "step": 4177
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878274568201661e-05,
+      "loss": 0.6599,
+      "step": 4178
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877685071623502e-05,
+      "loss": 0.6797,
+      "step": 4179
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877095439769487e-05,
+      "loss": 22.3434,
+      "step": 4180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.876505672660183e-05,
+      "loss": 22.3285,
+      "step": 4181
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875915770316164e-05,
+      "loss": 22.6931,
+      "step": 4182
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875325732758012e-05,
+      "loss": 20.4791,
+      "step": 4183
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87473556000631e-05,
+      "loss": 23.2018,
+      "step": 4184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.874145252081647e-05,
+      "loss": 19.9936,
+      "step": 4185
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873554809004614e-05,
+      "loss": 0.6958,
+      "step": 4186
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872964230795812e-05,
+      "loss": 23.7382,
+      "step": 4187
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87237351747584e-05,
+      "loss": 21.1293,
+      "step": 4188
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871782669065309e-05,
+      "loss": 22.1265,
+      "step": 4189
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871191685584832e-05,
+      "loss": 21.3424,
+      "step": 4190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870600567055021e-05,
+      "loss": 20.8564,
+      "step": 4191
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870009313496502e-05,
+      "loss": 22.0477,
+      "step": 4192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869417924929899e-05,
+      "loss": 22.7438,
+      "step": 4193
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868826401375845e-05,
+      "loss": 21.2353,
+      "step": 4194
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868234742854974e-05,
+      "loss": 0.7729,
+      "step": 4195
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867642949387927e-05,
+      "loss": 0.7263,
+      "step": 4196
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86705102099535e-05,
+      "loss": 20.7818,
+      "step": 4197
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86645895769789e-05,
+      "loss": 0.7674,
+      "step": 4198
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.865866759516203e-05,
+      "loss": 0.6794,
+      "step": 4199
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86527442647095e-05,
+      "loss": 20.4858,
+      "step": 4200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864681958582795e-05,
+      "loss": 0.7279,
+      "step": 4201
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864089355872405e-05,
+      "loss": 0.7531,
+      "step": 4202
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863496618360452e-05,
+      "loss": 22.6479,
+      "step": 4203
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862903746067618e-05,
+      "loss": 0.716,
+      "step": 4204
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862310739014582e-05,
+      "loss": 21.3207,
+      "step": 4205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861717597222035e-05,
+      "loss": 22.3369,
+      "step": 4206
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861124320710666e-05,
+      "loss": 20.7395,
+      "step": 4207
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.860530909501172e-05,
+      "loss": 22.0459,
+      "step": 4208
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859937363614258e-05,
+      "loss": 21.8055,
+      "step": 4209
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859343683070626e-05,
+      "loss": 21.688,
+      "step": 4210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858749867890992e-05,
+      "loss": 0.7193,
+      "step": 4211
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858155918096067e-05,
+      "loss": 20.8218,
+      "step": 4212
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857561833706572e-05,
+      "loss": 22.4084,
+      "step": 4213
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856967614743234e-05,
+      "loss": 24.3403,
+      "step": 4214
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856373261226782e-05,
+      "loss": 0.7057,
+      "step": 4215
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855778773177952e-05,
+      "loss": 0.7515,
+      "step": 4216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855184150617479e-05,
+      "loss": 22.2169,
+      "step": 4217
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85458939356611e-05,
+      "loss": 0.7657,
+      "step": 4218
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.853994502044593e-05,
+      "loss": 0.7293,
+      "step": 4219
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.853399476073682e-05,
+      "loss": 21.5003,
+      "step": 4220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852804315674134e-05,
+      "loss": 22.8257,
+      "step": 4221
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85220902086671e-05,
+      "loss": 21.2561,
+      "step": 4222
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.851613591672181e-05,
+      "loss": 22.4276,
+      "step": 4223
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.851018028111315e-05,
+      "loss": 0.6973,
+      "step": 4224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850422330204892e-05,
+      "loss": 22.6368,
+      "step": 4225
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849826497973692e-05,
+      "loss": 23.6535,
+      "step": 4226
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8492305314385e-05,
+      "loss": 22.4743,
+      "step": 4227
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848634430620107e-05,
+      "loss": 22.8407,
+      "step": 4228
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84803819553931e-05,
+      "loss": 0.6987,
+      "step": 4229
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.847441826216909e-05,
+      "loss": 22.6616,
+      "step": 4230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846845322673707e-05,
+      "loss": 23.1454,
+      "step": 4231
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846248684930514e-05,
+      "loss": 0.8175,
+      "step": 4232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845651913008145e-05,
+      "loss": 0.7202,
+      "step": 4233
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845055006927418e-05,
+      "loss": 0.6862,
+      "step": 4234
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844457966709157e-05,
+      "loss": 20.7657,
+      "step": 4235
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843860792374191e-05,
+      "loss": 0.7513,
+      "step": 4236
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843263483943351e-05,
+      "loss": 23.7388,
+      "step": 4237
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842666041437477e-05,
+      "loss": 21.2714,
+      "step": 4238
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842068464877407e-05,
+      "loss": 23.0538,
+      "step": 4239
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841470754283992e-05,
+      "loss": 23.2921,
+      "step": 4240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840872909678082e-05,
+      "loss": 0.7224,
+      "step": 4241
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840274931080535e-05,
+      "loss": 20.49,
+      "step": 4242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839676818512208e-05,
+      "loss": 0.76,
+      "step": 4243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839078571993971e-05,
+      "loss": 22.2772,
+      "step": 4244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838480191546689e-05,
+      "loss": 0.7225,
+      "step": 4245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837881677191242e-05,
+      "loss": 22.2647,
+      "step": 4246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837283028948509e-05,
+      "loss": 21.9136,
+      "step": 4247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.836684246839371e-05,
+      "loss": 0.7285,
+      "step": 4248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83608533088472e-05,
+      "loss": 0.7096,
+      "step": 4249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835486281105447e-05,
+      "loss": 20.8704,
+      "step": 4250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834887097522453e-05,
+      "loss": 0.6772,
+      "step": 4251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834287780156641e-05,
+      "loss": 22.0193,
+      "step": 4252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833688329028915e-05,
+      "loss": 0.7784,
+      "step": 4253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833088744160191e-05,
+      "loss": 21.6849,
+      "step": 4254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832489025571384e-05,
+      "loss": 19.8107,
+      "step": 4255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831889173283416e-05,
+      "loss": 0.7242,
+      "step": 4256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831289187317215e-05,
+      "loss": 20.5845,
+      "step": 4257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830689067693709e-05,
+      "loss": 0.703,
+      "step": 4258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830088814433835e-05,
+      "loss": 21.1842,
+      "step": 4259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.829488427558534e-05,
+      "loss": 20.5878,
+      "step": 4260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828887907088753e-05,
+      "loss": 21.0981,
+      "step": 4261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828287253045435e-05,
+      "loss": 0.7973,
+      "step": 4262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827686465449539e-05,
+      "loss": 0.7154,
+      "step": 4263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827085544322024e-05,
+      "loss": 21.8325,
+      "step": 4264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826484489683852e-05,
+      "loss": 23.3112,
+      "step": 4265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825883301555991e-05,
+      "loss": 22.2404,
+      "step": 4266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825281979959417e-05,
+      "loss": 21.5502,
+      "step": 4267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824680524915103e-05,
+      "loss": 0.735,
+      "step": 4268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824078936444034e-05,
+      "loss": 0.7249,
+      "step": 4269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823477214567197e-05,
+      "loss": 23.8068,
+      "step": 4270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822875359305582e-05,
+      "loss": 22.0173,
+      "step": 4271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822273370680188e-05,
+      "loss": 21.0059,
+      "step": 4272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821671248712011e-05,
+      "loss": 0.746,
+      "step": 4273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821068993422062e-05,
+      "loss": 20.623,
+      "step": 4274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820466604831345e-05,
+      "loss": 0.7106,
+      "step": 4275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819864082960882e-05,
+      "loss": 20.6763,
+      "step": 4276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819261427831686e-05,
+      "loss": 23.2607,
+      "step": 4277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818658639464784e-05,
+      "loss": 0.7756,
+      "step": 4278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818055717881204e-05,
+      "loss": 20.2472,
+      "step": 4279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817452663101981e-05,
+      "loss": 23.7143,
+      "step": 4280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81684947514815e-05,
+      "loss": 0.7015,
+      "step": 4281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.816246154040756e-05,
+      "loss": 0.7393,
+      "step": 4282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.815642699800847e-05,
+      "loss": 0.7535,
+      "step": 4283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.815039112449472e-05,
+      "loss": 21.4331,
+      "step": 4284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81443539200769e-05,
+      "loss": 0.721,
+      "step": 4285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81383153849656e-05,
+      "loss": 19.8951,
+      "step": 4286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.813227551937151e-05,
+      "loss": 21.7731,
+      "step": 4287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81262343235053e-05,
+      "loss": 22.3575,
+      "step": 4288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812019179757776e-05,
+      "loss": 18.8664,
+      "step": 4289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811414794179964e-05,
+      "loss": 0.7377,
+      "step": 4290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810810275638183e-05,
+      "loss": 21.3776,
+      "step": 4291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81020562415352e-05,
+      "loss": 0.7251,
+      "step": 4292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809600839747067e-05,
+      "loss": 20.6856,
+      "step": 4293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808995922439927e-05,
+      "loss": 0.706,
+      "step": 4294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808390872253199e-05,
+      "loss": 0.7727,
+      "step": 4295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807785689207992e-05,
+      "loss": 0.7187,
+      "step": 4296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807180373325418e-05,
+      "loss": 22.3579,
+      "step": 4297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806574924626593e-05,
+      "loss": 20.1183,
+      "step": 4298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80596934313264e-05,
+      "loss": 21.7835,
+      "step": 4299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805363628864686e-05,
+      "loss": 21.0584,
+      "step": 4300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80475778184386e-05,
+      "loss": 22.067,
+      "step": 4301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.804151802091296e-05,
+      "loss": 0.7673,
+      "step": 4302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803545689628138e-05,
+      "loss": 0.6985,
+      "step": 4303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802939444475527e-05,
+      "loss": 22.1154,
+      "step": 4304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802333066654614e-05,
+      "loss": 22.7086,
+      "step": 4305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801726556186554e-05,
+      "loss": 19.7233,
+      "step": 4306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801119913092504e-05,
+      "loss": 0.7381,
+      "step": 4307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800513137393626e-05,
+      "loss": 21.6479,
+      "step": 4308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79990622911109e-05,
+      "loss": 22.2858,
+      "step": 4309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799299188266068e-05,
+      "loss": 0.7552,
+      "step": 4310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798692014879737e-05,
+      "loss": 21.4695,
+      "step": 4311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798084708973278e-05,
+      "loss": 0.6164,
+      "step": 4312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79747727056788e-05,
+      "loss": 20.5795,
+      "step": 4313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796869699684729e-05,
+      "loss": 21.2036,
+      "step": 4314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796261996345024e-05,
+      "loss": 21.8017,
+      "step": 4315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795654160569964e-05,
+      "loss": 0.6929,
+      "step": 4316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795046192380756e-05,
+      "loss": 21.2454,
+      "step": 4317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.794438091798607e-05,
+      "loss": 0.7162,
+      "step": 4318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793829858844731e-05,
+      "loss": 0.7961,
+      "step": 4319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793221493540347e-05,
+      "loss": 0.7796,
+      "step": 4320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79261299590668e-05,
+      "loss": 0.7349,
+      "step": 4321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792004365964956e-05,
+      "loss": 0.7318,
+      "step": 4322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.791395603736407e-05,
+      "loss": 26.3463,
+      "step": 4323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790786709242271e-05,
+      "loss": 20.253,
+      "step": 4324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790177682503791e-05,
+      "loss": 22.0861,
+      "step": 4325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789568523542213e-05,
+      "loss": 0.6629,
+      "step": 4326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788959232378785e-05,
+      "loss": 22.6531,
+      "step": 4327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788349809034767e-05,
+      "loss": 21.8917,
+      "step": 4328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787740253531415e-05,
+      "loss": 0.7572,
+      "step": 4329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787130565889996e-05,
+      "loss": 0.703,
+      "step": 4330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786520746131779e-05,
+      "loss": 0.7601,
+      "step": 4331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785910794278037e-05,
+      "loss": 22.8973,
+      "step": 4332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785300710350051e-05,
+      "loss": 20.8687,
+      "step": 4333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784690494369103e-05,
+      "loss": 0.7055,
+      "step": 4334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78408014635648e-05,
+      "loss": 0.6772,
+      "step": 4335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783469666333475e-05,
+      "loss": 0.6845,
+      "step": 4336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782859054321386e-05,
+      "loss": 23.3779,
+      "step": 4337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782248310341514e-05,
+      "loss": 0.7121,
+      "step": 4338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781637434415165e-05,
+      "loss": 24.2776,
+      "step": 4339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78102642656365e-05,
+      "loss": 22.205,
+      "step": 4340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780415286808283e-05,
+      "loss": 22.3692,
+      "step": 4341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779804015170386e-05,
+      "loss": 21.1305,
+      "step": 4342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779192611671285e-05,
+      "loss": 21.4411,
+      "step": 4343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778581076332306e-05,
+      "loss": 0.7476,
+      "step": 4344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777969409174782e-05,
+      "loss": 24.6445,
+      "step": 4345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777357610220056e-05,
+      "loss": 21.8032,
+      "step": 4346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776745679489469e-05,
+      "loss": 22.3542,
+      "step": 4347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776133617004367e-05,
+      "loss": 21.3872,
+      "step": 4348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775521422786104e-05,
+      "loss": 19.3585,
+      "step": 4349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774909096856036e-05,
+      "loss": 0.7651,
+      "step": 4350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774296639235526e-05,
+      "loss": 0.7632,
+      "step": 4351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773684049945938e-05,
+      "loss": 21.5021,
+      "step": 4352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773071329008644e-05,
+      "loss": 0.7721,
+      "step": 4353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772458476445018e-05,
+      "loss": 0.8335,
+      "step": 4354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771845492276442e-05,
+      "loss": 23.5824,
+      "step": 4355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771232376524297e-05,
+      "loss": 21.3248,
+      "step": 4356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770619129209976e-05,
+      "loss": 22.705,
+      "step": 4357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770005750354868e-05,
+      "loss": 22.3758,
+      "step": 4358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.769392239980376e-05,
+      "loss": 0.7169,
+      "step": 4359
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7687785981079e-05,
+      "loss": 21.1882,
+      "step": 4360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768164824758846e-05,
+      "loss": 0.7228,
+      "step": 4361
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767550919954629e-05,
+      "loss": 24.0475,
+      "step": 4362
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766936883716664e-05,
+      "loss": 0.6861,
+      "step": 4363
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766322716066372e-05,
+      "loss": 0.7333,
+      "step": 4364
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76570841702518e-05,
+      "loss": 21.6876,
+      "step": 4365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.765093986614516e-05,
+      "loss": 19.4224,
+      "step": 4366
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764479424855817e-05,
+      "loss": 0.7111,
+      "step": 4367
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763864731770522e-05,
+      "loss": 21.724,
+      "step": 4368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763249907380074e-05,
+      "loss": 22.7569,
+      "step": 4369
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762634951705922e-05,
+      "loss": 0.8007,
+      "step": 4370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762019864769518e-05,
+      "loss": 22.3383,
+      "step": 4371
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761404646592322e-05,
+      "loss": 22.7126,
+      "step": 4372
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760789297195796e-05,
+      "loss": 20.9943,
+      "step": 4373
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760173816601406e-05,
+      "loss": 0.7345,
+      "step": 4374
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759558204830624e-05,
+      "loss": 21.9019,
+      "step": 4375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758942461904925e-05,
+      "loss": 21.6868,
+      "step": 4376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758326587845791e-05,
+      "loss": 19.4165,
+      "step": 4377
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757710582674707e-05,
+      "loss": 21.4048,
+      "step": 4378
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757094446413163e-05,
+      "loss": 23.1546,
+      "step": 4379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756478179082652e-05,
+      "loss": 20.4465,
+      "step": 4380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755861780704674e-05,
+      "loss": 20.5669,
+      "step": 4381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755245251300732e-05,
+      "loss": 20.0007,
+      "step": 4382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754628590892336e-05,
+      "loss": 0.7171,
+      "step": 4383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754011799500994e-05,
+      "loss": 19.9288,
+      "step": 4384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75339487714823e-05,
+      "loss": 19.9046,
+      "step": 4385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752777823855558e-05,
+      "loss": 23.1866,
+      "step": 4386
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75216063964451e-05,
+      "loss": 0.7407,
+      "step": 4387
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751543324536617e-05,
+      "loss": 23.2862,
+      "step": 4388
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75092587855341e-05,
+      "loss": 19.4089,
+      "step": 4389
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.750308301716433e-05,
+      "loss": 20.8379,
+      "step": 4390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74969059404723e-05,
+      "loss": 0.6572,
+      "step": 4391
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749072755567351e-05,
+      "loss": 0.7222,
+      "step": 4392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.748454786298346e-05,
+      "loss": 22.3099,
+      "step": 4393
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747836686261775e-05,
+      "loss": 22.5611,
+      "step": 4394
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747218455479204e-05,
+      "loss": 0.753,
+      "step": 4395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746600093972197e-05,
+      "loss": 21.7537,
+      "step": 4396
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745981601762327e-05,
+      "loss": 21.7823,
+      "step": 4397
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74536297887117e-05,
+      "loss": 0.7528,
+      "step": 4398
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744744225320307e-05,
+      "loss": 20.0922,
+      "step": 4399
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744125341131325e-05,
+      "loss": 20.1674,
+      "step": 4400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743506326325815e-05,
+      "loss": 21.0321,
+      "step": 4401
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742887180925367e-05,
+      "loss": 23.211,
+      "step": 4402
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742267904951587e-05,
+      "loss": 20.9202,
+      "step": 4403
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741648498426074e-05,
+      "loss": 0.7175,
+      "step": 4404
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741028961370437e-05,
+      "loss": 19.7194,
+      "step": 4405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740409293806292e-05,
+      "loss": 21.7537,
+      "step": 4406
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739789495755253e-05,
+      "loss": 22.1979,
+      "step": 4407
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739169567238945e-05,
+      "loss": 25.095,
+      "step": 4408
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738549508278993e-05,
+      "loss": 22.224,
+      "step": 4409
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737929318897027e-05,
+      "loss": 20.506,
+      "step": 4410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737308999114686e-05,
+      "loss": 22.5838,
+      "step": 4411
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.736688548953607e-05,
+      "loss": 21.563,
+      "step": 4412
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.736067968435438e-05,
+      "loss": 21.6711,
+      "step": 4413
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735447257581826e-05,
+      "loss": 21.6004,
+      "step": 4414
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734826416414426e-05,
+      "loss": 0.7103,
+      "step": 4415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734205444954895e-05,
+      "loss": 0.6773,
+      "step": 4416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733584343224899e-05,
+      "loss": 0.7288,
+      "step": 4417
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732963111246101e-05,
+      "loss": 20.1317,
+      "step": 4418
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73234174904018e-05,
+      "loss": 0.7751,
+      "step": 4419
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731720256628808e-05,
+      "loss": 0.6844,
+      "step": 4420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731098634033664e-05,
+      "loss": 19.4993,
+      "step": 4421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.730476881276439e-05,
+      "loss": 21.3304,
+      "step": 4422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72985499837882e-05,
+      "loss": 21.955,
+      "step": 4423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.729232985362504e-05,
+      "loss": 0.7722,
+      "step": 4424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728610842249187e-05,
+      "loss": 0.7097,
+      "step": 4425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727988569060577e-05,
+      "loss": 21.5209,
+      "step": 4426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72736616581838e-05,
+      "loss": 0.7363,
+      "step": 4427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726743632544309e-05,
+      "loss": 22.6013,
+      "step": 4428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726120969260081e-05,
+      "loss": 22.8127,
+      "step": 4429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72549817598742e-05,
+      "loss": 23.7017,
+      "step": 4430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724875252748053e-05,
+      "loss": 0.7412,
+      "step": 4431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724252199563709e-05,
+      "loss": 0.6334,
+      "step": 4432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723629016456125e-05,
+      "loss": 21.7318,
+      "step": 4433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72300570344704e-05,
+      "loss": 0.7594,
+      "step": 4434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.7223822605582e-05,
+      "loss": 23.6116,
+      "step": 4435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721758687811352e-05,
+      "loss": 21.3142,
+      "step": 4436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721134985228252e-05,
+      "loss": 0.6975,
+      "step": 4437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72051115283066e-05,
+      "loss": 21.5505,
+      "step": 4438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719887190640333e-05,
+      "loss": 0.6921,
+      "step": 4439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719263098679045e-05,
+      "loss": 0.7023,
+      "step": 4440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718638876968564e-05,
+      "loss": 22.99,
+      "step": 4441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718014525530667e-05,
+      "loss": 19.9056,
+      "step": 4442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717390044387135e-05,
+      "loss": 0.6902,
+      "step": 4443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716765433559753e-05,
+      "loss": 22.5991,
+      "step": 4444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716140693070314e-05,
+      "loss": 22.2352,
+      "step": 4445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715515822940607e-05,
+      "loss": 22.2572,
+      "step": 4446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714890823192439e-05,
+      "loss": 21.3939,
+      "step": 4447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714265693847606e-05,
+      "loss": 20.4475,
+      "step": 4448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713640434927918e-05,
+      "loss": 22.5694,
+      "step": 4449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713015046455191e-05,
+      "loss": 19.3187,
+      "step": 4450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.712389528451236e-05,
+      "loss": 21.1442,
+      "step": 4451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711763880937882e-05,
+      "loss": 21.2802,
+      "step": 4452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71113810393695e-05,
+      "loss": 19.8817,
+      "step": 4453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710512197470273e-05,
+      "loss": 0.723,
+      "step": 4454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709886161559686e-05,
+      "loss": 0.7948,
+      "step": 4455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709259996227027e-05,
+      "loss": 20.3038,
+      "step": 4456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708633701494141e-05,
+      "loss": 21.4167,
+      "step": 4457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708007277382877e-05,
+      "loss": 0.7896,
+      "step": 4458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70738072391509e-05,
+      "loss": 18.6614,
+      "step": 4459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706754041112634e-05,
+      "loss": 18.8372,
+      "step": 4460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706127228997376e-05,
+      "loss": 0.6929,
+      "step": 4461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70550028759118e-05,
+      "loss": 20.8951,
+      "step": 4462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704873216915915e-05,
+      "loss": 21.4049,
+      "step": 4463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70424601699346e-05,
+      "loss": 0.7654,
+      "step": 4464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703618687845696e-05,
+      "loss": 20.6262,
+      "step": 4465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702991229494506e-05,
+      "loss": 0.6777,
+      "step": 4466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70236364196178e-05,
+      "loss": 0.7351,
+      "step": 4467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70173592526941e-05,
+      "loss": 0.7502,
+      "step": 4468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701108079439298e-05,
+      "loss": 21.3021,
+      "step": 4469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700480104493345e-05,
+      "loss": 20.7365,
+      "step": 4470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699852000453457e-05,
+      "loss": 21.4811,
+      "step": 4471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699223767341549e-05,
+      "loss": 21.3078,
+      "step": 4472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698595405179532e-05,
+      "loss": 0.748,
+      "step": 4473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697966913989335e-05,
+      "loss": 19.5379,
+      "step": 4474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697338293792876e-05,
+      "loss": 21.6968,
+      "step": 4475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696709544612088e-05,
+      "loss": 0.6753,
+      "step": 4476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696080666468905e-05,
+      "loss": 0.7327,
+      "step": 4477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.695451659385265e-05,
+      "loss": 20.9338,
+      "step": 4478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694822523383113e-05,
+      "loss": 0.7518,
+      "step": 4479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694193258484396e-05,
+      "loss": 0.7057,
+      "step": 4480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693563864711067e-05,
+      "loss": 20.5304,
+      "step": 4481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692934342085082e-05,
+      "loss": 22.1783,
+      "step": 4482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692304690628403e-05,
+      "loss": 21.9125,
+      "step": 4483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691674910362995e-05,
+      "loss": 21.1666,
+      "step": 4484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691045001310829e-05,
+      "loss": 21.4454,
+      "step": 4485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690414963493878e-05,
+      "loss": 22.4389,
+      "step": 4486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689784796934125e-05,
+      "loss": 0.733,
+      "step": 4487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689154501653552e-05,
+      "loss": 0.7474,
+      "step": 4488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.688524077674147e-05,
+      "loss": 0.7068,
+      "step": 4489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687893525017902e-05,
+      "loss": 0.7425,
+      "step": 4490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687262843706816e-05,
+      "loss": 22.4111,
+      "step": 4491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.686632033762889e-05,
+      "loss": 0.7048,
+      "step": 4492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.686001095208129e-05,
+      "loss": 21.4035,
+      "step": 4493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.685370028064546e-05,
+      "loss": 0.7475,
+      "step": 4494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684738832354155e-05,
+      "loss": 0.735,
+      "step": 4495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684107508098976e-05,
+      "loss": 0.7108,
+      "step": 4496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683476055321034e-05,
+      "loss": 21.4523,
+      "step": 4497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682844474042356e-05,
+      "loss": 22.4376,
+      "step": 4498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682212764284976e-05,
+      "loss": 0.7788,
+      "step": 4499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.681580926070931e-05,
+      "loss": 22.8211,
+      "step": 4500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680948959422266e-05,
+      "loss": 23.1248,
+      "step": 4501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680316864361024e-05,
+      "loss": 23.3011,
+      "step": 4502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679684640909258e-05,
+      "loss": 22.2537,
+      "step": 4503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679052289089023e-05,
+      "loss": 0.7079,
+      "step": 4504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678419808922378e-05,
+      "loss": 22.4604,
+      "step": 4505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677787200431389e-05,
+      "loss": 0.719,
+      "step": 4506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677154463638127e-05,
+      "loss": 0.6816,
+      "step": 4507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.676521598564659e-05,
+      "loss": 0.6572,
+      "step": 4508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675888605233071e-05,
+      "loss": 22.0067,
+      "step": 4509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67525548366544e-05,
+      "loss": 19.6184,
+      "step": 4510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674622233883854e-05,
+      "loss": 21.1623,
+      "step": 4511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.673988855910406e-05,
+      "loss": 20.5091,
+      "step": 4512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.673355349767189e-05,
+      "loss": 22.8309,
+      "step": 4513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672721715476307e-05,
+      "loss": 21.3205,
+      "step": 4514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672087953059862e-05,
+      "loss": 21.5571,
+      "step": 4515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.671454062539964e-05,
+      "loss": 0.7112,
+      "step": 4516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670820043938728e-05,
+      "loss": 19.9405,
+      "step": 4517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67018589727827e-05,
+      "loss": 21.7136,
+      "step": 4518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669551622580713e-05,
+      "loss": 0.6991,
+      "step": 4519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668917219868185e-05,
+      "loss": 0.7726,
+      "step": 4520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668282689162816e-05,
+      "loss": 19.3088,
+      "step": 4521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667648030486746e-05,
+      "loss": 0.6778,
+      "step": 4522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667013243862113e-05,
+      "loss": 19.6473,
+      "step": 4523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66637832931106e-05,
+      "loss": 0.7431,
+      "step": 4524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665743286855741e-05,
+      "loss": 0.7106,
+      "step": 4525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665108116518306e-05,
+      "loss": 21.0187,
+      "step": 4526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664472818320913e-05,
+      "loss": 0.7121,
+      "step": 4527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66383739228573e-05,
+      "loss": 22.8684,
+      "step": 4528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663201838434918e-05,
+      "loss": 22.1049,
+      "step": 4529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662566156790653e-05,
+      "loss": 0.6479,
+      "step": 4530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661930347375108e-05,
+      "loss": 20.8206,
+      "step": 4531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661294410210467e-05,
+      "loss": 20.0429,
+      "step": 4532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660658345318915e-05,
+      "loss": 20.2039,
+      "step": 4533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660022152722639e-05,
+      "loss": 22.9728,
+      "step": 4534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659385832443834e-05,
+      "loss": 0.7406,
+      "step": 4535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658749384504699e-05,
+      "loss": 0.6993,
+      "step": 4536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658112808927437e-05,
+      "loss": 0.762,
+      "step": 4537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657476105734254e-05,
+      "loss": 0.733,
+      "step": 4538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656839274947364e-05,
+      "loss": 19.3502,
+      "step": 4539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656202316588982e-05,
+      "loss": 23.2173,
+      "step": 4540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.655565230681329e-05,
+      "loss": 20.1385,
+      "step": 4541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654928017246631e-05,
+      "loss": 0.7509,
+      "step": 4542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654290676307115e-05,
+      "loss": 0.7214,
+      "step": 4543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653653207885019e-05,
+      "loss": 0.7238,
+      "step": 4544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653015612002577e-05,
+      "loss": 0.6981,
+      "step": 4545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652377888682037e-05,
+      "loss": 0.7432,
+      "step": 4546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651740037945643e-05,
+      "loss": 22.6182,
+      "step": 4547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651102059815648e-05,
+      "loss": 20.7321,
+      "step": 4548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650463954314309e-05,
+      "loss": 0.696,
+      "step": 4549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649825721463884e-05,
+      "loss": 23.77,
+      "step": 4550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649187361286641e-05,
+      "loss": 22.7567,
+      "step": 4551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64854887380485e-05,
+      "loss": 0.7016,
+      "step": 4552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647910259040783e-05,
+      "loss": 0.7165,
+      "step": 4553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64727151701672e-05,
+      "loss": 21.8712,
+      "step": 4554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.646632647754942e-05,
+      "loss": 24.2864,
+      "step": 4555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64599365127774e-05,
+      "loss": 0.6923,
+      "step": 4556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645354527607402e-05,
+      "loss": 24.4601,
+      "step": 4557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644715276766227e-05,
+      "loss": 0.831,
+      "step": 4558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644075898776514e-05,
+      "loss": 0.7164,
+      "step": 4559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643436393660568e-05,
+      "loss": 0.6778,
+      "step": 4560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6427967614407e-05,
+      "loss": 24.6206,
+      "step": 4561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642157002139226e-05,
+      "loss": 21.188,
+      "step": 4562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64151711577846e-05,
+      "loss": 18.8528,
+      "step": 4563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640877102380726e-05,
+      "loss": 18.8218,
+      "step": 4564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640236961968355e-05,
+      "loss": 0.6837,
+      "step": 4565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639596694563673e-05,
+      "loss": 22.4293,
+      "step": 4566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638956300189021e-05,
+      "loss": 23.8914,
+      "step": 4567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638315778866738e-05,
+      "loss": 22.2386,
+      "step": 4568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63767513061917e-05,
+      "loss": 23.7703,
+      "step": 4569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.637034355468663e-05,
+      "loss": 21.8549,
+      "step": 4570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636393453437573e-05,
+      "loss": 20.8649,
+      "step": 4571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63575242454826e-05,
+      "loss": 0.7398,
+      "step": 4572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.635111268823084e-05,
+      "loss": 22.7867,
+      "step": 4573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634469986284415e-05,
+      "loss": 0.7328,
+      "step": 4574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633828576954622e-05,
+      "loss": 0.6615,
+      "step": 4575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633187040856083e-05,
+      "loss": 0.7019,
+      "step": 4576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.632545378011178e-05,
+      "loss": 0.7646,
+      "step": 4577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63190358844229e-05,
+      "loss": 20.367,
+      "step": 4578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.631261672171811e-05,
+      "loss": 0.7028,
+      "step": 4579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630619629222132e-05,
+      "loss": 20.1959,
+      "step": 4580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629977459615655e-05,
+      "loss": 0.6623,
+      "step": 4581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629335163374779e-05,
+      "loss": 23.1707,
+      "step": 4582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628692740521911e-05,
+      "loss": 22.2766,
+      "step": 4583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628050191079466e-05,
+      "loss": 22.8066,
+      "step": 4584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627407515069855e-05,
+      "loss": 20.1438,
+      "step": 4585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626764712515502e-05,
+      "loss": 20.0972,
+      "step": 4586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62612178343883e-05,
+      "loss": 0.7041,
+      "step": 4587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625478727862268e-05,
+      "loss": 0.7107,
+      "step": 4588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62483554580825e-05,
+      "loss": 19.2141,
+      "step": 4589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624192237299213e-05,
+      "loss": 21.3298,
+      "step": 4590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6235488023576e-05,
+      "loss": 20.4299,
+      "step": 4591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622905241005858e-05,
+      "loss": 20.0144,
+      "step": 4592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622261553266438e-05,
+      "loss": 19.6479,
+      "step": 4593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621617739161793e-05,
+      "loss": 0.7495,
+      "step": 4594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.620973798714386e-05,
+      "loss": 20.4247,
+      "step": 4595
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.620329731946682e-05,
+      "loss": 21.8571,
+      "step": 4596
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619685538881149e-05,
+      "loss": 20.7026,
+      "step": 4597
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619041219540255e-05,
+      "loss": 20.3948,
+      "step": 4598
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.618396773946487e-05,
+      "loss": 0.6895,
+      "step": 4599
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617752202122318e-05,
+      "loss": 0.6818,
+      "step": 4600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617107504090239e-05,
+      "loss": 21.4644,
+      "step": 4601
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61646267987274e-05,
+      "loss": 20.3188,
+      "step": 4602
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615817729492317e-05,
+      "loss": 0.739,
+      "step": 4603
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615172652971468e-05,
+      "loss": 0.7103,
+      "step": 4604
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.6145274503327e-05,
+      "loss": 0.7568,
+      "step": 4605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.613882121598515e-05,
+      "loss": 0.7169,
+      "step": 4606
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.613236666791431e-05,
+      "loss": 0.7535,
+      "step": 4607
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612591085933965e-05,
+      "loss": 21.9711,
+      "step": 4608
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611945379048635e-05,
+      "loss": 20.5478,
+      "step": 4609
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611299546157974e-05,
+      "loss": 0.6382,
+      "step": 4610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610653587284505e-05,
+      "loss": 20.8559,
+      "step": 4611
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610007502450765e-05,
+      "loss": 0.7594,
+      "step": 4612
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609361291679294e-05,
+      "loss": 20.8506,
+      "step": 4613
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608714954992633e-05,
+      "loss": 0.7189,
+      "step": 4614
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608068492413336e-05,
+      "loss": 0.745,
+      "step": 4615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60742190396395e-05,
+      "loss": 0.713,
+      "step": 4616
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606775189667032e-05,
+      "loss": 21.2064,
+      "step": 4617
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606128349545147e-05,
+      "loss": 19.353,
+      "step": 4618
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.605481383620855e-05,
+      "loss": 21.6718,
+      "step": 4619
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604834291916728e-05,
+      "loss": 0.7163,
+      "step": 4620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604187074455343e-05,
+      "loss": 19.3752,
+      "step": 4621
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603539731259276e-05,
+      "loss": 21.8844,
+      "step": 4622
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60289226235111e-05,
+      "loss": 22.2648,
+      "step": 4623
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602244667753433e-05,
+      "loss": 21.6454,
+      "step": 4624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601596947488837e-05,
+      "loss": 0.8037,
+      "step": 4625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600949101579918e-05,
+      "loss": 0.7531,
+      "step": 4626
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600301130049275e-05,
+      "loss": 20.2733,
+      "step": 4627
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599653032919517e-05,
+      "loss": 20.5671,
+      "step": 4628
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599004810213249e-05,
+      "loss": 20.4882,
+      "step": 4629
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.598356461953088e-05,
+      "loss": 0.7226,
+      "step": 4630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597707988161649e-05,
+      "loss": 0.7454,
+      "step": 4631
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597059388861556e-05,
+      "loss": 0.708,
+      "step": 4632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596410664075438e-05,
+      "loss": 0.6935,
+      "step": 4633
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595761813825924e-05,
+      "loss": 0.7567,
+      "step": 4634
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595112838135648e-05,
+      "loss": 21.0245,
+      "step": 4635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.594463737027254e-05,
+      "loss": 22.3081,
+      "step": 4636
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593814510523382e-05,
+      "loss": 20.7043,
+      "step": 4637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593165158646686e-05,
+      "loss": 0.7164,
+      "step": 4638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592515681419813e-05,
+      "loss": 19.7001,
+      "step": 4639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591866078865425e-05,
+      "loss": 21.637,
+      "step": 4640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591216351006181e-05,
+      "loss": 0.7752,
+      "step": 4641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590566497864749e-05,
+      "loss": 22.2623,
+      "step": 4642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5899165194638e-05,
+      "loss": 20.7056,
+      "step": 4643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589266415826007e-05,
+      "loss": 20.1087,
+      "step": 4644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588616186974052e-05,
+      "loss": 0.6977,
+      "step": 4645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587965832930616e-05,
+      "loss": 20.1668,
+      "step": 4646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587315353718388e-05,
+      "loss": 20.2205,
+      "step": 4647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586664749360063e-05,
+      "loss": 0.7671,
+      "step": 4648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586014019878332e-05,
+      "loss": 20.501,
+      "step": 4649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585363165295903e-05,
+      "loss": 20.7544,
+      "step": 4650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584712185635478e-05,
+      "loss": 21.2867,
+      "step": 4651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584061080919765e-05,
+      "loss": 20.4206,
+      "step": 4652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58340985117148e-05,
+      "loss": 19.1839,
+      "step": 4653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582758496413344e-05,
+      "loss": 21.8528,
+      "step": 4654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582107016668078e-05,
+      "loss": 21.5828,
+      "step": 4655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581455411958408e-05,
+      "loss": 0.6984,
+      "step": 4656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580803682307068e-05,
+      "loss": 22.0603,
+      "step": 4657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580151827736794e-05,
+      "loss": 0.6904,
+      "step": 4658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579499848270323e-05,
+      "loss": 21.5474,
+      "step": 4659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578847743930403e-05,
+      "loss": 22.4679,
+      "step": 4660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578195514739784e-05,
+      "loss": 0.7123,
+      "step": 4661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577543160721217e-05,
+      "loss": 22.2568,
+      "step": 4662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57689068189746e-05,
+      "loss": 0.7409,
+      "step": 4663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576238078291277e-05,
+      "loss": 21.3973,
+      "step": 4664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575585349925433e-05,
+      "loss": 0.6956,
+      "step": 4665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574932496822699e-05,
+      "loss": 0.7227,
+      "step": 4666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574279519005853e-05,
+      "loss": 21.1014,
+      "step": 4667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573626416497668e-05,
+      "loss": 0.7358,
+      "step": 4668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572973189320935e-05,
+      "loss": 0.6838,
+      "step": 4669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572319837498439e-05,
+      "loss": 21.5283,
+      "step": 4670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571666361052974e-05,
+      "loss": 0.7156,
+      "step": 4671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571012760007335e-05,
+      "loss": 0.7609,
+      "step": 4672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570359034384326e-05,
+      "loss": 0.6748,
+      "step": 4673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56970518420675e-05,
+      "loss": 0.6612,
+      "step": 4674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569051209497419e-05,
+      "loss": 19.8613,
+      "step": 4675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568397110279146e-05,
+      "loss": 20.3878,
+      "step": 4676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56774288657475e-05,
+      "loss": 0.763,
+      "step": 4677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.567088538407055e-05,
+      "loss": 19.4455,
+      "step": 4678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.566434065798889e-05,
+      "loss": 20.8362,
+      "step": 4679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56577946877308e-05,
+      "loss": 19.873,
+      "step": 4680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565124747352469e-05,
+      "loss": 20.2793,
+      "step": 4681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564469901559893e-05,
+      "loss": 20.3064,
+      "step": 4682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563814931418199e-05,
+      "loss": 22.7676,
+      "step": 4683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563159836950232e-05,
+      "loss": 19.7494,
+      "step": 4684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562504618178852e-05,
+      "loss": 22.5051,
+      "step": 4685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561849275126911e-05,
+      "loss": 22.6475,
+      "step": 4686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561193807817275e-05,
+      "loss": 20.7981,
+      "step": 4687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560538216272808e-05,
+      "loss": 21.347,
+      "step": 4688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559882500516381e-05,
+      "loss": 0.7729,
+      "step": 4689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559226660570869e-05,
+      "loss": 0.6978,
+      "step": 4690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558570696459154e-05,
+      "loss": 22.5002,
+      "step": 4691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557914608204119e-05,
+      "loss": 20.9371,
+      "step": 4692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557258395828648e-05,
+      "loss": 0.7021,
+      "step": 4693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556602059355639e-05,
+      "loss": 0.7684,
+      "step": 4694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.555945598807984e-05,
+      "loss": 21.2703,
+      "step": 4695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.555289014208588e-05,
+      "loss": 21.8368,
+      "step": 4696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554632305580354e-05,
+      "loss": 20.5565,
+      "step": 4697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.553975472946193e-05,
+      "loss": 0.8056,
+      "step": 4698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.553318516329019e-05,
+      "loss": 19.9387,
+      "step": 4699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552661435751749e-05,
+      "loss": 21.3777,
+      "step": 4700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552004231237307e-05,
+      "loss": 21.139,
+      "step": 4701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551346902808622e-05,
+      "loss": 0.7169,
+      "step": 4702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550689450488621e-05,
+      "loss": 0.7309,
+      "step": 4703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550031874300242e-05,
+      "loss": 21.0032,
+      "step": 4704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.549374174266424e-05,
+      "loss": 0.7288,
+      "step": 4705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548716350410114e-05,
+      "loss": 19.3072,
+      "step": 4706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548058402754258e-05,
+      "loss": 22.1547,
+      "step": 4707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547400331321811e-05,
+      "loss": 20.7403,
+      "step": 4708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546742136135729e-05,
+      "loss": 21.7197,
+      "step": 4709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546083817218973e-05,
+      "loss": 0.7077,
+      "step": 4710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545425374594511e-05,
+      "loss": 21.9699,
+      "step": 4711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544766808285311e-05,
+      "loss": 24.0501,
+      "step": 4712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544108118314348e-05,
+      "loss": 20.4474,
+      "step": 4713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.543449304704604e-05,
+      "loss": 18.5693,
+      "step": 4714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542790367479058e-05,
+      "loss": 0.6664,
+      "step": 4715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542131306660699e-05,
+      "loss": 21.2522,
+      "step": 4716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.541472122272518e-05,
+      "loss": 18.3035,
+      "step": 4717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540812814337515e-05,
+      "loss": 20.4358,
+      "step": 4718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540153382878686e-05,
+      "loss": 21.1445,
+      "step": 4719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539493827919036e-05,
+      "loss": 21.6162,
+      "step": 4720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.538834149481577e-05,
+      "loss": 21.7971,
+      "step": 4721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.538174347589321e-05,
+      "loss": 0.7065,
+      "step": 4722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537514422265284e-05,
+      "loss": 18.8795,
+      "step": 4723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53685437353249e-05,
+      "loss": 20.7139,
+      "step": 4724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536194201413965e-05,
+      "loss": 0.7337,
+      "step": 4725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535533905932738e-05,
+      "loss": 0.7158,
+      "step": 4726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534873487111845e-05,
+      "loss": 20.5957,
+      "step": 4727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534212944974327e-05,
+      "loss": 0.6701,
+      "step": 4728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.533552279543224e-05,
+      "loss": 20.651,
+      "step": 4729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.532891490841588e-05,
+      "loss": 22.0606,
+      "step": 4730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.532230578892467e-05,
+      "loss": 0.7253,
+      "step": 4731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531569543718918e-05,
+      "loss": 0.7718,
+      "step": 4732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530908385344005e-05,
+      "loss": 21.2351,
+      "step": 4733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530247103790789e-05,
+      "loss": 0.7036,
+      "step": 4734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.529585699082341e-05,
+      "loss": 0.7317,
+      "step": 4735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528924171241736e-05,
+      "loss": 0.7416,
+      "step": 4736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52826252029205e-05,
+      "loss": 0.6991,
+      "step": 4737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.527600746256365e-05,
+      "loss": 22.807,
+      "step": 4738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526938849157769e-05,
+      "loss": 19.8728,
+      "step": 4739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526276829019353e-05,
+      "loss": 18.6708,
+      "step": 4740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52561468586421e-05,
+      "loss": 24.0947,
+      "step": 4741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524952419715439e-05,
+      "loss": 0.7238,
+      "step": 4742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524290030596145e-05,
+      "loss": 0.7813,
+      "step": 4743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523627518529436e-05,
+      "loss": 0.7302,
+      "step": 4744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522964883538424e-05,
+      "loss": 22.1749,
+      "step": 4745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522302125646227e-05,
+      "loss": 22.6541,
+      "step": 4746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.521639244875963e-05,
+      "loss": 20.4574,
+      "step": 4747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52097624125076e-05,
+      "loss": 19.5599,
+      "step": 4748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520313114793743e-05,
+      "loss": 23.8596,
+      "step": 4749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51964986552805e-05,
+      "loss": 0.6872,
+      "step": 4750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518986493476818e-05,
+      "loss": 0.6508,
+      "step": 4751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518322998663189e-05,
+      "loss": 0.7063,
+      "step": 4752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517659381110308e-05,
+      "loss": 0.7603,
+      "step": 4753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516995640841327e-05,
+      "loss": 0.7311,
+      "step": 4754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5163317778794e-05,
+      "loss": 24.3013,
+      "step": 4755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.515667792247689e-05,
+      "loss": 25.5688,
+      "step": 4756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.515003683969354e-05,
+      "loss": 0.7059,
+      "step": 4757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514339453067566e-05,
+      "loss": 23.415,
+      "step": 4758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513675099565495e-05,
+      "loss": 23.9802,
+      "step": 4759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513010623486321e-05,
+      "loss": 20.6795,
+      "step": 4760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.512346024853219e-05,
+      "loss": 20.5163,
+      "step": 4761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511681303689378e-05,
+      "loss": 0.7314,
+      "step": 4762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511016460017988e-05,
+      "loss": 24.5553,
+      "step": 4763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51035149386224e-05,
+      "loss": 18.5757,
+      "step": 4764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509686405245332e-05,
+      "loss": 22.7681,
+      "step": 4765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509021194190469e-05,
+      "loss": 22.0851,
+      "step": 4766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508355860720853e-05,
+      "loss": 23.4702,
+      "step": 4767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.507690404859698e-05,
+      "loss": 0.7139,
+      "step": 4768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50702482663022e-05,
+      "loss": 21.4008,
+      "step": 4769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506359126055633e-05,
+      "loss": 0.7248,
+      "step": 4770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505693303159164e-05,
+      "loss": 0.7112,
+      "step": 4771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50502735796404e-05,
+      "loss": 20.9642,
+      "step": 4772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.504361290493493e-05,
+      "loss": 0.706,
+      "step": 4773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50369510077076e-05,
+      "loss": 20.8525,
+      "step": 4774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50302878881908e-05,
+      "loss": 0.7311,
+      "step": 4775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.5023623546617e-05,
+      "loss": 23.4007,
+      "step": 4776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501695798321867e-05,
+      "loss": 0.69,
+      "step": 4777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501029119822836e-05,
+      "loss": 22.0256,
+      "step": 4778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500362319187862e-05,
+      "loss": 0.7582,
+      "step": 4779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49969539644021e-05,
+      "loss": 21.8425,
+      "step": 4780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.499028351603144e-05,
+      "loss": 20.6211,
+      "step": 4781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498361184699934e-05,
+      "loss": 0.7043,
+      "step": 4782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497693895753856e-05,
+      "loss": 0.7438,
+      "step": 4783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497026484788189e-05,
+      "loss": 0.6978,
+      "step": 4784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.496358951826215e-05,
+      "loss": 0.7062,
+      "step": 4785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.495691296891221e-05,
+      "loss": 21.3071,
+      "step": 4786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4950235200065e-05,
+      "loss": 0.6634,
+      "step": 4787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494355621195347e-05,
+      "loss": 20.6578,
+      "step": 4788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.493687600481064e-05,
+      "loss": 22.3314,
+      "step": 4789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.493019457886953e-05,
+      "loss": 0.6991,
+      "step": 4790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492351193436321e-05,
+      "loss": 0.7303,
+      "step": 4791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491682807152487e-05,
+      "loss": 22.0659,
+      "step": 4792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491014299058762e-05,
+      "loss": 21.8189,
+      "step": 4793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490345669178471e-05,
+      "loss": 0.6559,
+      "step": 4794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489676917534938e-05,
+      "loss": 0.727,
+      "step": 4795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489008044151494e-05,
+      "loss": 20.7846,
+      "step": 4796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488339049051473e-05,
+      "loss": 21.1302,
+      "step": 4797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487669932258212e-05,
+      "loss": 0.7394,
+      "step": 4798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487000693795056e-05,
+      "loss": 21.7625,
+      "step": 4799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486331333685348e-05,
+      "loss": 21.4537,
+      "step": 4800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.485661851952442e-05,
+      "loss": 21.6111,
+      "step": 4801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484992248619694e-05,
+      "loss": 0.7206,
+      "step": 4802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484322523710463e-05,
+      "loss": 0.7067,
+      "step": 4803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48365267724811e-05,
+      "loss": 21.3457,
+      "step": 4804
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482982709256007e-05,
+      "loss": 22.8741,
+      "step": 4805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482312619757526e-05,
+      "loss": 20.368,
+      "step": 4806
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48164240877604e-05,
+      "loss": 0.7324,
+      "step": 4807
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480972076334933e-05,
+      "loss": 20.8651,
+      "step": 4808
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480301622457589e-05,
+      "loss": 0.6838,
+      "step": 4809
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.479631047167398e-05,
+      "loss": 19.0722,
+      "step": 4810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.478960350487752e-05,
+      "loss": 0.6478,
+      "step": 4811
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47828953244205e-05,
+      "loss": 21.3951,
+      "step": 4812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477618593053693e-05,
+      "loss": 0.6889,
+      "step": 4813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476947532346089e-05,
+      "loss": 21.7022,
+      "step": 4814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476276350342646e-05,
+      "loss": 20.115,
+      "step": 4815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47560504706678e-05,
+      "loss": 0.6437,
+      "step": 4816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47493362254191e-05,
+      "loss": 21.1071,
+      "step": 4817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47426207679146e-05,
+      "loss": 20.986,
+      "step": 4818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.473590409838856e-05,
+      "loss": 21.8996,
+      "step": 4819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472918621707528e-05,
+      "loss": 19.5215,
+      "step": 4820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472246712420917e-05,
+      "loss": 19.8053,
+      "step": 4821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471574682002458e-05,
+      "loss": 20.5745,
+      "step": 4822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470902530475598e-05,
+      "loss": 0.7149,
+      "step": 4823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470230257863782e-05,
+      "loss": 0.8022,
+      "step": 4824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469557864190467e-05,
+      "loss": 0.7856,
+      "step": 4825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468885349479108e-05,
+      "loss": 21.8593,
+      "step": 4826
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468212713753166e-05,
+      "loss": 20.3941,
+      "step": 4827
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467539957036106e-05,
+      "loss": 19.114,
+      "step": 4828
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4668670793514e-05,
+      "loss": 0.6914,
+      "step": 4829
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466194080722521e-05,
+      "loss": 0.7013,
+      "step": 4830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465520961172943e-05,
+      "loss": 0.7228,
+      "step": 4831
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464847720726155e-05,
+      "loss": 19.4841,
+      "step": 4832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464174359405637e-05,
+      "loss": 0.7435,
+      "step": 4833
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463500877234885e-05,
+      "loss": 21.0018,
+      "step": 4834
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462827274237391e-05,
+      "loss": 0.6721,
+      "step": 4835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462153550436653e-05,
+      "loss": 20.8703,
+      "step": 4836
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461479705856177e-05,
+      "loss": 0.7122,
+      "step": 4837
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460805740519469e-05,
+      "loss": 22.6594,
+      "step": 4838
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460131654450044e-05,
+      "loss": 21.9461,
+      "step": 4839
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459457447671413e-05,
+      "loss": 19.4654,
+      "step": 4840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4587831202071e-05,
+      "loss": 17.7448,
+      "step": 4841
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.458108672080624e-05,
+      "loss": 0.7583,
+      "step": 4842
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457434103315523e-05,
+      "loss": 21.8003,
+      "step": 4843
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.456759413935321e-05,
+      "loss": 21.2962,
+      "step": 4844
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.456084603963558e-05,
+      "loss": 19.0939,
+      "step": 4845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.455409673423777e-05,
+      "loss": 0.6974,
+      "step": 4846
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454734622339524e-05,
+      "loss": 18.3731,
+      "step": 4847
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454059450734345e-05,
+      "loss": 21.4162,
+      "step": 4848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453384158631795e-05,
+      "loss": 21.4878,
+      "step": 4849
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.452708746055433e-05,
+      "loss": 19.54,
+      "step": 4850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.452033213028822e-05,
+      "loss": 0.7182,
+      "step": 4851
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451357559575526e-05,
+      "loss": 0.7013,
+      "step": 4852
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450681785719117e-05,
+      "loss": 0.7346,
+      "step": 4853
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45000589148317e-05,
+      "loss": 19.2318,
+      "step": 4854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.449329876891264e-05,
+      "loss": 0.6994,
+      "step": 4855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448653741966983e-05,
+      "loss": 0.7069,
+      "step": 4856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447977486733912e-05,
+      "loss": 0.7207,
+      "step": 4857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447301111215646e-05,
+      "loss": 20.8589,
+      "step": 4858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.446624615435779e-05,
+      "loss": 21.4298,
+      "step": 4859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445947999417909e-05,
+      "loss": 19.9552,
+      "step": 4860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445271263185646e-05,
+      "loss": 19.1647,
+      "step": 4861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444594406762591e-05,
+      "loss": 0.7456,
+      "step": 4862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443917430172363e-05,
+      "loss": 0.757,
+      "step": 4863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443240333438575e-05,
+      "loss": 0.7872,
+      "step": 4864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442563116584851e-05,
+      "loss": 20.9237,
+      "step": 4865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.441885779634814e-05,
+      "loss": 20.7355,
+      "step": 4866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.441208322612095e-05,
+      "loss": 20.5294,
+      "step": 4867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440530745540325e-05,
+      "loss": 21.3903,
+      "step": 4868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439853048443142e-05,
+      "loss": 0.6946,
+      "step": 4869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439175231344192e-05,
+      "loss": 20.4063,
+      "step": 4870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.438497294267117e-05,
+      "loss": 0.6656,
+      "step": 4871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437819237235569e-05,
+      "loss": 19.422,
+      "step": 4872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437141060273201e-05,
+      "loss": 0.7405,
+      "step": 4873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436462763403674e-05,
+      "loss": 21.0685,
+      "step": 4874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43578434665065e-05,
+      "loss": 0.7617,
+      "step": 4875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435105810037795e-05,
+      "loss": 0.7228,
+      "step": 4876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.434427153588783e-05,
+      "loss": 21.2313,
+      "step": 4877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433748377327287e-05,
+      "loss": 20.8287,
+      "step": 4878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433069481276986e-05,
+      "loss": 19.5585,
+      "step": 4879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.432390465461565e-05,
+      "loss": 0.7655,
+      "step": 4880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431711329904712e-05,
+      "loss": 0.6878,
+      "step": 4881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431032074630118e-05,
+      "loss": 20.8681,
+      "step": 4882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.430352699661484e-05,
+      "loss": 21.2809,
+      "step": 4883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429673205022504e-05,
+      "loss": 19.5789,
+      "step": 4884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428993590736885e-05,
+      "loss": 19.0656,
+      "step": 4885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428313856828336e-05,
+      "loss": 18.5769,
+      "step": 4886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427634003320573e-05,
+      "loss": 22.8321,
+      "step": 4887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426954030237309e-05,
+      "loss": 23.4843,
+      "step": 4888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426273937602265e-05,
+      "loss": 22.061,
+      "step": 4889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42559372543917e-05,
+      "loss": 22.3784,
+      "step": 4890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424913393771751e-05,
+      "loss": 0.653,
+      "step": 4891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424232942623745e-05,
+      "loss": 20.4709,
+      "step": 4892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423552372018888e-05,
+      "loss": 22.5313,
+      "step": 4893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42287168198092e-05,
+      "loss": 23.0855,
+      "step": 4894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.422190872533591e-05,
+      "loss": 22.6567,
+      "step": 4895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42150994370065e-05,
+      "loss": 0.7217,
+      "step": 4896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420828895505852e-05,
+      "loss": 20.549,
+      "step": 4897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420147727972955e-05,
+      "loss": 0.7174,
+      "step": 4898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419466441125723e-05,
+      "loss": 21.099,
+      "step": 4899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418785034987921e-05,
+      "loss": 20.7315,
+      "step": 4900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418103509583323e-05,
+      "loss": 0.7251,
+      "step": 4901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417421864935707e-05,
+      "loss": 19.6847,
+      "step": 4902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.416740101068846e-05,
+      "loss": 0.7367,
+      "step": 4903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.416058218006529e-05,
+      "loss": 22.5947,
+      "step": 4904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415376215772542e-05,
+      "loss": 0.7512,
+      "step": 4905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.414694094390677e-05,
+      "loss": 0.7456,
+      "step": 4906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41401185388473e-05,
+      "loss": 22.6717,
+      "step": 4907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.4133294942785e-05,
+      "loss": 0.747,
+      "step": 4908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.412647015595796e-05,
+      "loss": 20.4449,
+      "step": 4909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411964417860426e-05,
+      "loss": 17.8663,
+      "step": 4910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411281701096199e-05,
+      "loss": 0.6793,
+      "step": 4911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410598865326934e-05,
+      "loss": 0.7347,
+      "step": 4912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409915910576453e-05,
+      "loss": 21.6094,
+      "step": 4913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409232836868582e-05,
+      "loss": 20.1865,
+      "step": 4914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40854964422715e-05,
+      "loss": 22.7681,
+      "step": 4915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407866332675988e-05,
+      "loss": 20.334,
+      "step": 4916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40718290223894e-05,
+      "loss": 0.7185,
+      "step": 4917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40649935293984e-05,
+      "loss": 0.8245,
+      "step": 4918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40581568480254e-05,
+      "loss": 21.6839,
+      "step": 4919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.405131897850887e-05,
+      "loss": 21.2774,
+      "step": 4920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404447992108738e-05,
+      "loss": 0.7784,
+      "step": 4921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40376396759995e-05,
+      "loss": 0.7505,
+      "step": 4922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403079824348385e-05,
+      "loss": 0.7242,
+      "step": 4923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.402395562377914e-05,
+      "loss": 20.6862,
+      "step": 4924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401711181712402e-05,
+      "loss": 0.6996,
+      "step": 4925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401026682375728e-05,
+      "loss": 18.286,
+      "step": 4926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.400342064391769e-05,
+      "loss": 20.3524,
+      "step": 4927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399657327784411e-05,
+      "loss": 0.6458,
+      "step": 4928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39897247257754e-05,
+      "loss": 0.73,
+      "step": 4929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.398287498795047e-05,
+      "loss": 20.6869,
+      "step": 4930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39760240646083e-05,
+      "loss": 20.8415,
+      "step": 4931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396917195598786e-05,
+      "loss": 19.7336,
+      "step": 4932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396231866232821e-05,
+      "loss": 20.1544,
+      "step": 4933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395546418386843e-05,
+      "loss": 0.6985,
+      "step": 4934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394860852084763e-05,
+      "loss": 22.9593,
+      "step": 4935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.3941751673505e-05,
+      "loss": 19.5197,
+      "step": 4936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.393489364207974e-05,
+      "loss": 21.5261,
+      "step": 4937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392803442681108e-05,
+      "loss": 0.7443,
+      "step": 4938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392117402793833e-05,
+      "loss": 0.7166,
+      "step": 4939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.391431244570078e-05,
+      "loss": 18.7427,
+      "step": 4940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390744968033786e-05,
+      "loss": 19.486,
+      "step": 4941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390058573208895e-05,
+      "loss": 20.2992,
+      "step": 4942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.389372060119348e-05,
+      "loss": 21.142,
+      "step": 4943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.388685428789099e-05,
+      "loss": 22.5666,
+      "step": 4944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3879986792421e-05,
+      "loss": 0.6645,
+      "step": 4945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.387311811502308e-05,
+      "loss": 22.0286,
+      "step": 4946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386624825593686e-05,
+      "loss": 20.2227,
+      "step": 4947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385937721540199e-05,
+      "loss": 21.5647,
+      "step": 4948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385250499365817e-05,
+      "loss": 17.335,
+      "step": 4949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.384563159094516e-05,
+      "loss": 0.6713,
+      "step": 4950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383875700750272e-05,
+      "loss": 19.8688,
+      "step": 4951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383188124357069e-05,
+      "loss": 20.1025,
+      "step": 4952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382500429938893e-05,
+      "loss": 0.7711,
+      "step": 4953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381812617519735e-05,
+      "loss": 0.6639,
+      "step": 4954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381124687123591e-05,
+      "loss": 0.7726,
+      "step": 4955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.380436638774458e-05,
+      "loss": 21.0302,
+      "step": 4956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37974847249634e-05,
+      "loss": 21.4714,
+      "step": 4957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379060188313244e-05,
+      "loss": 23.0868,
+      "step": 4958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378371786249182e-05,
+      "loss": 18.9798,
+      "step": 4959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377683266328167e-05,
+      "loss": 22.1315,
+      "step": 4960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376994628574219e-05,
+      "loss": 20.2037,
+      "step": 4961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376305873011366e-05,
+      "loss": 20.6367,
+      "step": 4962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375616999663632e-05,
+      "loss": 22.2806,
+      "step": 4963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.374928008555047e-05,
+      "loss": 0.7393,
+      "step": 4964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.374238899709652e-05,
+      "loss": 22.234,
+      "step": 4965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373549673151485e-05,
+      "loss": 18.8142,
+      "step": 4966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.372860328904586e-05,
+      "loss": 21.6546,
+      "step": 4967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37217086699301e-05,
+      "loss": 22.3482,
+      "step": 4968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371481287440805e-05,
+      "loss": 0.7519,
+      "step": 4969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370791590272029e-05,
+      "loss": 19.2881,
+      "step": 4970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370101775510742e-05,
+      "loss": 22.0604,
+      "step": 4971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.36941184318101e-05,
+      "loss": 18.9784,
+      "step": 4972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3687217933069e-05,
+      "loss": 0.7032,
+      "step": 4973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368031625912484e-05,
+      "loss": 23.4335,
+      "step": 4974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.367341341021844e-05,
+      "loss": 21.3529,
+      "step": 4975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366650938659055e-05,
+      "loss": 20.7671,
+      "step": 4976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365960418848206e-05,
+      "loss": 21.1801,
+      "step": 4977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365269781613385e-05,
+      "loss": 20.0044,
+      "step": 4978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364579026978687e-05,
+      "loss": 20.0875,
+      "step": 4979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.363888154968206e-05,
+      "loss": 0.7116,
+      "step": 4980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.363197165606047e-05,
+      "loss": 21.6524,
+      "step": 4981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362506058916315e-05,
+      "loss": 21.0876,
+      "step": 4982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361814834923119e-05,
+      "loss": 0.7375,
+      "step": 4983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361123493650571e-05,
+      "loss": 0.7109,
+      "step": 4984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360432035122792e-05,
+      "loss": 0.7478,
+      "step": 4985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.359740459363903e-05,
+      "loss": 0.7309,
+      "step": 4986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.359048766398031e-05,
+      "loss": 19.4666,
+      "step": 4987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358356956249305e-05,
+      "loss": 0.6764,
+      "step": 4988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.357665028941859e-05,
+      "loss": 18.8867,
+      "step": 4989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35697298449983e-05,
+      "loss": 22.9144,
+      "step": 4990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.356280822947365e-05,
+      "loss": 20.6926,
+      "step": 4991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355588544308606e-05,
+      "loss": 24.0017,
+      "step": 4992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.354896148607708e-05,
+      "loss": 20.4299,
+      "step": 4993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.354203635868823e-05,
+      "loss": 20.5492,
+      "step": 4994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353511006116108e-05,
+      "loss": 0.7432,
+      "step": 4995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35281825937373e-05,
+      "loss": 21.9465,
+      "step": 4996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.352125395665853e-05,
+      "loss": 0.756,
+      "step": 4997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35143241501665e-05,
+      "loss": 22.0286,
+      "step": 4998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350739317450294e-05,
+      "loss": 19.5865,
+      "step": 4999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350046102990967e-05,
+      "loss": 0.7658,
+      "step": 5000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349352771662849e-05,
+      "loss": 0.7424,
+      "step": 5001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.348659323490129e-05,
+      "loss": 0.6456,
+      "step": 5002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347965758496999e-05,
+      "loss": 21.9792,
+      "step": 5003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347272076707655e-05,
+      "loss": 20.7,
+      "step": 5004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.346578278146295e-05,
+      "loss": 20.848,
+      "step": 5005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345884362837123e-05,
+      "loss": 20.3136,
+      "step": 5006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345190330804346e-05,
+      "loss": 19.7799,
+      "step": 5007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344496182072179e-05,
+      "loss": 20.0927,
+      "step": 5008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343801916664835e-05,
+      "loss": 21.1657,
+      "step": 5009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343107534606535e-05,
+      "loss": 0.7543,
+      "step": 5010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342413035921502e-05,
+      "loss": 0.75,
+      "step": 5011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341718420633964e-05,
+      "loss": 0.6961,
+      "step": 5012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341023688768157e-05,
+      "loss": 0.6921,
+      "step": 5013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.34032884034831e-05,
+      "loss": 0.7449,
+      "step": 5014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339633875398671e-05,
+      "loss": 19.8853,
+      "step": 5015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338938793943478e-05,
+      "loss": 0.6616,
+      "step": 5016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338243596006985e-05,
+      "loss": 20.4824,
+      "step": 5017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33754828161344e-05,
+      "loss": 0.7565,
+      "step": 5018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3368528507871e-05,
+      "loss": 0.778,
+      "step": 5019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33615730355223e-05,
+      "loss": 23.2246,
+      "step": 5020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33546163993309e-05,
+      "loss": 0.7538,
+      "step": 5021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33476585995395e-05,
+      "loss": 21.7817,
+      "step": 5022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.334069963639083e-05,
+      "loss": 20.9234,
+      "step": 5023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333373951012766e-05,
+      "loss": 0.7643,
+      "step": 5024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332677822099281e-05,
+      "loss": 18.8509,
+      "step": 5025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33198157692291e-05,
+      "loss": 19.3362,
+      "step": 5026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.331285215507944e-05,
+      "loss": 22.3925,
+      "step": 5027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330588737878677e-05,
+      "loss": 0.7165,
+      "step": 5028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329892144059405e-05,
+      "loss": 0.7116,
+      "step": 5029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329195434074429e-05,
+      "loss": 0.7591,
+      "step": 5030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328498607948055e-05,
+      "loss": 22.0518,
+      "step": 5031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327801665704588e-05,
+      "loss": 0.7438,
+      "step": 5032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327104607368349e-05,
+      "loss": 0.6798,
+      "step": 5033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32640743296365e-05,
+      "loss": 21.1396,
+      "step": 5034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325710142514815e-05,
+      "loss": 0.7544,
+      "step": 5035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325012736046167e-05,
+      "loss": 19.5002,
+      "step": 5036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324315213582037e-05,
+      "loss": 0.7112,
+      "step": 5037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323617575146759e-05,
+      "loss": 20.8977,
+      "step": 5038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.322919820764669e-05,
+      "loss": 0.691,
+      "step": 5039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.322221950460111e-05,
+      "loss": 19.5678,
+      "step": 5040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321523964257431e-05,
+      "loss": 0.68,
+      "step": 5041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.320825862180974e-05,
+      "loss": 0.7724,
+      "step": 5042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.320127644255098e-05,
+      "loss": 21.4809,
+      "step": 5043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319429310504161e-05,
+      "loss": 19.1101,
+      "step": 5044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.318730860952522e-05,
+      "loss": 0.6981,
+      "step": 5045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31803229562455e-05,
+      "loss": 19.0188,
+      "step": 5046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317333614544614e-05,
+      "loss": 0.7239,
+      "step": 5047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.316634817737086e-05,
+      "loss": 0.6602,
+      "step": 5048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315935905226346e-05,
+      "loss": 19.4946,
+      "step": 5049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315236877036776e-05,
+      "loss": 0.7041,
+      "step": 5050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314537733192762e-05,
+      "loss": 21.5244,
+      "step": 5051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.313838473718694e-05,
+      "loss": 20.656,
+      "step": 5052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.313139098638965e-05,
+      "loss": 20.463,
+      "step": 5053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.312439607977976e-05,
+      "loss": 20.3203,
+      "step": 5054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311740001760126e-05,
+      "loss": 21.2753,
+      "step": 5055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311040280009826e-05,
+      "loss": 0.7075,
+      "step": 5056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.310340442751479e-05,
+      "loss": 0.7423,
+      "step": 5057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309640490009505e-05,
+      "loss": 20.5093,
+      "step": 5058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308940421808322e-05,
+      "loss": 18.4186,
+      "step": 5059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308240238172348e-05,
+      "loss": 0.714,
+      "step": 5060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.307539939126016e-05,
+      "loss": 22.4387,
+      "step": 5061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306839524693753e-05,
+      "loss": 22.4455,
+      "step": 5062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306138994899993e-05,
+      "loss": 0.7594,
+      "step": 5063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305438349769174e-05,
+      "loss": 0.7531,
+      "step": 5064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304737589325739e-05,
+      "loss": 21.1366,
+      "step": 5065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304036713594136e-05,
+      "loss": 20.1494,
+      "step": 5066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.303335722598816e-05,
+      "loss": 0.7153,
+      "step": 5067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302634616364231e-05,
+      "loss": 0.6995,
+      "step": 5068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301933394914842e-05,
+      "loss": 19.3954,
+      "step": 5069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301232058275108e-05,
+      "loss": 19.7786,
+      "step": 5070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3005306064695e-05,
+      "loss": 20.0708,
+      "step": 5071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299829039522487e-05,
+      "loss": 21.9179,
+      "step": 5072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299127357458542e-05,
+      "loss": 20.1304,
+      "step": 5073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298425560302146e-05,
+      "loss": 19.4082,
+      "step": 5074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.297723648077779e-05,
+      "loss": 20.3778,
+      "step": 5075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29702162080993e-05,
+      "loss": 0.6788,
+      "step": 5076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.296319478523091e-05,
+      "loss": 20.5584,
+      "step": 5077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295617221241755e-05,
+      "loss": 0.7082,
+      "step": 5078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.294914848990419e-05,
+      "loss": 0.6874,
+      "step": 5079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.294212361793588e-05,
+      "loss": 20.6724,
+      "step": 5080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.293509759675769e-05,
+      "loss": 18.7361,
+      "step": 5081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292807042661472e-05,
+      "loss": 20.2349,
+      "step": 5082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292104210775211e-05,
+      "loss": 19.7643,
+      "step": 5083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291401264041507e-05,
+      "loss": 0.6827,
+      "step": 5084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290698202484882e-05,
+      "loss": 19.8089,
+      "step": 5085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28999502612986e-05,
+      "loss": 0.7191,
+      "step": 5086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289291735000975e-05,
+      "loss": 0.7283,
+      "step": 5087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288588329122761e-05,
+      "loss": 21.2987,
+      "step": 5088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.287884808519756e-05,
+      "loss": 0.7117,
+      "step": 5089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.287181173216503e-05,
+      "loss": 20.3888,
+      "step": 5090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28647742323755e-05,
+      "loss": 21.1549,
+      "step": 5091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28577355860745e-05,
+      "loss": 18.7084,
+      "step": 5092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285069579350752e-05,
+      "loss": 0.756,
+      "step": 5093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.284365485492019e-05,
+      "loss": 0.7446,
+      "step": 5094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283661277055814e-05,
+      "loss": 0.6622,
+      "step": 5095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.2829569540667e-05,
+      "loss": 0.7235,
+      "step": 5096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282252516549251e-05,
+      "loss": 21.0802,
+      "step": 5097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281547964528043e-05,
+      "loss": 18.3285,
+      "step": 5098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280843298027652e-05,
+      "loss": 20.0678,
+      "step": 5099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28013851707266e-05,
+      "loss": 0.7709,
+      "step": 5100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279433621687657e-05,
+      "loss": 20.7423,
+      "step": 5101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278728611897232e-05,
+      "loss": 0.6817,
+      "step": 5102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278023487725982e-05,
+      "loss": 0.7669,
+      "step": 5103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.2773182491985e-05,
+      "loss": 20.0292,
+      "step": 5104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276612896339395e-05,
+      "loss": 0.683,
+      "step": 5105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275907429173272e-05,
+      "loss": 0.6876,
+      "step": 5106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275201847724741e-05,
+      "loss": 20.4809,
+      "step": 5107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274496152018414e-05,
+      "loss": 0.708,
+      "step": 5108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273790342078916e-05,
+      "loss": 19.7283,
+      "step": 5109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273084417930865e-05,
+      "loss": 22.462,
+      "step": 5110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.272378379598889e-05,
+      "loss": 22.7738,
+      "step": 5111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271672227107619e-05,
+      "loss": 19.1542,
+      "step": 5112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270965960481689e-05,
+      "loss": 20.6867,
+      "step": 5113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270259579745737e-05,
+      "loss": 20.2405,
+      "step": 5114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.269553084924407e-05,
+      "loss": 0.686,
+      "step": 5115
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268846476042346e-05,
+      "loss": 0.7109,
+      "step": 5116
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268139753124202e-05,
+      "loss": 21.585,
+      "step": 5117
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267432916194633e-05,
+      "loss": 0.6971,
+      "step": 5118
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.266725965278294e-05,
+      "loss": 0.7175,
+      "step": 5119
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26601890039985e-05,
+      "loss": 0.7339,
+      "step": 5120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265311721583966e-05,
+      "loss": 0.7069,
+      "step": 5121
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264604428855314e-05,
+      "loss": 20.326,
+      "step": 5122
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263897022238567e-05,
+      "loss": 19.6717,
+      "step": 5123
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263189501758404e-05,
+      "loss": 0.7237,
+      "step": 5124
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.262481867439507e-05,
+      "loss": 0.7856,
+      "step": 5125
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.261774119306563e-05,
+      "loss": 0.702,
+      "step": 5126
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.261066257384262e-05,
+      "loss": 22.1129,
+      "step": 5127
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.260358281697298e-05,
+      "loss": 0.7099,
+      "step": 5128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25965019227037e-05,
+      "loss": 0.7495,
+      "step": 5129
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258941989128177e-05,
+      "loss": 20.1767,
+      "step": 5130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25823367229543e-05,
+      "loss": 0.744,
+      "step": 5131
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.257525241796838e-05,
+      "loss": 21.6307,
+      "step": 5132
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256816697657113e-05,
+      "loss": 19.0033,
+      "step": 5133
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256108039900974e-05,
+      "loss": 20.6175,
+      "step": 5134
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255399268553144e-05,
+      "loss": 22.6081,
+      "step": 5135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.254690383638346e-05,
+      "loss": 21.8261,
+      "step": 5136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253981385181313e-05,
+      "loss": 20.2749,
+      "step": 5137
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253272273206778e-05,
+      "loss": 20.1467,
+      "step": 5138
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25256304773948e-05,
+      "loss": 19.1671,
+      "step": 5139
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25185370880416e-05,
+      "loss": 0.7681,
+      "step": 5140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.251144256425563e-05,
+      "loss": 0.7485,
+      "step": 5141
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250434690628439e-05,
+      "loss": 21.2434,
+      "step": 5142
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249725011437544e-05,
+      "loss": 0.6593,
+      "step": 5143
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249015218877632e-05,
+      "loss": 19.4665,
+      "step": 5144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248305312973465e-05,
+      "loss": 0.7392,
+      "step": 5145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.247595293749813e-05,
+      "loss": 20.5366,
+      "step": 5146
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.246885161231441e-05,
+      "loss": 0.7242,
+      "step": 5147
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.246174915443125e-05,
+      "loss": 0.7232,
+      "step": 5148
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24546455640964e-05,
+      "loss": 19.8944,
+      "step": 5149
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24475408415577e-05,
+      "loss": 22.153,
+      "step": 5150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244043498706299e-05,
+      "loss": 19.3735,
+      "step": 5151
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243332800086017e-05,
+      "loss": 0.7269,
+      "step": 5152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242621988319715e-05,
+      "loss": 21.4091,
+      "step": 5153
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241911063432193e-05,
+      "loss": 20.7133,
+      "step": 5154
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241200025448251e-05,
+      "loss": 18.4332,
+      "step": 5155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240488874392693e-05,
+      "loss": 22.7428,
+      "step": 5156
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239777610290332e-05,
+      "loss": 22.3612,
+      "step": 5157
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239066233165975e-05,
+      "loss": 0.6568,
+      "step": 5158
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.238354743044443e-05,
+      "loss": 21.7236,
+      "step": 5159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237643139950555e-05,
+      "loss": 22.3126,
+      "step": 5160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236931423909138e-05,
+      "loss": 0.7057,
+      "step": 5161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236219594945021e-05,
+      "loss": 19.3703,
+      "step": 5162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.235507653083032e-05,
+      "loss": 0.7346,
+      "step": 5163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234795598348011e-05,
+      "loss": 20.9397,
+      "step": 5164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234083430764798e-05,
+      "loss": 22.1243,
+      "step": 5165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23337115035824e-05,
+      "loss": 20.2409,
+      "step": 5166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232658757153182e-05,
+      "loss": 20.9243,
+      "step": 5167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.231946251174476e-05,
+      "loss": 18.9612,
+      "step": 5168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.231233632446981e-05,
+      "loss": 21.7137,
+      "step": 5169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.230520900995555e-05,
+      "loss": 19.3442,
+      "step": 5170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229808056845065e-05,
+      "loss": 20.2436,
+      "step": 5171
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229095100020377e-05,
+      "loss": 21.5926,
+      "step": 5172
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228382030546361e-05,
+      "loss": 0.7166,
+      "step": 5173
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.227668848447896e-05,
+      "loss": 0.6816,
+      "step": 5174
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226955553749862e-05,
+      "loss": 23.1958,
+      "step": 5175
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226242146477141e-05,
+      "loss": 0.7068,
+      "step": 5176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.225528626654623e-05,
+      "loss": 17.9756,
+      "step": 5177
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.224814994307197e-05,
+      "loss": 19.9473,
+      "step": 5178
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.224101249459759e-05,
+      "loss": 0.7553,
+      "step": 5179
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22338739213721e-05,
+      "loss": 0.7549,
+      "step": 5180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222673422364452e-05,
+      "loss": 21.1937,
+      "step": 5181
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221959340166395e-05,
+      "loss": 24.07,
+      "step": 5182
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221245145567949e-05,
+      "loss": 20.6039,
+      "step": 5183
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220530838594025e-05,
+      "loss": 20.7674,
+      "step": 5184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.219816419269548e-05,
+      "loss": 21.4268,
+      "step": 5185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.219101887619438e-05,
+      "loss": 20.1421,
+      "step": 5186
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218387243668623e-05,
+      "loss": 20.1734,
+      "step": 5187
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217672487442034e-05,
+      "loss": 17.7665,
+      "step": 5188
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.216957618964605e-05,
+      "loss": 18.6337,
+      "step": 5189
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.216242638261276e-05,
+      "loss": 20.6577,
+      "step": 5190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215527545356987e-05,
+      "loss": 0.7214,
+      "step": 5191
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.214812340276686e-05,
+      "loss": 0.7728,
+      "step": 5192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.214097023045325e-05,
+      "loss": 0.6825,
+      "step": 5193
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.213381593687856e-05,
+      "loss": 22.7628,
+      "step": 5194
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212666052229239e-05,
+      "loss": 22.0626,
+      "step": 5195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211950398694435e-05,
+      "loss": 20.4164,
+      "step": 5196
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211234633108411e-05,
+      "loss": 21.8705,
+      "step": 5197
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210518755496138e-05,
+      "loss": 0.6645,
+      "step": 5198
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209802765882586e-05,
+      "loss": 0.7284,
+      "step": 5199
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209086664292735e-05,
+      "loss": 22.5638,
+      "step": 5200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208370450751567e-05,
+      "loss": 20.9789,
+      "step": 5201
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20765412528407e-05,
+      "loss": 0.6752,
+      "step": 5202
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20693768791523e-05,
+      "loss": 23.4427,
+      "step": 5203
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.206221138670042e-05,
+      "loss": 17.4974,
+      "step": 5204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205504477573503e-05,
+      "loss": 21.7728,
+      "step": 5205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.204787704650614e-05,
+      "loss": 0.6908,
+      "step": 5206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20407081992638e-05,
+      "loss": 19.9522,
+      "step": 5207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203353823425809e-05,
+      "loss": 22.5766,
+      "step": 5208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.202636715173916e-05,
+      "loss": 19.9355,
+      "step": 5209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20191949519572e-05,
+      "loss": 0.7547,
+      "step": 5210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201202163516234e-05,
+      "loss": 21.4394,
+      "step": 5211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200484720160491e-05,
+      "loss": 21.3761,
+      "step": 5212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199767165153515e-05,
+      "loss": 20.6491,
+      "step": 5213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199049498520341e-05,
+      "loss": 0.7225,
+      "step": 5214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.198331720286004e-05,
+      "loss": 0.7029,
+      "step": 5215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.197613830475541e-05,
+      "loss": 0.6795,
+      "step": 5216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.196895829114e-05,
+      "loss": 0.6602,
+      "step": 5217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19617771622643e-05,
+      "loss": 17.6949,
+      "step": 5218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19545949183788e-05,
+      "loss": 21.4183,
+      "step": 5219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194741155973407e-05,
+      "loss": 22.3434,
+      "step": 5220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19402270865807e-05,
+      "loss": 22.3058,
+      "step": 5221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.193304149916934e-05,
+      "loss": 20.5979,
+      "step": 5222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.192585479775063e-05,
+      "loss": 18.7078,
+      "step": 5223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191866698257533e-05,
+      "loss": 20.0933,
+      "step": 5224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191147805389417e-05,
+      "loss": 20.5189,
+      "step": 5225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.190428801195795e-05,
+      "loss": 20.991,
+      "step": 5226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.189709685701747e-05,
+      "loss": 0.6898,
+      "step": 5227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.188990458932361e-05,
+      "loss": 21.4239,
+      "step": 5228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.188271120912732e-05,
+      "loss": 0.7188,
+      "step": 5229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18755167166795e-05,
+      "loss": 20.6909,
+      "step": 5230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186832111223115e-05,
+      "loss": 22.0971,
+      "step": 5231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186112439603329e-05,
+      "loss": 0.7556,
+      "step": 5232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1853926568337e-05,
+      "loss": 20.8236,
+      "step": 5233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.184672762939334e-05,
+      "loss": 20.3128,
+      "step": 5234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183952757945347e-05,
+      "loss": 0.8075,
+      "step": 5235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183232641876861e-05,
+      "loss": 0.7267,
+      "step": 5236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.182512414758992e-05,
+      "loss": 0.659,
+      "step": 5237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18179207661687e-05,
+      "loss": 22.7281,
+      "step": 5238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18107162747562e-05,
+      "loss": 0.6738,
+      "step": 5239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.180351067360379e-05,
+      "loss": 19.4316,
+      "step": 5240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179630396296283e-05,
+      "loss": 21.8338,
+      "step": 5241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178909614308474e-05,
+      "loss": 19.7691,
+      "step": 5242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178188721422096e-05,
+      "loss": 0.7493,
+      "step": 5243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177467717662298e-05,
+      "loss": 21.66,
+      "step": 5244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176746603054233e-05,
+      "loss": 21.0252,
+      "step": 5245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176025377623058e-05,
+      "loss": 18.674,
+      "step": 5246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175304041393931e-05,
+      "loss": 22.171,
+      "step": 5247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.17458259439202e-05,
+      "loss": 19.1039,
+      "step": 5248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173861036642492e-05,
+      "loss": 0.6872,
+      "step": 5249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173139368170518e-05,
+      "loss": 0.7572,
+      "step": 5250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.172417589001274e-05,
+      "loss": 0.7415,
+      "step": 5251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.17169569915994e-05,
+      "loss": 0.701,
+      "step": 5252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170973698671703e-05,
+      "loss": 0.7589,
+      "step": 5253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170251587561744e-05,
+      "loss": 20.3896,
+      "step": 5254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169529365855259e-05,
+      "loss": 0.6659,
+      "step": 5255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168807033577442e-05,
+      "loss": 0.6857,
+      "step": 5256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168084590753491e-05,
+      "loss": 22.5438,
+      "step": 5257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167362037408613e-05,
+      "loss": 0.717,
+      "step": 5258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16663937356801e-05,
+      "loss": 20.7593,
+      "step": 5259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.165916599256895e-05,
+      "loss": 0.6846,
+      "step": 5260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.165193714500481e-05,
+      "loss": 19.883,
+      "step": 5261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164470719323989e-05,
+      "loss": 21.4824,
+      "step": 5262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163747613752638e-05,
+      "loss": 19.8691,
+      "step": 5263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163024397811656e-05,
+      "loss": 0.7124,
+      "step": 5264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.162301071526274e-05,
+      "loss": 0.7359,
+      "step": 5265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161577634921724e-05,
+      "loss": 21.4825,
+      "step": 5266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160854088023242e-05,
+      "loss": 0.7097,
+      "step": 5267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160130430856074e-05,
+      "loss": 21.3458,
+      "step": 5268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.159406663445462e-05,
+      "loss": 20.3561,
+      "step": 5269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.158682785816654e-05,
+      "loss": 0.7709,
+      "step": 5270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157958797994907e-05,
+      "loss": 21.1153,
+      "step": 5271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157234700005475e-05,
+      "loss": 0.7386,
+      "step": 5272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156510491873618e-05,
+      "loss": 19.3621,
+      "step": 5273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155786173624605e-05,
+      "loss": 0.7258,
+      "step": 5274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155061745283699e-05,
+      "loss": 19.257,
+      "step": 5275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154337206876176e-05,
+      "loss": 0.7818,
+      "step": 5276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153612558427311e-05,
+      "loss": 19.7243,
+      "step": 5277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15288779996238e-05,
+      "loss": 0.7233,
+      "step": 5278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152162931506673e-05,
+      "loss": 0.6865,
+      "step": 5279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151437953085475e-05,
+      "loss": 0.6799,
+      "step": 5280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.150712864724075e-05,
+      "loss": 0.7036,
+      "step": 5281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149987666447772e-05,
+      "loss": 20.009,
+      "step": 5282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149262358281864e-05,
+      "loss": 21.1809,
+      "step": 5283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.148536940251652e-05,
+      "loss": 19.0115,
+      "step": 5284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.147811412382444e-05,
+      "loss": 20.6396,
+      "step": 5285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.147085774699552e-05,
+      "loss": 21.0696,
+      "step": 5286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.146360027228288e-05,
+      "loss": 19.0804,
+      "step": 5287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145634169993972e-05,
+      "loss": 0.7472,
+      "step": 5288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.144908203021925e-05,
+      "loss": 0.6915,
+      "step": 5289
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.144182126337472e-05,
+      "loss": 22.5543,
+      "step": 5290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143455939965945e-05,
+      "loss": 20.3162,
+      "step": 5291
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.142729643932676e-05,
+      "loss": 20.2511,
+      "step": 5292
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.142003238263003e-05,
+      "loss": 0.7166,
+      "step": 5293
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141276722982267e-05,
+      "loss": 22.765,
+      "step": 5294
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140550098115814e-05,
+      "loss": 20.5243,
+      "step": 5295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139823363688991e-05,
+      "loss": 19.2257,
+      "step": 5296
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139096519727152e-05,
+      "loss": 0.716,
+      "step": 5297
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138369566255652e-05,
+      "loss": 19.5304,
+      "step": 5298
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137642503299854e-05,
+      "loss": 0.7138,
+      "step": 5299
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136915330885118e-05,
+      "loss": 18.2003,
+      "step": 5300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136188049036815e-05,
+      "loss": 20.8325,
+      "step": 5301
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135460657780317e-05,
+      "loss": 0.7147,
+      "step": 5302
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134733157141e-05,
+      "loss": 18.6782,
+      "step": 5303
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134005547144238e-05,
+      "loss": 19.569,
+      "step": 5304
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133277827815422e-05,
+      "loss": 0.6901,
+      "step": 5305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132549999179933e-05,
+      "loss": 0.7414,
+      "step": 5306
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131822061263166e-05,
+      "loss": 20.9517,
+      "step": 5307
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131094014090513e-05,
+      "loss": 0.783,
+      "step": 5308
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130365857687372e-05,
+      "loss": 0.7357,
+      "step": 5309
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129637592079147e-05,
+      "loss": 0.7139,
+      "step": 5310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128909217291245e-05,
+      "loss": 20.5255,
+      "step": 5311
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128180733349073e-05,
+      "loss": 0.6664,
+      "step": 5312
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127452140278047e-05,
+      "loss": 0.6863,
+      "step": 5313
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.126723438103583e-05,
+      "loss": 20.6331,
+      "step": 5314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125994626851103e-05,
+      "loss": 20.7101,
+      "step": 5315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125265706546032e-05,
+      "loss": 0.738,
+      "step": 5316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.1245366772138e-05,
+      "loss": 20.3506,
+      "step": 5317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123807538879838e-05,
+      "loss": 0.6912,
+      "step": 5318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123078291569582e-05,
+      "loss": 20.6653,
+      "step": 5319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122348935308475e-05,
+      "loss": 0.65,
+      "step": 5320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12161947012196e-05,
+      "loss": 0.6576,
+      "step": 5321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120889896035483e-05,
+      "loss": 21.1778,
+      "step": 5322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120160213074497e-05,
+      "loss": 19.4245,
+      "step": 5323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.119430421264459e-05,
+      "loss": 21.2838,
+      "step": 5324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.118700520630826e-05,
+      "loss": 20.6704,
+      "step": 5325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117970511199065e-05,
+      "loss": 21.6346,
+      "step": 5326
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117240392994638e-05,
+      "loss": 20.4237,
+      "step": 5327
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116510166043017e-05,
+      "loss": 19.1136,
+      "step": 5328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115779830369678e-05,
+      "loss": 0.7394,
+      "step": 5329
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.1150493860001e-05,
+      "loss": 0.7716,
+      "step": 5330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.114318832959762e-05,
+      "loss": 0.739,
+      "step": 5331
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113588171274152e-05,
+      "loss": 0.7063,
+      "step": 5332
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112857400968759e-05,
+      "loss": 0.7722,
+      "step": 5333
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112126522069077e-05,
+      "loss": 21.1625,
+      "step": 5334
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.111395534600603e-05,
+      "loss": 20.6395,
+      "step": 5335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110664438588839e-05,
+      "loss": 0.7116,
+      "step": 5336
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109933234059288e-05,
+      "loss": 20.8521,
+      "step": 5337
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109201921037459e-05,
+      "loss": 0.7299,
+      "step": 5338
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.108470499548863e-05,
+      "loss": 0.7268,
+      "step": 5339
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10773896961902e-05,
+      "loss": 18.6439,
+      "step": 5340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.107007331273449e-05,
+      "loss": 19.518,
+      "step": 5341
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106275584537673e-05,
+      "loss": 0.6986,
+      "step": 5342
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105543729437216e-05,
+      "loss": 0.7127,
+      "step": 5343
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.104811765997614e-05,
+      "loss": 20.2981,
+      "step": 5344
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.1040796942444e-05,
+      "loss": 0.7282,
+      "step": 5345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103347514203114e-05,
+      "loss": 22.676,
+      "step": 5346
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.102615225899298e-05,
+      "loss": 19.5431,
+      "step": 5347
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101882829358498e-05,
+      "loss": 0.6834,
+      "step": 5348
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101150324606265e-05,
+      "loss": 21.766,
+      "step": 5349
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.100417711668151e-05,
+      "loss": 19.6516,
+      "step": 5350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099684990569717e-05,
+      "loss": 0.7572,
+      "step": 5351
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098952161336523e-05,
+      "loss": 22.5784,
+      "step": 5352
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098219223994132e-05,
+      "loss": 18.6344,
+      "step": 5353
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.097486178568116e-05,
+      "loss": 20.1082,
+      "step": 5354
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09675302508405e-05,
+      "loss": 0.6938,
+      "step": 5355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.096019763567503e-05,
+      "loss": 18.4439,
+      "step": 5356
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095286394044062e-05,
+      "loss": 20.3329,
+      "step": 5357
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09455291653931e-05,
+      "loss": 19.7186,
+      "step": 5358
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093819331078832e-05,
+      "loss": 20.0357,
+      "step": 5359
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093085637688222e-05,
+      "loss": 0.6755,
+      "step": 5360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.092351836393076e-05,
+      "loss": 0.6589,
+      "step": 5361
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091617927218992e-05,
+      "loss": 0.6894,
+      "step": 5362
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090883910191572e-05,
+      "loss": 0.7519,
+      "step": 5363
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090149785336425e-05,
+      "loss": 0.7117,
+      "step": 5364
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089415552679159e-05,
+      "loss": 19.8539,
+      "step": 5365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08868121224539e-05,
+      "loss": 20.0559,
+      "step": 5366
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.087946764060736e-05,
+      "loss": 0.7607,
+      "step": 5367
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08721220815082e-05,
+      "loss": 21.5084,
+      "step": 5368
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086477544541264e-05,
+      "loss": 18.9915,
+      "step": 5369
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0857427732577e-05,
+      "loss": 22.4022,
+      "step": 5370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085007894325758e-05,
+      "loss": 21.5201,
+      "step": 5371
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08427290777108e-05,
+      "loss": 19.636,
+      "step": 5372
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.083537813619302e-05,
+      "loss": 22.0096,
+      "step": 5373
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082802611896069e-05,
+      "loss": 18.6611,
+      "step": 5374
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08206730262703e-05,
+      "loss": 18.9061,
+      "step": 5375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081331885837837e-05,
+      "loss": 0.6993,
+      "step": 5376
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080596361554146e-05,
+      "loss": 0.6962,
+      "step": 5377
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079860729801615e-05,
+      "loss": 0.6843,
+      "step": 5378
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079124990605907e-05,
+      "loss": 21.2788,
+      "step": 5379
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07838914399269e-05,
+      "loss": 18.3963,
+      "step": 5380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.077653189987633e-05,
+      "loss": 0.7468,
+      "step": 5381
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07691712861641e-05,
+      "loss": 0.6975,
+      "step": 5382
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076180959904702e-05,
+      "loss": 20.3337,
+      "step": 5383
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07544468387819e-05,
+      "loss": 19.9629,
+      "step": 5384
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074708300562557e-05,
+      "loss": 0.8069,
+      "step": 5385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.073971809983494e-05,
+      "loss": 20.3821,
+      "step": 5386
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.073235212166692e-05,
+      "loss": 20.9017,
+      "step": 5387
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072498507137852e-05,
+      "loss": 20.4007,
+      "step": 5388
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.071761694922671e-05,
+      "loss": 17.2817,
+      "step": 5389
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.071024775546854e-05,
+      "loss": 21.4581,
+      "step": 5390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07028774903611e-05,
+      "loss": 0.6753,
+      "step": 5391
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06955061541615e-05,
+      "loss": 0.6528,
+      "step": 5392
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.068813374712688e-05,
+      "loss": 0.7545,
+      "step": 5393
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.068076026951445e-05,
+      "loss": 21.3264,
+      "step": 5394
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067338572158146e-05,
+      "loss": 20.7566,
+      "step": 5395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066601010358511e-05,
+      "loss": 18.8798,
+      "step": 5396
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.065863341578277e-05,
+      "loss": 22.2198,
+      "step": 5397
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.065125565843175e-05,
+      "loss": 0.6698,
+      "step": 5398
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.064387683178942e-05,
+      "loss": 20.7222,
+      "step": 5399
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063649693611321e-05,
+      "loss": 20.711,
+      "step": 5400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06291159716606e-05,
+      "loss": 0.6926,
+      "step": 5401
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062173393868904e-05,
+      "loss": 19.7748,
+      "step": 5402
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061435083745605e-05,
+      "loss": 19.7432,
+      "step": 5403
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060696666821923e-05,
+      "loss": 20.213,
+      "step": 5404
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059958143123618e-05,
+      "loss": 22.1219,
+      "step": 5405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059219512676451e-05,
+      "loss": 18.9847,
+      "step": 5406
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058480775506192e-05,
+      "loss": 0.7238,
+      "step": 5407
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057741931638612e-05,
+      "loss": 19.2047,
+      "step": 5408
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057002981099486e-05,
+      "loss": 18.8977,
+      "step": 5409
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.056263923914592e-05,
+      "loss": 22.0461,
+      "step": 5410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055524760109714e-05,
+      "loss": 20.9519,
+      "step": 5411
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054785489710637e-05,
+      "loss": 19.669,
+      "step": 5412
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05404611274315e-05,
+      "loss": 20.3665,
+      "step": 5413
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05330662923305e-05,
+      "loss": 19.8052,
+      "step": 5414
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.052567039206131e-05,
+      "loss": 19.3227,
+      "step": 5415
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.051827342688198e-05,
+      "loss": 0.7553,
+      "step": 5416
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.051087539705052e-05,
+      "loss": 20.2252,
+      "step": 5417
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050347630282503e-05,
+      "loss": 0.7021,
+      "step": 5418
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049607614446364e-05,
+      "loss": 0.7494,
+      "step": 5419
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04886749222245e-05,
+      "loss": 0.6842,
+      "step": 5420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048127263636583e-05,
+      "loss": 0.7564,
+      "step": 5421
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.047386928714582e-05,
+      "loss": 20.3697,
+      "step": 5422
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.046646487482277e-05,
+      "loss": 18.4348,
+      "step": 5423
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.045905939965499e-05,
+      "loss": 0.7059,
+      "step": 5424
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.045165286190082e-05,
+      "loss": 20.2356,
+      "step": 5425
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044424526181866e-05,
+      "loss": 0.7609,
+      "step": 5426
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043683659966688e-05,
+      "loss": 21.4688,
+      "step": 5427
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042942687570399e-05,
+      "loss": 19.6463,
+      "step": 5428
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042201609018845e-05,
+      "loss": 19.8885,
+      "step": 5429
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04146042433788e-05,
+      "loss": 0.6916,
+      "step": 5430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.040719133553363e-05,
+      "loss": 18.1718,
+      "step": 5431
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03997773669115e-05,
+      "loss": 20.7918,
+      "step": 5432
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.039236233777109e-05,
+      "loss": 18.8004,
+      "step": 5433
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038494624837106e-05,
+      "loss": 19.1889,
+      "step": 5434
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037752909897014e-05,
+      "loss": 0.7077,
+      "step": 5435
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037011088982707e-05,
+      "loss": 20.6571,
+      "step": 5436
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036269162120063e-05,
+      "loss": 18.8745,
+      "step": 5437
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035527129334968e-05,
+      "loss": 0.715,
+      "step": 5438
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034784990653306e-05,
+      "loss": 23.3977,
+      "step": 5439
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034042746100964e-05,
+      "loss": 0.7609,
+      "step": 5440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.033300395703845e-05,
+      "loss": 21.1029,
+      "step": 5441
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.032557939487838e-05,
+      "loss": 0.7372,
+      "step": 5442
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.031815377478843e-05,
+      "loss": 22.5266,
+      "step": 5443
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.031072709702772e-05,
+      "loss": 17.9623,
+      "step": 5444
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03032993618553e-05,
+      "loss": 18.2795,
+      "step": 5445
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02958705695303e-05,
+      "loss": 19.7006,
+      "step": 5446
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028844072031185e-05,
+      "loss": 0.7153,
+      "step": 5447
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028100981445918e-05,
+      "loss": 23.7709,
+      "step": 5448
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027357785223151e-05,
+      "loss": 0.6921,
+      "step": 5449
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.026614483388811e-05,
+      "loss": 19.8696,
+      "step": 5450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.025871075968828e-05,
+      "loss": 21.489,
+      "step": 5451
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.025127562989135e-05,
+      "loss": 0.7032,
+      "step": 5452
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024383944475673e-05,
+      "loss": 19.2567,
+      "step": 5453
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023640220454382e-05,
+      "loss": 0.7698,
+      "step": 5454
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02289639095121e-05,
+      "loss": 21.0913,
+      "step": 5455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022152455992102e-05,
+      "loss": 19.8329,
+      "step": 5456
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021408415603011e-05,
+      "loss": 20.2448,
+      "step": 5457
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.020664269809898e-05,
+      "loss": 18.1739,
+      "step": 5458
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019920018638718e-05,
+      "loss": 0.73,
+      "step": 5459
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019175662115437e-05,
+      "loss": 19.9554,
+      "step": 5460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018431200266023e-05,
+      "loss": 20.3966,
+      "step": 5461
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.017686633116445e-05,
+      "loss": 20.7335,
+      "step": 5462
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016941960692681e-05,
+      "loss": 0.7327,
+      "step": 5463
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016197183020705e-05,
+      "loss": 0.7349,
+      "step": 5464
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.015452300126502e-05,
+      "loss": 22.1293,
+      "step": 5465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014707312036059e-05,
+      "loss": 0.7265,
+      "step": 5466
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013962218775363e-05,
+      "loss": 19.1339,
+      "step": 5467
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013217020370407e-05,
+      "loss": 0.7678,
+      "step": 5468
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01247171684719e-05,
+      "loss": 0.6819,
+      "step": 5469
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01172630823171e-05,
+      "loss": 21.4633,
+      "step": 5470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010980794549972e-05,
+      "loss": 20.9375,
+      "step": 5471
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010235175827985e-05,
+      "loss": 21.7742,
+      "step": 5472
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.009489452091758e-05,
+      "loss": 0.6906,
+      "step": 5473
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00874362336731e-05,
+      "loss": 23.0092,
+      "step": 5474
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007997689680655e-05,
+      "loss": 22.2576,
+      "step": 5475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007251651057819e-05,
+      "loss": 0.7456,
+      "step": 5476
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006505507524829e-05,
+      "loss": 19.6351,
+      "step": 5477
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.005759259107709e-05,
+      "loss": 0.6826,
+      "step": 5478
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.005012905832499e-05,
+      "loss": 0.6944,
+      "step": 5479
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00426644772523e-05,
+      "loss": 0.6992,
+      "step": 5480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.003519884811948e-05,
+      "loss": 25.5689,
+      "step": 5481
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002773217118696e-05,
+      "loss": 20.4312,
+      "step": 5482
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00202644467152e-05,
+      "loss": 0.7278,
+      "step": 5483
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001279567496474e-05,
+      "loss": 19.6346,
+      "step": 5484
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00053258561961e-05,
+      "loss": 19.063,
+      "step": 5485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999785499066992e-05,
+      "loss": 19.8757,
+      "step": 5486
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999038307864678e-05,
+      "loss": 20.6311,
+      "step": 5487
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.998291012038736e-05,
+      "loss": 0.7202,
+      "step": 5488
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99754361161524e-05,
+      "loss": 19.9091,
+      "step": 5489
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996796106620257e-05,
+      "loss": 0.7342,
+      "step": 5490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996048497079868e-05,
+      "loss": 0.7208,
+      "step": 5491
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995300783020152e-05,
+      "loss": 0.654,
+      "step": 5492
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.994552964467196e-05,
+      "loss": 0.7106,
+      "step": 5493
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993805041447086e-05,
+      "loss": 22.2792,
+      "step": 5494
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993057013985915e-05,
+      "loss": 21.0153,
+      "step": 5495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.992308882109778e-05,
+      "loss": 19.7918,
+      "step": 5496
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991560645844773e-05,
+      "loss": 18.2614,
+      "step": 5497
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990812305217006e-05,
+      "loss": 0.7125,
+      "step": 5498
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990063860252582e-05,
+      "loss": 20.2752,
+      "step": 5499
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.989315310977608e-05,
+      "loss": 19.7388,
+      "step": 5500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988566657418201e-05,
+      "loss": 0.686,
+      "step": 5501
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987817899600478e-05,
+      "loss": 0.7271,
+      "step": 5502
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987069037550559e-05,
+      "loss": 20.3488,
+      "step": 5503
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.98632007129457e-05,
+      "loss": 21.0729,
+      "step": 5504
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985571000858638e-05,
+      "loss": 20.8775,
+      "step": 5505
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984821826268894e-05,
+      "loss": 20.204,
+      "step": 5506
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984072547551475e-05,
+      "loss": 21.3468,
+      "step": 5507
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.983323164732522e-05,
+      "loss": 0.7714,
+      "step": 5508
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982573677838172e-05,
+      "loss": 20.7437,
+      "step": 5509
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981824086894576e-05,
+      "loss": 22.6467,
+      "step": 5510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981074391927884e-05,
+      "loss": 21.5693,
+      "step": 5511
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980324592964248e-05,
+      "loss": 0.7042,
+      "step": 5512
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.979574690029825e-05,
+      "loss": 20.1327,
+      "step": 5513
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978824683150779e-05,
+      "loss": 0.6858,
+      "step": 5514
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978074572353269e-05,
+      "loss": 17.6504,
+      "step": 5515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.977324357663469e-05,
+      "loss": 0.7643,
+      "step": 5516
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976574039107547e-05,
+      "loss": 21.2166,
+      "step": 5517
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975823616711682e-05,
+      "loss": 20.0327,
+      "step": 5518
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975073090502048e-05,
+      "loss": 0.7165,
+      "step": 5519
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.974322460504832e-05,
+      "loss": 0.7137,
+      "step": 5520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.973571726746218e-05,
+      "loss": 20.829,
+      "step": 5521
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972820889252396e-05,
+      "loss": 18.1426,
+      "step": 5522
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972069948049561e-05,
+      "loss": 0.6782,
+      "step": 5523
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971318903163908e-05,
+      "loss": 20.5434,
+      "step": 5524
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97056775462164e-05,
+      "loss": 20.2555,
+      "step": 5525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969816502448959e-05,
+      "loss": 19.5886,
+      "step": 5526
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969065146672074e-05,
+      "loss": 20.7231,
+      "step": 5527
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.968313687317199e-05,
+      "loss": 19.9343,
+      "step": 5528
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967562124410546e-05,
+      "loss": 21.7896,
+      "step": 5529
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966810457978335e-05,
+      "loss": 0.6921,
+      "step": 5530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966058688046787e-05,
+      "loss": 20.8747,
+      "step": 5531
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965306814642132e-05,
+      "loss": 0.7312,
+      "step": 5532
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964554837790594e-05,
+      "loss": 19.5068,
+      "step": 5533
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96380275751841e-05,
+      "loss": 19.92,
+      "step": 5534
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963050573851818e-05,
+      "loss": 17.9062,
+      "step": 5535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962298286817055e-05,
+      "loss": 20.9306,
+      "step": 5536
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961545896440366e-05,
+      "loss": 0.8041,
+      "step": 5537
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960793402748002e-05,
+      "loss": 18.422,
+      "step": 5538
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960040805766209e-05,
+      "loss": 0.7149,
+      "step": 5539
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959288105521245e-05,
+      "loss": 0.7028,
+      "step": 5540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.958535302039367e-05,
+      "loss": 0.7436,
+      "step": 5541
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95778239534684e-05,
+      "loss": 21.1758,
+      "step": 5542
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.957029385469928e-05,
+      "loss": 19.025,
+      "step": 5543
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956276272434898e-05,
+      "loss": 21.2403,
+      "step": 5544
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955523056268026e-05,
+      "loss": 0.7583,
+      "step": 5545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.954769736995589e-05,
+      "loss": 20.7059,
+      "step": 5546
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.954016314643863e-05,
+      "loss": 20.0116,
+      "step": 5547
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.953262789239136e-05,
+      "loss": 18.4587,
+      "step": 5548
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.952509160807693e-05,
+      "loss": 20.1882,
+      "step": 5549
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951755429375825e-05,
+      "loss": 0.7114,
+      "step": 5550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951001594969827e-05,
+      "loss": 20.593,
+      "step": 5551
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.950247657615998e-05,
+      "loss": 0.7006,
+      "step": 5552
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.949493617340637e-05,
+      "loss": 0.7238,
+      "step": 5553
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948739474170053e-05,
+      "loss": 0.7086,
+      "step": 5554
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.94798522813055e-05,
+      "loss": 0.7424,
+      "step": 5555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.947230879248444e-05,
+      "loss": 0.759,
+      "step": 5556
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946476427550051e-05,
+      "loss": 22.1223,
+      "step": 5557
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945721873061689e-05,
+      "loss": 20.8681,
+      "step": 5558
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944967215809683e-05,
+      "loss": 20.0851,
+      "step": 5559
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944212455820358e-05,
+      "loss": 20.7828,
+      "step": 5560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943457593120045e-05,
+      "loss": 19.4776,
+      "step": 5561
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942702627735078e-05,
+      "loss": 21.7625,
+      "step": 5562
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941947559691792e-05,
+      "loss": 22.7659,
+      "step": 5563
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941192389016534e-05,
+      "loss": 19.7016,
+      "step": 5564
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.940437115735642e-05,
+      "loss": 0.6791,
+      "step": 5565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93968173987547e-05,
+      "loss": 18.5158,
+      "step": 5566
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938926261462366e-05,
+      "loss": 0.6911,
+      "step": 5567
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938170680522687e-05,
+      "loss": 19.8251,
+      "step": 5568
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.937414997082793e-05,
+      "loss": 0.6997,
+      "step": 5569
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.936659211169045e-05,
+      "loss": 19.1256,
+      "step": 5570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.935903322807807e-05,
+      "loss": 18.7505,
+      "step": 5571
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.935147332025452e-05,
+      "loss": 0.6534,
+      "step": 5572
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.934391238848353e-05,
+      "loss": 21.3458,
+      "step": 5573
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933635043302887e-05,
+      "loss": 19.5601,
+      "step": 5574
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932878745415433e-05,
+      "loss": 17.7489,
+      "step": 5575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932122345212376e-05,
+      "loss": 23.1725,
+      "step": 5576
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931365842720102e-05,
+      "loss": 17.8956,
+      "step": 5577
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.930609237965005e-05,
+      "loss": 19.3752,
+      "step": 5578
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929852530973476e-05,
+      "loss": 0.6999,
+      "step": 5579
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929095721771917e-05,
+      "loss": 0.7106,
+      "step": 5580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928338810386729e-05,
+      "loss": 18.6161,
+      "step": 5581
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927581796844314e-05,
+      "loss": 18.5225,
+      "step": 5582
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.926824681171087e-05,
+      "loss": 21.2515,
+      "step": 5583
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.926067463393454e-05,
+      "loss": 18.9826,
+      "step": 5584
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925310143537836e-05,
+      "loss": 0.7324,
+      "step": 5585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92455272163065e-05,
+      "loss": 20.3184,
+      "step": 5586
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923795197698322e-05,
+      "loss": 18.0527,
+      "step": 5587
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923037571767275e-05,
+      "loss": 20.4791,
+      "step": 5588
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.922279843863942e-05,
+      "loss": 0.727,
+      "step": 5589
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.921522014014755e-05,
+      "loss": 20.1647,
+      "step": 5590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920764082246153e-05,
+      "loss": 20.5145,
+      "step": 5591
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920006048584576e-05,
+      "loss": 20.1014,
+      "step": 5592
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91924791305647e-05,
+      "loss": 18.7933,
+      "step": 5593
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918489675688281e-05,
+      "loss": 19.7465,
+      "step": 5594
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.917731336506464e-05,
+      "loss": 0.7154,
+      "step": 5595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916972895537471e-05,
+      "loss": 17.6426,
+      "step": 5596
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91621435280776e-05,
+      "loss": 0.749,
+      "step": 5597
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.915455708343797e-05,
+      "loss": 0.7367,
+      "step": 5598
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.914696962172046e-05,
+      "loss": 21.0745,
+      "step": 5599
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.913938114318977e-05,
+      "loss": 0.7461,
+      "step": 5600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91317916481106e-05,
+      "loss": 20.6916,
+      "step": 5601
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.912420113674776e-05,
+      "loss": 20.882,
+      "step": 5602
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911660960936602e-05,
+      "loss": 20.1276,
+      "step": 5603
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910901706623022e-05,
+      "loss": 0.7575,
+      "step": 5604
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910142350760525e-05,
+      "loss": 18.7341,
+      "step": 5605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909382893375598e-05,
+      "loss": 0.6923,
+      "step": 5606
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90862333449474e-05,
+      "loss": 21.9817,
+      "step": 5607
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907863674144445e-05,
+      "loss": 22.161,
+      "step": 5608
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907103912351217e-05,
+      "loss": 0.7299,
+      "step": 5609
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.906344049141559e-05,
+      "loss": 18.6946,
+      "step": 5610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905584084541978e-05,
+      "loss": 0.7037,
+      "step": 5611
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90482401857899e-05,
+      "loss": 19.6652,
+      "step": 5612
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.904063851279106e-05,
+      "loss": 17.802,
+      "step": 5613
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.903303582668849e-05,
+      "loss": 19.7793,
+      "step": 5614
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902543212774739e-05,
+      "loss": 19.7927,
+      "step": 5615
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901782741623303e-05,
+      "loss": 20.0378,
+      "step": 5616
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90102216924107e-05,
+      "loss": 20.2555,
+      "step": 5617
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900261495654574e-05,
+      "loss": 20.3204,
+      "step": 5618
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899500720890352e-05,
+      "loss": 21.1234,
+      "step": 5619
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89873984497494e-05,
+      "loss": 0.7454,
+      "step": 5620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897978867934887e-05,
+      "loss": 0.7548,
+      "step": 5621
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89721778979674e-05,
+      "loss": 19.0701,
+      "step": 5622
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.896456610587046e-05,
+      "loss": 19.9906,
+      "step": 5623
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89569533033236e-05,
+      "loss": 21.5347,
+      "step": 5624
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894933949059245e-05,
+      "loss": 0.7462,
+      "step": 5625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894172466794256e-05,
+      "loss": 18.0413,
+      "step": 5626
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89341088356396e-05,
+      "loss": 20.6616,
+      "step": 5627
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.892649199394925e-05,
+      "loss": 0.7283,
+      "step": 5628
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.891887414313726e-05,
+      "loss": 20.4837,
+      "step": 5629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.891125528346935e-05,
+      "loss": 17.8926,
+      "step": 5630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89036354152113e-05,
+      "loss": 0.6763,
+      "step": 5631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.889601453862898e-05,
+      "loss": 19.9179,
+      "step": 5632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888839265398822e-05,
+      "loss": 20.1223,
+      "step": 5633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888076976155492e-05,
+      "loss": 18.9808,
+      "step": 5634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.8873145861595e-05,
+      "loss": 18.8288,
+      "step": 5635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.886552095437444e-05,
+      "loss": 18.8344,
+      "step": 5636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.885789504015924e-05,
+      "loss": 17.2854,
+      "step": 5637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.885026811921541e-05,
+      "loss": 0.7016,
+      "step": 5638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884264019180907e-05,
+      "loss": 0.6889,
+      "step": 5639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.883501125820627e-05,
+      "loss": 18.7585,
+      "step": 5640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88273813186732e-05,
+      "loss": 18.6936,
+      "step": 5641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8819750373476e-05,
+      "loss": 19.4034,
+      "step": 5642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.881211842288091e-05,
+      "loss": 21.6408,
+      "step": 5643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880448546715416e-05,
+      "loss": 19.2502,
+      "step": 5644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8796851506562e-05,
+      "loss": 20.3843,
+      "step": 5645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.878921654137082e-05,
+      "loss": 20.4958,
+      "step": 5646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.878158057184691e-05,
+      "loss": 20.8433,
+      "step": 5647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877394359825667e-05,
+      "loss": 19.4216,
+      "step": 5648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876630562086653e-05,
+      "loss": 19.7448,
+      "step": 5649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875866663994295e-05,
+      "loss": 21.9349,
+      "step": 5650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87510266557524e-05,
+      "loss": 19.4603,
+      "step": 5651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874338566856145e-05,
+      "loss": 0.7038,
+      "step": 5652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873574367863661e-05,
+      "loss": 21.0836,
+      "step": 5653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872810068624451e-05,
+      "loss": 19.0767,
+      "step": 5654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872045669165177e-05,
+      "loss": 18.8714,
+      "step": 5655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.871281169512506e-05,
+      "loss": 21.8754,
+      "step": 5656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870516569693106e-05,
+      "loss": 20.7671,
+      "step": 5657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869751869733654e-05,
+      "loss": 19.3921,
+      "step": 5658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.868987069660824e-05,
+      "loss": 21.1001,
+      "step": 5659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8682221695013e-05,
+      "loss": 19.8389,
+      "step": 5660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.867457169281765e-05,
+      "loss": 0.7413,
+      "step": 5661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866692069028904e-05,
+      "loss": 20.7364,
+      "step": 5662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86592686876941e-05,
+      "loss": 0.6447,
+      "step": 5663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86516156852998e-05,
+      "loss": 21.2267,
+      "step": 5664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.864396168337307e-05,
+      "loss": 0.7061,
+      "step": 5665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.863630668218095e-05,
+      "loss": 0.7156,
+      "step": 5666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86286506819905e-05,
+      "loss": 20.618,
+      "step": 5667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862099368306879e-05,
+      "loss": 20.3418,
+      "step": 5668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.861333568568296e-05,
+      "loss": 0.6499,
+      "step": 5669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.860567669010014e-05,
+      "loss": 20.8281,
+      "step": 5670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859801669658753e-05,
+      "loss": 20.6848,
+      "step": 5671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859035570541235e-05,
+      "loss": 21.3981,
+      "step": 5672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858269371684188e-05,
+      "loss": 20.6127,
+      "step": 5673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857503073114338e-05,
+      "loss": 20.7301,
+      "step": 5674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856736674858421e-05,
+      "loss": 18.1101,
+      "step": 5675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855970176943173e-05,
+      "loss": 19.9607,
+      "step": 5676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855203579395331e-05,
+      "loss": 0.719,
+      "step": 5677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.85443688224164e-05,
+      "loss": 0.7543,
+      "step": 5678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.853670085508849e-05,
+      "loss": 0.7066,
+      "step": 5679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852903189223705e-05,
+      "loss": 0.7111,
+      "step": 5680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852136193412963e-05,
+      "loss": 20.9199,
+      "step": 5681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.851369098103381e-05,
+      "loss": 0.7033,
+      "step": 5682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850601903321716e-05,
+      "loss": 19.6185,
+      "step": 5683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84983460909474e-05,
+      "loss": 0.6889,
+      "step": 5684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.849067215449212e-05,
+      "loss": 0.7665,
+      "step": 5685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.848299722411908e-05,
+      "loss": 19.503,
+      "step": 5686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.847532130009601e-05,
+      "loss": 18.8875,
+      "step": 5687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846764438269067e-05,
+      "loss": 18.6861,
+      "step": 5688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845996647217093e-05,
+      "loss": 0.7699,
+      "step": 5689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845228756880461e-05,
+      "loss": 20.1171,
+      "step": 5690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.844460767285956e-05,
+      "loss": 21.2773,
+      "step": 5691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.843692678460377e-05,
+      "loss": 20.2226,
+      "step": 5692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842924490430512e-05,
+      "loss": 19.4493,
+      "step": 5693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842156203223165e-05,
+      "loss": 19.0427,
+      "step": 5694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841387816865136e-05,
+      "loss": 22.1278,
+      "step": 5695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840619331383232e-05,
+      "loss": 0.6864,
+      "step": 5696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839850746804261e-05,
+      "loss": 20.1464,
+      "step": 5697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839082063155034e-05,
+      "loss": 19.1595,
+      "step": 5698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83831328046237e-05,
+      "loss": 0.6777,
+      "step": 5699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.837544398753087e-05,
+      "loss": 20.0275,
+      "step": 5700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83677541805401e-05,
+      "loss": 21.8623,
+      "step": 5701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836006338391963e-05,
+      "loss": 0.7228,
+      "step": 5702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.835237159793778e-05,
+      "loss": 0.7189,
+      "step": 5703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.834467882286286e-05,
+      "loss": 21.3146,
+      "step": 5704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.833698505896325e-05,
+      "loss": 0.6813,
+      "step": 5705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.832929030650738e-05,
+      "loss": 0.7104,
+      "step": 5706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.832159456576362e-05,
+      "loss": 21.3695,
+      "step": 5707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.831389783700051e-05,
+      "loss": 0.7352,
+      "step": 5708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830620012048651e-05,
+      "loss": 19.7544,
+      "step": 5709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82985014164902e-05,
+      "loss": 18.7949,
+      "step": 5710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829080172528011e-05,
+      "loss": 20.1013,
+      "step": 5711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828310104712489e-05,
+      "loss": 0.7298,
+      "step": 5712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.827539938229315e-05,
+      "loss": 22.4856,
+      "step": 5713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82676967310536e-05,
+      "loss": 17.0258,
+      "step": 5714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.825999309367493e-05,
+      "loss": 0.7053,
+      "step": 5715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82522884704259e-05,
+      "loss": 18.7951,
+      "step": 5716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.824458286157527e-05,
+      "loss": 18.6896,
+      "step": 5717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823687626739187e-05,
+      "loss": 0.7491,
+      "step": 5718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.822916868814456e-05,
+      "loss": 20.1386,
+      "step": 5719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.822146012410221e-05,
+      "loss": 22.1131,
+      "step": 5720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821375057553376e-05,
+      "loss": 0.7171,
+      "step": 5721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.820604004270813e-05,
+      "loss": 19.6962,
+      "step": 5722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.819832852589435e-05,
+      "loss": 0.7293,
+      "step": 5723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81906160253614e-05,
+      "loss": 20.3765,
+      "step": 5724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818290254137835e-05,
+      "loss": 20.2616,
+      "step": 5725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817518807421429e-05,
+      "loss": 19.4773,
+      "step": 5726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816747262413837e-05,
+      "loss": 19.4191,
+      "step": 5727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815975619141971e-05,
+      "loss": 22.0106,
+      "step": 5728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81520387763275e-05,
+      "loss": 22.7808,
+      "step": 5729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814432037913102e-05,
+      "loss": 0.7028,
+      "step": 5730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81366010000995e-05,
+      "loss": 0.673,
+      "step": 5731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812888063950222e-05,
+      "loss": 18.7819,
+      "step": 5732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812115929760853e-05,
+      "loss": 0.6857,
+      "step": 5733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81134369746878e-05,
+      "loss": 0.6904,
+      "step": 5734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810571367100941e-05,
+      "loss": 20.3087,
+      "step": 5735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809798938684282e-05,
+      "loss": 19.4344,
+      "step": 5736
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809026412245749e-05,
+      "loss": 0.7066,
+      "step": 5737
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808253787812288e-05,
+      "loss": 18.2079,
+      "step": 5738
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807481065410856e-05,
+      "loss": 18.6043,
+      "step": 5739
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.806708245068411e-05,
+      "loss": 19.3313,
+      "step": 5740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805935326811912e-05,
+      "loss": 0.7427,
+      "step": 5741
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805162310668324e-05,
+      "loss": 21.722,
+      "step": 5742
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.804389196664612e-05,
+      "loss": 20.2069,
+      "step": 5743
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.803615984827747e-05,
+      "loss": 0.7307,
+      "step": 5744
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802842675184706e-05,
+      "loss": 22.3139,
+      "step": 5745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802069267762462e-05,
+      "loss": 20.7076,
+      "step": 5746
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801295762588e-05,
+      "loss": 0.6982,
+      "step": 5747
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.800522159688301e-05,
+      "loss": 20.7062,
+      "step": 5748
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799748459090356e-05,
+      "loss": 0.7221,
+      "step": 5749
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798974660821155e-05,
+      "loss": 20.6332,
+      "step": 5750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79820076490769e-05,
+      "loss": 19.9081,
+      "step": 5751
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79742677137696e-05,
+      "loss": 19.8426,
+      "step": 5752
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79665268025597e-05,
+      "loss": 19.4208,
+      "step": 5753
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79587849157172e-05,
+      "loss": 0.6961,
+      "step": 5754
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79510420535122e-05,
+      "loss": 0.6914,
+      "step": 5755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79432982162148e-05,
+      "loss": 0.7702,
+      "step": 5756
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.793555340409519e-05,
+      "loss": 0.725,
+      "step": 5757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792780761742353e-05,
+      "loss": 0.7252,
+      "step": 5758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792006085647001e-05,
+      "loss": 0.7433,
+      "step": 5759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.791231312150492e-05,
+      "loss": 0.7628,
+      "step": 5760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790456441279853e-05,
+      "loss": 20.7715,
+      "step": 5761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.789681473062116e-05,
+      "loss": 20.5203,
+      "step": 5762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788906407524317e-05,
+      "loss": 19.9179,
+      "step": 5763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788131244693494e-05,
+      "loss": 20.8608,
+      "step": 5764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78735598459669e-05,
+      "loss": 21.0933,
+      "step": 5765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.786580627260948e-05,
+      "loss": 17.5522,
+      "step": 5766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78580517271332e-05,
+      "loss": 19.2796,
+      "step": 5767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785029620980858e-05,
+      "loss": 18.565,
+      "step": 5768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784253972090614e-05,
+      "loss": 0.6817,
+      "step": 5769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.783478226069651e-05,
+      "loss": 0.7083,
+      "step": 5770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.782702382945031e-05,
+      "loss": 20.6378,
+      "step": 5771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781926442743818e-05,
+      "loss": 18.646,
+      "step": 5772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781150405493083e-05,
+      "loss": 18.2005,
+      "step": 5773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.780374271219897e-05,
+      "loss": 0.7735,
+      "step": 5774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.779598039951337e-05,
+      "loss": 0.7457,
+      "step": 5775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778821711714484e-05,
+      "loss": 20.0747,
+      "step": 5776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778045286536418e-05,
+      "loss": 20.4025,
+      "step": 5777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.777268764444228e-05,
+      "loss": 0.6779,
+      "step": 5778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776492145464999e-05,
+      "loss": 21.5819,
+      "step": 5779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.775715429625829e-05,
+      "loss": 19.2018,
+      "step": 5780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774938616953811e-05,
+      "loss": 19.3326,
+      "step": 5781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774161707476044e-05,
+      "loss": 0.6576,
+      "step": 5782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.773384701219635e-05,
+      "loss": 20.4948,
+      "step": 5783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.772607598211687e-05,
+      "loss": 0.6942,
+      "step": 5784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.771830398479311e-05,
+      "loss": 0.7394,
+      "step": 5785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.771053102049621e-05,
+      "loss": 20.0724,
+      "step": 5786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770275708949732e-05,
+      "loss": 19.7966,
+      "step": 5787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.769498219206766e-05,
+      "loss": 21.5278,
+      "step": 5788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.768720632847842e-05,
+      "loss": 0.7309,
+      "step": 5789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767942949900093e-05,
+      "loss": 18.3215,
+      "step": 5790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767165170390643e-05,
+      "loss": 18.9274,
+      "step": 5791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.766387294346628e-05,
+      "loss": 0.7107,
+      "step": 5792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765609321795186e-05,
+      "loss": 18.2666,
+      "step": 5793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764831252763456e-05,
+      "loss": 17.5174,
+      "step": 5794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764053087278582e-05,
+      "loss": 0.674,
+      "step": 5795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.763274825367708e-05,
+      "loss": 19.4644,
+      "step": 5796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76249646705799e-05,
+      "loss": 18.3267,
+      "step": 5797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761718012376574e-05,
+      "loss": 19.0823,
+      "step": 5798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.760939461350623e-05,
+      "loss": 18.3714,
+      "step": 5799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.760160814007296e-05,
+      "loss": 17.588,
+      "step": 5800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.759382070373755e-05,
+      "loss": 0.7002,
+      "step": 5801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758603230477168e-05,
+      "loss": 19.6849,
+      "step": 5802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757824294344707e-05,
+      "loss": 0.6825,
+      "step": 5803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757045262003544e-05,
+      "loss": 18.0018,
+      "step": 5804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.756266133480856e-05,
+      "loss": 0.737,
+      "step": 5805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755486908803824e-05,
+      "loss": 19.3067,
+      "step": 5806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75470758799963e-05,
+      "loss": 0.6741,
+      "step": 5807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753928171095463e-05,
+      "loss": 0.688,
+      "step": 5808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753148658118515e-05,
+      "loss": 20.6014,
+      "step": 5809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.752369049095978e-05,
+      "loss": 0.7556,
+      "step": 5810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.75158934405505e-05,
+      "loss": 19.7278,
+      "step": 5811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750809543022929e-05,
+      "loss": 18.7118,
+      "step": 5812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750029646026821e-05,
+      "loss": 19.4723,
+      "step": 5813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.749249653093935e-05,
+      "loss": 0.7097,
+      "step": 5814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748469564251478e-05,
+      "loss": 0.7104,
+      "step": 5815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747689379526668e-05,
+      "loss": 20.8253,
+      "step": 5816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.746909098946718e-05,
+      "loss": 21.7945,
+      "step": 5817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.746128722538853e-05,
+      "loss": 19.5901,
+      "step": 5818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.745348250330292e-05,
+      "loss": 0.7595,
+      "step": 5819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744567682348266e-05,
+      "loss": 0.7496,
+      "step": 5820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743787018620006e-05,
+      "loss": 20.8001,
+      "step": 5821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743006259172744e-05,
+      "loss": 0.7011,
+      "step": 5822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.742225404033718e-05,
+      "loss": 18.5684,
+      "step": 5823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741444453230169e-05,
+      "loss": 20.6669,
+      "step": 5824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74066340678934e-05,
+      "loss": 0.7634,
+      "step": 5825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.73988226473848e-05,
+      "loss": 19.5379,
+      "step": 5826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739101027104841e-05,
+      "loss": 0.6775,
+      "step": 5827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.738319693915672e-05,
+      "loss": 0.7015,
+      "step": 5828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737538265198234e-05,
+      "loss": 21.6587,
+      "step": 5829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736756740979788e-05,
+      "loss": 0.6877,
+      "step": 5830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735975121287598e-05,
+      "loss": 21.0111,
+      "step": 5831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735193406148929e-05,
+      "loss": 19.5812,
+      "step": 5832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734411595591053e-05,
+      "loss": 20.2001,
+      "step": 5833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.733629689641245e-05,
+      "loss": 21.0224,
+      "step": 5834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732847688326781e-05,
+      "loss": 0.6505,
+      "step": 5835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732065591674943e-05,
+      "loss": 19.8239,
+      "step": 5836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.731283399713014e-05,
+      "loss": 0.7247,
+      "step": 5837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730501112468282e-05,
+      "loss": 0.7869,
+      "step": 5838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729718729968037e-05,
+      "loss": 0.8068,
+      "step": 5839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728936252239571e-05,
+      "loss": 0.7688,
+      "step": 5840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728153679310187e-05,
+      "loss": 21.3356,
+      "step": 5841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72737101120718e-05,
+      "loss": 17.7794,
+      "step": 5842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726588247957857e-05,
+      "loss": 18.3207,
+      "step": 5843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725805389589525e-05,
+      "loss": 19.1382,
+      "step": 5844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725022436129492e-05,
+      "loss": 0.7356,
+      "step": 5845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.724239387605074e-05,
+      "loss": 18.7028,
+      "step": 5846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72345624404359e-05,
+      "loss": 0.6678,
+      "step": 5847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.722673005472358e-05,
+      "loss": 18.4722,
+      "step": 5848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.721889671918703e-05,
+      "loss": 0.6698,
+      "step": 5849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72110624340995e-05,
+      "loss": 20.2102,
+      "step": 5850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.720322719973431e-05,
+      "loss": 20.1484,
+      "step": 5851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719539101636483e-05,
+      "loss": 0.6984,
+      "step": 5852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.718755388426437e-05,
+      "loss": 18.542,
+      "step": 5853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.717971580370638e-05,
+      "loss": 0.7288,
+      "step": 5854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.717187677496427e-05,
+      "loss": 20.3368,
+      "step": 5855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.716403679831152e-05,
+      "loss": 19.2525,
+      "step": 5856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.715619587402164e-05,
+      "loss": 0.692,
+      "step": 5857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714835400236816e-05,
+      "loss": 21.5353,
+      "step": 5858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714051118362466e-05,
+      "loss": 0.7369,
+      "step": 5859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.713266741806474e-05,
+      "loss": 21.5287,
+      "step": 5860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.712482270596199e-05,
+      "loss": 0.7124,
+      "step": 5861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.711697704759013e-05,
+      "loss": 20.0023,
+      "step": 5862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710913044322286e-05,
+      "loss": 22.286,
+      "step": 5863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710128289313389e-05,
+      "loss": 0.7277,
+      "step": 5864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.709343439759702e-05,
+      "loss": 18.562,
+      "step": 5865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708558495688601e-05,
+      "loss": 21.2715,
+      "step": 5866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.707773457127472e-05,
+      "loss": 19.7513,
+      "step": 5867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706988324103701e-05,
+      "loss": 21.0491,
+      "step": 5868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70620309664468e-05,
+      "loss": 18.6354,
+      "step": 5869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.705417774777796e-05,
+      "loss": 0.7028,
+      "step": 5870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.704632358530453e-05,
+      "loss": 0.6904,
+      "step": 5871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.703846847930047e-05,
+      "loss": 18.4638,
+      "step": 5872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70306124300398e-05,
+      "loss": 0.697,
+      "step": 5873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.702275543779662e-05,
+      "loss": 0.7251,
+      "step": 5874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.7014897502845e-05,
+      "loss": 0.7192,
+      "step": 5875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.700703862545908e-05,
+      "loss": 19.0616,
+      "step": 5876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699917880591301e-05,
+      "loss": 0.7127,
+      "step": 5877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6991318044481e-05,
+      "loss": 21.3061,
+      "step": 5878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.698345634143729e-05,
+      "loss": 19.1469,
+      "step": 5879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.697559369705609e-05,
+      "loss": 18.2407,
+      "step": 5880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.696773011161176e-05,
+      "loss": 0.7024,
+      "step": 5881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695986558537858e-05,
+      "loss": 18.0964,
+      "step": 5882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695200011863093e-05,
+      "loss": 18.5365,
+      "step": 5883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69441337116432e-05,
+      "loss": 21.05,
+      "step": 5884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69362663646898e-05,
+      "loss": 20.1383,
+      "step": 5885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.692839807804521e-05,
+      "loss": 18.8142,
+      "step": 5886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.692052885198394e-05,
+      "loss": 21.4481,
+      "step": 5887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691265868678045e-05,
+      "loss": 20.3302,
+      "step": 5888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.690478758270933e-05,
+      "loss": 20.8789,
+      "step": 5889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.689691554004519e-05,
+      "loss": 20.0656,
+      "step": 5890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.688904255906263e-05,
+      "loss": 0.6255,
+      "step": 5891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68811686400363e-05,
+      "loss": 19.0552,
+      "step": 5892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68732937832409e-05,
+      "loss": 21.0807,
+      "step": 5893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686541798895115e-05,
+      "loss": 0.7728,
+      "step": 5894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.685754125744179e-05,
+      "loss": 19.5538,
+      "step": 5895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684966358898762e-05,
+      "loss": 21.0416,
+      "step": 5896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684178498386345e-05,
+      "loss": 18.3891,
+      "step": 5897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.683390544234412e-05,
+      "loss": 0.6886,
+      "step": 5898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682602496470455e-05,
+      "loss": 19.1001,
+      "step": 5899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681814355121961e-05,
+      "loss": 19.5157,
+      "step": 5900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681026120216429e-05,
+      "loss": 20.5796,
+      "step": 5901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680237791781354e-05,
+      "loss": 0.7486,
+      "step": 5902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679449369844241e-05,
+      "loss": 0.7521,
+      "step": 5903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.678660854432589e-05,
+      "loss": 20.988,
+      "step": 5904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677872245573912e-05,
+      "loss": 20.8629,
+      "step": 5905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677083543295716e-05,
+      "loss": 0.748,
+      "step": 5906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.676294747625519e-05,
+      "loss": 19.5605,
+      "step": 5907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675505858590839e-05,
+      "loss": 22.329,
+      "step": 5908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.674716876219194e-05,
+      "loss": 19.2661,
+      "step": 5909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67392780053811e-05,
+      "loss": 21.5689,
+      "step": 5910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.673138631575115e-05,
+      "loss": 18.6077,
+      "step": 5911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672349369357738e-05,
+      "loss": 19.6981,
+      "step": 5912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671560013913515e-05,
+      "loss": 19.6808,
+      "step": 5913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.670770565269983e-05,
+      "loss": 20.4781,
+      "step": 5914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669981023454682e-05,
+      "loss": 18.366,
+      "step": 5915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669191388495153e-05,
+      "loss": 18.9494,
+      "step": 5916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.668401660418947e-05,
+      "loss": 18.804,
+      "step": 5917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.667611839253613e-05,
+      "loss": 19.4047,
+      "step": 5918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666821925026704e-05,
+      "loss": 21.9216,
+      "step": 5919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666031917765778e-05,
+      "loss": 20.1353,
+      "step": 5920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665241817498391e-05,
+      "loss": 17.8442,
+      "step": 5921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66445162425211e-05,
+      "loss": 20.1136,
+      "step": 5922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.663661338054502e-05,
+      "loss": 19.3089,
+      "step": 5923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662870958933134e-05,
+      "loss": 0.6929,
+      "step": 5924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662080486915582e-05,
+      "loss": 0.6943,
+      "step": 5925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.661289922029419e-05,
+      "loss": 0.7611,
+      "step": 5926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.660499264302225e-05,
+      "loss": 22.1558,
+      "step": 5927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659708513761586e-05,
+      "loss": 20.424,
+      "step": 5928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658917670435084e-05,
+      "loss": 20.5218,
+      "step": 5929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658126734350309e-05,
+      "loss": 18.4779,
+      "step": 5930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.657335705534856e-05,
+      "loss": 0.7599,
+      "step": 5931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.656544584016316e-05,
+      "loss": 20.5794,
+      "step": 5932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655753369822291e-05,
+      "loss": 20.9954,
+      "step": 5933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654962062980384e-05,
+      "loss": 17.4468,
+      "step": 5934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6541706635182e-05,
+      "loss": 17.1878,
+      "step": 5935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.653379171463343e-05,
+      "loss": 18.6117,
+      "step": 5936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.652587586843429e-05,
+      "loss": 20.9185,
+      "step": 5937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651795909686074e-05,
+      "loss": 20.4138,
+      "step": 5938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651004140018894e-05,
+      "loss": 18.7825,
+      "step": 5939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.650212277869508e-05,
+      "loss": 21.1001,
+      "step": 5940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.649420323265545e-05,
+      "loss": 17.3315,
+      "step": 5941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.648628276234633e-05,
+      "loss": 0.7729,
+      "step": 5942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647836136804402e-05,
+      "loss": 19.461,
+      "step": 5943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647043905002484e-05,
+      "loss": 0.7255,
+      "step": 5944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64625158085652e-05,
+      "loss": 0.7445,
+      "step": 5945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645459164394149e-05,
+      "loss": 20.4946,
+      "step": 5946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.644666655643013e-05,
+      "loss": 19.2675,
+      "step": 5947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643874054630764e-05,
+      "loss": 0.7249,
+      "step": 5948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643081361385048e-05,
+      "loss": 0.7187,
+      "step": 5949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642288575933523e-05,
+      "loss": 0.6975,
+      "step": 5950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.641495698303843e-05,
+      "loss": 0.7428,
+      "step": 5951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.640702728523668e-05,
+      "loss": 19.3384,
+      "step": 5952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.639909666620663e-05,
+      "loss": 19.1189,
+      "step": 5953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.639116512622493e-05,
+      "loss": 19.5465,
+      "step": 5954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638323266556828e-05,
+      "loss": 19.0701,
+      "step": 5955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637529928451342e-05,
+      "loss": 0.6846,
+      "step": 5956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63673649833371e-05,
+      "loss": 22.4875,
+      "step": 5957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635942976231613e-05,
+      "loss": 20.5265,
+      "step": 5958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635149362172732e-05,
+      "loss": 18.0671,
+      "step": 5959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634355656184752e-05,
+      "loss": 18.8442,
+      "step": 5960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.633561858295364e-05,
+      "loss": 18.7427,
+      "step": 5961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63276796853226e-05,
+      "loss": 19.4597,
+      "step": 5962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631973986923135e-05,
+      "loss": 20.6165,
+      "step": 5963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631179913495685e-05,
+      "loss": 0.7142,
+      "step": 5964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.630385748277618e-05,
+      "loss": 0.7226,
+      "step": 5965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.629591491296635e-05,
+      "loss": 17.9975,
+      "step": 5966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628797142580442e-05,
+      "loss": 17.1005,
+      "step": 5967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628002702156755e-05,
+      "loss": 17.9114,
+      "step": 5968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.627208170053286e-05,
+      "loss": 0.7046,
+      "step": 5969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.626413546297755e-05,
+      "loss": 19.184,
+      "step": 5970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62561883091788e-05,
+      "loss": 0.7183,
+      "step": 5971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624824023941387e-05,
+      "loss": 0.689,
+      "step": 5972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624029125396004e-05,
+      "loss": 19.4104,
+      "step": 5973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.623234135309459e-05,
+      "loss": 0.7434,
+      "step": 5974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.622439053709489e-05,
+      "loss": 19.734,
+      "step": 5975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62164388062383e-05,
+      "loss": 0.7601,
+      "step": 5976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.620848616080221e-05,
+      "loss": 19.331,
+      "step": 5977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.620053260106406e-05,
+      "loss": 0.696,
+      "step": 5978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.619257812730131e-05,
+      "loss": 19.6432,
+      "step": 5979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.61846227397915e-05,
+      "loss": 0.7237,
+      "step": 5980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.617666643881209e-05,
+      "loss": 0.7142,
+      "step": 5981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.616870922464069e-05,
+      "loss": 0.6665,
+      "step": 5982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.616075109755488e-05,
+      "loss": 20.3859,
+      "step": 5983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615279205783228e-05,
+      "loss": 19.5367,
+      "step": 5984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614483210575055e-05,
+      "loss": 19.1462,
+      "step": 5985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.613687124158737e-05,
+      "loss": 0.7328,
+      "step": 5986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612890946562049e-05,
+      "loss": 17.6712,
+      "step": 5987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612094677812764e-05,
+      "loss": 0.7143,
+      "step": 5988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.61129831793866e-05,
+      "loss": 0.8344,
+      "step": 5989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.610501866967518e-05,
+      "loss": 18.9059,
+      "step": 5990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609705324927126e-05,
+      "loss": 19.4738,
+      "step": 5991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60890869184527e-05,
+      "loss": 0.6598,
+      "step": 5992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.608111967749741e-05,
+      "loss": 19.145,
+      "step": 5993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60731515266833e-05,
+      "loss": 0.6972,
+      "step": 5994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.606518246628841e-05,
+      "loss": 19.7575,
+      "step": 5995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.605721249659072e-05,
+      "loss": 19.7519,
+      "step": 5996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.604924161786827e-05,
+      "loss": 20.2949,
+      "step": 5997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60412698303991e-05,
+      "loss": 18.5267,
+      "step": 5998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.603329713446135e-05,
+      "loss": 20.6289,
+      "step": 5999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.602532353033313e-05,
+      "loss": 20.5514,
+      "step": 6000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601734901829261e-05,
+      "loss": 0.7655,
+      "step": 6001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6009373598618e-05,
+      "loss": 0.7694,
+      "step": 6002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60013972715875e-05,
+      "loss": 20.1851,
+      "step": 6003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.599342003747941e-05,
+      "loss": 18.3731,
+      "step": 6004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5985441896572e-05,
+      "loss": 17.7535,
+      "step": 6005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597746284914358e-05,
+      "loss": 19.5814,
+      "step": 6006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596948289547253e-05,
+      "loss": 20.3369,
+      "step": 6007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596150203583719e-05,
+      "loss": 0.6929,
+      "step": 6008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.595352027051605e-05,
+      "loss": 18.1033,
+      "step": 6009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594553759978751e-05,
+      "loss": 18.6723,
+      "step": 6010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.593755402393006e-05,
+      "loss": 20.0738,
+      "step": 6011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.59295695432222e-05,
+      "loss": 0.7357,
+      "step": 6012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.592158415794252e-05,
+      "loss": 20.8538,
+      "step": 6013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.591359786836956e-05,
+      "loss": 20.2297,
+      "step": 6014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.590561067478193e-05,
+      "loss": 19.6461,
+      "step": 6015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.589762257745828e-05,
+      "loss": 0.7059,
+      "step": 6016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.588963357667725e-05,
+      "loss": 0.7322,
+      "step": 6017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.58816436727176e-05,
+      "loss": 17.7051,
+      "step": 6018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5873652865858e-05,
+      "loss": 0.7207,
+      "step": 6019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.586566115637725e-05,
+      "loss": 0.7194,
+      "step": 6020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.585766854455414e-05,
+      "loss": 21.2053,
+      "step": 6021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584967503066751e-05,
+      "loss": 19.1969,
+      "step": 6022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584168061499621e-05,
+      "loss": 21.0754,
+      "step": 6023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.583368529781912e-05,
+      "loss": 21.371,
+      "step": 6024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582568907941519e-05,
+      "loss": 19.1444,
+      "step": 6025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.581769196006331e-05,
+      "loss": 19.5083,
+      "step": 6026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.580969394004256e-05,
+      "loss": 20.7072,
+      "step": 6027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.580169501963188e-05,
+      "loss": 0.7359,
+      "step": 6028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579369519911035e-05,
+      "loss": 19.0655,
+      "step": 6029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.578569447875706e-05,
+      "loss": 19.3076,
+      "step": 6030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.577769285885109e-05,
+      "loss": 19.4805,
+      "step": 6031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57696903396716e-05,
+      "loss": 20.9132,
+      "step": 6032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576168692149776e-05,
+      "loss": 19.1466,
+      "step": 6033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57536826046088e-05,
+      "loss": 21.0354,
+      "step": 6034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.574567738928393e-05,
+      "loss": 18.2896,
+      "step": 6035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57376712758024e-05,
+      "loss": 19.6118,
+      "step": 6036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.572966426444355e-05,
+      "loss": 18.6133,
+      "step": 6037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.572165635548667e-05,
+      "loss": 17.3118,
+      "step": 6038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.571364754921115e-05,
+      "loss": 0.6758,
+      "step": 6039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570563784589639e-05,
+      "loss": 21.5285,
+      "step": 6040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569762724582179e-05,
+      "loss": 0.7034,
+      "step": 6041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.568961574926683e-05,
+      "loss": 18.7759,
+      "step": 6042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.568160335651095e-05,
+      "loss": 20.8322,
+      "step": 6043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567359006783374e-05,
+      "loss": 20.6643,
+      "step": 6044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.566557588351468e-05,
+      "loss": 21.612,
+      "step": 6045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.56575608038334e-05,
+      "loss": 0.7398,
+      "step": 6046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.564954482906948e-05,
+      "loss": 20.2071,
+      "step": 6047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.564152795950256e-05,
+      "loss": 18.6716,
+      "step": 6048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.563351019541237e-05,
+      "loss": 19.3102,
+      "step": 6049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562549153707856e-05,
+      "loss": 0.7233,
+      "step": 6050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.561747198478088e-05,
+      "loss": 18.999,
+      "step": 6051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.560945153879911e-05,
+      "loss": 0.7446,
+      "step": 6052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.560143019941301e-05,
+      "loss": 19.6011,
+      "step": 6053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.559340796690247e-05,
+      "loss": 0.7068,
+      "step": 6054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.558538484154729e-05,
+      "loss": 21.1891,
+      "step": 6055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.557736082362743e-05,
+      "loss": 0.6876,
+      "step": 6056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556933591342275e-05,
+      "loss": 19.5972,
+      "step": 6057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556131011121323e-05,
+      "loss": 18.1697,
+      "step": 6058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.555328341727888e-05,
+      "loss": 0.7264,
+      "step": 6059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554525583189969e-05,
+      "loss": 18.4151,
+      "step": 6060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55372273553557e-05,
+      "loss": 19.7899,
+      "step": 6061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.552919798792701e-05,
+      "loss": 0.7102,
+      "step": 6062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.552116772989373e-05,
+      "loss": 0.7315,
+      "step": 6063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.551313658153598e-05,
+      "loss": 20.793,
+      "step": 6064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.550510454313396e-05,
+      "loss": 21.7241,
+      "step": 6065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.549707161496783e-05,
+      "loss": 21.3982,
+      "step": 6066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548903779731788e-05,
+      "loss": 20.0006,
+      "step": 6067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548100309046436e-05,
+      "loss": 19.2593,
+      "step": 6068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547296749468754e-05,
+      "loss": 18.3285,
+      "step": 6069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.546493101026778e-05,
+      "loss": 0.7055,
+      "step": 6070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54568936374854e-05,
+      "loss": 19.5656,
+      "step": 6071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544885537662083e-05,
+      "loss": 0.6846,
+      "step": 6072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544081622795444e-05,
+      "loss": 0.7334,
+      "step": 6073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.543277619176676e-05,
+      "loss": 0.7458,
+      "step": 6074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54247352683382e-05,
+      "loss": 0.7666,
+      "step": 6075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54166934579493e-05,
+      "loss": 19.6955,
+      "step": 6076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540865076088063e-05,
+      "loss": 21.2078,
+      "step": 6077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540060717741273e-05,
+      "loss": 0.716,
+      "step": 6078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53925627078262e-05,
+      "loss": 0.7533,
+      "step": 6079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53845173524017e-05,
+      "loss": 0.6769,
+      "step": 6080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53764711114199e-05,
+      "loss": 19.3715,
+      "step": 6081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.536842398516149e-05,
+      "loss": 17.9219,
+      "step": 6082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53603759739072e-05,
+      "loss": 22.165,
+      "step": 6083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535232707793778e-05,
+      "loss": 0.689,
+      "step": 6084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.534427729753404e-05,
+      "loss": 0.6936,
+      "step": 6085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.533622663297679e-05,
+      "loss": 18.3053,
+      "step": 6086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.532817508454688e-05,
+      "loss": 0.72,
+      "step": 6087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.532012265252523e-05,
+      "loss": 19.1247,
+      "step": 6088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.53120693371927e-05,
+      "loss": 0.7009,
+      "step": 6089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.530401513883028e-05,
+      "loss": 19.2176,
+      "step": 6090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529596005771893e-05,
+      "loss": 0.6737,
+      "step": 6091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.528790409413965e-05,
+      "loss": 0.6479,
+      "step": 6092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527984724837349e-05,
+      "loss": 18.8384,
+      "step": 6093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.52717895207015e-05,
+      "loss": 0.7558,
+      "step": 6094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.526373091140482e-05,
+      "loss": 19.1181,
+      "step": 6095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.525567142076454e-05,
+      "loss": 0.7244,
+      "step": 6096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.524761104906183e-05,
+      "loss": 0.6711,
+      "step": 6097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.52395497965779e-05,
+      "loss": 19.6927,
+      "step": 6098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523148766359395e-05,
+      "loss": 19.9553,
+      "step": 6099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522342465039123e-05,
+      "loss": 19.1703,
+      "step": 6100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.521536075725106e-05,
+      "loss": 18.9897,
+      "step": 6101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520729598445473e-05,
+      "loss": 0.7074,
+      "step": 6102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.519923033228358e-05,
+      "loss": 0.6707,
+      "step": 6103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5191163801019e-05,
+      "loss": 16.7422,
+      "step": 6104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51830963909424e-05,
+      "loss": 19.5632,
+      "step": 6105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51750281023352e-05,
+      "loss": 0.6999,
+      "step": 6106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516695893547887e-05,
+      "loss": 0.6657,
+      "step": 6107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515888889065492e-05,
+      "loss": 19.4031,
+      "step": 6108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515081796814486e-05,
+      "loss": 0.7266,
+      "step": 6109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.514274616823028e-05,
+      "loss": 0.6754,
+      "step": 6110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.513467349119275e-05,
+      "loss": 0.7063,
+      "step": 6111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51265999373139e-05,
+      "loss": 0.6889,
+      "step": 6112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.511852550687537e-05,
+      "loss": 19.8983,
+      "step": 6113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.511045020015885e-05,
+      "loss": 20.1548,
+      "step": 6114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.510237401744607e-05,
+      "loss": 0.7185,
+      "step": 6115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.509429695901875e-05,
+      "loss": 19.269,
+      "step": 6116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508621902515864e-05,
+      "loss": 0.674,
+      "step": 6117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507814021614761e-05,
+      "loss": 0.6884,
+      "step": 6118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507006053226746e-05,
+      "loss": 19.1352,
+      "step": 6119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506197997380003e-05,
+      "loss": 17.7133,
+      "step": 6120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505389854102727e-05,
+      "loss": 18.294,
+      "step": 6121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.504581623423105e-05,
+      "loss": 21.1058,
+      "step": 6122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503773305369338e-05,
+      "loss": 20.8291,
+      "step": 6123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.502964899969622e-05,
+      "loss": 21.1766,
+      "step": 6124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.502156407252159e-05,
+      "loss": 20.2158,
+      "step": 6125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.501347827245153e-05,
+      "loss": 17.8379,
+      "step": 6126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500539159976813e-05,
+      "loss": 0.7388,
+      "step": 6127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.499730405475352e-05,
+      "loss": 0.7245,
+      "step": 6128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49892156376898e-05,
+      "loss": 0.708,
+      "step": 6129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498112634885917e-05,
+      "loss": 20.4003,
+      "step": 6130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497303618854381e-05,
+      "loss": 0.6764,
+      "step": 6131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.496494515702595e-05,
+      "loss": 17.3358,
+      "step": 6132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49568532545879e-05,
+      "loss": 22.016,
+      "step": 6133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.494876048151188e-05,
+      "loss": 0.763,
+      "step": 6134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.494066683808025e-05,
+      "loss": 21.6505,
+      "step": 6135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.493257232457538e-05,
+      "loss": 19.8204,
+      "step": 6136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49244769412796e-05,
+      "loss": 20.2563,
+      "step": 6137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.491638068847538e-05,
+      "loss": 0.7604,
+      "step": 6138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.490828356644513e-05,
+      "loss": 0.649,
+      "step": 6139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.490018557547134e-05,
+      "loss": 19.1732,
+      "step": 6140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.48920867158365e-05,
+      "loss": 0.7442,
+      "step": 6141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488398698782316e-05,
+      "loss": 20.3288,
+      "step": 6142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.487588639171386e-05,
+      "loss": 19.0614,
+      "step": 6143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.486778492779121e-05,
+      "loss": 17.2693,
+      "step": 6144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485968259633786e-05,
+      "loss": 20.9867,
+      "step": 6145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485157939763643e-05,
+      "loss": 19.5738,
+      "step": 6146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.484347533196961e-05,
+      "loss": 0.7375,
+      "step": 6147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483537039962014e-05,
+      "loss": 20.8361,
+      "step": 6148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482726460087074e-05,
+      "loss": 0.7715,
+      "step": 6149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.48191579360042e-05,
+      "loss": 19.4513,
+      "step": 6150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.481105040530334e-05,
+      "loss": 0.6665,
+      "step": 6151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.480294200905097e-05,
+      "loss": 20.1612,
+      "step": 6152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.479483274752998e-05,
+      "loss": 19.4584,
+      "step": 6153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.478672262102324e-05,
+      "loss": 20.6223,
+      "step": 6154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.47786116298137e-05,
+      "loss": 0.698,
+      "step": 6155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.477049977418432e-05,
+      "loss": 18.5484,
+      "step": 6156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.476238705441807e-05,
+      "loss": 0.6482,
+      "step": 6157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.475427347079799e-05,
+      "loss": 19.8203,
+      "step": 6158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474615902360711e-05,
+      "loss": 17.7983,
+      "step": 6159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.473804371312852e-05,
+      "loss": 0.719,
+      "step": 6160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472992753964532e-05,
+      "loss": 0.6456,
+      "step": 6161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472181050344065e-05,
+      "loss": 19.4003,
+      "step": 6162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.471369260479769e-05,
+      "loss": 20.219,
+      "step": 6163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.47055738439996e-05,
+      "loss": 19.0508,
+      "step": 6164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.469745422132968e-05,
+      "loss": 0.6819,
+      "step": 6165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468933373707112e-05,
+      "loss": 0.7505,
+      "step": 6166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468121239150725e-05,
+      "loss": 18.3764,
+      "step": 6167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.467309018492137e-05,
+      "loss": 0.7246,
+      "step": 6168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.466496711759684e-05,
+      "loss": 0.6848,
+      "step": 6169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465684318981702e-05,
+      "loss": 18.1412,
+      "step": 6170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.464871840186535e-05,
+      "loss": 0.714,
+      "step": 6171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.464059275402524e-05,
+      "loss": 17.3902,
+      "step": 6172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.463246624658018e-05,
+      "loss": 18.5242,
+      "step": 6173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.462433887981365e-05,
+      "loss": 0.721,
+      "step": 6174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.46162106540092e-05,
+      "loss": 0.7211,
+      "step": 6175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.460808156945036e-05,
+      "loss": 18.4562,
+      "step": 6176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459995162642077e-05,
+      "loss": 20.1405,
+      "step": 6177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4591820825204e-05,
+      "loss": 0.7193,
+      "step": 6178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.45836891660837e-05,
+      "loss": 0.7132,
+      "step": 6179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457555664934357e-05,
+      "loss": 20.4815,
+      "step": 6180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.456742327526731e-05,
+      "loss": 18.8096,
+      "step": 6181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455928904413868e-05,
+      "loss": 17.1379,
+      "step": 6182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455115395624143e-05,
+      "loss": 20.1894,
+      "step": 6183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.454301801185935e-05,
+      "loss": 20.927,
+      "step": 6184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453488121127626e-05,
+      "loss": 0.731,
+      "step": 6185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.452674355477604e-05,
+      "loss": 18.8941,
+      "step": 6186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451860504264261e-05,
+      "loss": 0.6852,
+      "step": 6187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451046567515981e-05,
+      "loss": 19.544,
+      "step": 6188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.450232545261165e-05,
+      "loss": 19.1468,
+      "step": 6189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.449418437528209e-05,
+      "loss": 21.2798,
+      "step": 6190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.44860424434551e-05,
+      "loss": 21.3904,
+      "step": 6191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.447789965741478e-05,
+      "loss": 0.7314,
+      "step": 6192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.446975601744516e-05,
+      "loss": 19.7522,
+      "step": 6193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.446161152383036e-05,
+      "loss": 19.6525,
+      "step": 6194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445346617685447e-05,
+      "loss": 0.7391,
+      "step": 6195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444531997680167e-05,
+      "loss": 19.7679,
+      "step": 6196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.443717292395615e-05,
+      "loss": 0.7559,
+      "step": 6197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442902501860211e-05,
+      "loss": 0.7296,
+      "step": 6198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442087626102382e-05,
+      "loss": 0.7318,
+      "step": 6199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.441272665150553e-05,
+      "loss": 0.7695,
+      "step": 6200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.440457619033156e-05,
+      "loss": 0.7235,
+      "step": 6201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.439642487778624e-05,
+      "loss": 19.9878,
+      "step": 6202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438827271415393e-05,
+      "loss": 19.28,
+      "step": 6203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438011969971903e-05,
+      "loss": 19.8219,
+      "step": 6204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.437196583476596e-05,
+      "loss": 22.2373,
+      "step": 6205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.436381111957918e-05,
+      "loss": 19.4213,
+      "step": 6206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.435565555444317e-05,
+      "loss": 20.6278,
+      "step": 6207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434749913964246e-05,
+      "loss": 19.3375,
+      "step": 6208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.433934187546154e-05,
+      "loss": 0.7214,
+      "step": 6209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.433118376218504e-05,
+      "loss": 20.0978,
+      "step": 6210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432302480009752e-05,
+      "loss": 20.2293,
+      "step": 6211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.431486498948365e-05,
+      "loss": 19.4793,
+      "step": 6212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430670433062806e-05,
+      "loss": 0.7234,
+      "step": 6213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.429854282381545e-05,
+      "loss": 19.0607,
+      "step": 6214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.429038046933056e-05,
+      "loss": 0.694,
+      "step": 6215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428221726745808e-05,
+      "loss": 17.0551,
+      "step": 6216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.427405321848286e-05,
+      "loss": 20.143,
+      "step": 6217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.426588832268968e-05,
+      "loss": 0.7287,
+      "step": 6218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425772258036336e-05,
+      "loss": 0.7168,
+      "step": 6219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.424955599178881e-05,
+      "loss": 19.6544,
+      "step": 6220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.424138855725088e-05,
+      "loss": 18.2577,
+      "step": 6221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.423322027703452e-05,
+      "loss": 0.7073,
+      "step": 6222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422505115142469e-05,
+      "loss": 0.7591,
+      "step": 6223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.421688118070637e-05,
+      "loss": 0.6908,
+      "step": 6224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420871036516459e-05,
+      "loss": 19.3091,
+      "step": 6225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420053870508437e-05,
+      "loss": 19.0896,
+      "step": 6226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.419236620075079e-05,
+      "loss": 20.7327,
+      "step": 6227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.418419285244897e-05,
+      "loss": 0.7206,
+      "step": 6228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.417601866046403e-05,
+      "loss": 20.3372,
+      "step": 6229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.416784362508114e-05,
+      "loss": 0.6993,
+      "step": 6230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.415966774658548e-05,
+      "loss": 0.7128,
+      "step": 6231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.415149102526228e-05,
+      "loss": 0.7173,
+      "step": 6232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.414331346139678e-05,
+      "loss": 18.7119,
+      "step": 6233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413513505527429e-05,
+      "loss": 0.655,
+      "step": 6234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41269558071801e-05,
+      "loss": 19.9946,
+      "step": 6235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411877571739953e-05,
+      "loss": 19.4945,
+      "step": 6236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411059478621798e-05,
+      "loss": 19.6507,
+      "step": 6237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.410241301392082e-05,
+      "loss": 0.6553,
+      "step": 6238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.409423040079352e-05,
+      "loss": 18.4656,
+      "step": 6239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.408604694712149e-05,
+      "loss": 0.7301,
+      "step": 6240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.407786265319023e-05,
+      "loss": 20.4367,
+      "step": 6241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.406967751928528e-05,
+      "loss": 17.7719,
+      "step": 6242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.406149154569214e-05,
+      "loss": 0.6881,
+      "step": 6243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405330473269641e-05,
+      "loss": 19.6929,
+      "step": 6244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.40451170805837e-05,
+      "loss": 21.1983,
+      "step": 6245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.403692858963962e-05,
+      "loss": 19.8026,
+      "step": 6246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402873926014986e-05,
+      "loss": 19.2934,
+      "step": 6247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402054909240009e-05,
+      "loss": 19.3636,
+      "step": 6248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.401235808667603e-05,
+      "loss": 19.9724,
+      "step": 6249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400416624326342e-05,
+      "loss": 0.694,
+      "step": 6250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.399597356244806e-05,
+      "loss": 0.6916,
+      "step": 6251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.398778004451576e-05,
+      "loss": 18.7689,
+      "step": 6252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.397958568975234e-05,
+      "loss": 18.4407,
+      "step": 6253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.397139049844367e-05,
+      "loss": 0.7079,
+      "step": 6254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.396319447087564e-05,
+      "loss": 18.5369,
+      "step": 6255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.39549976073342e-05,
+      "loss": 19.4337,
+      "step": 6256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.394679990810528e-05,
+      "loss": 20.2358,
+      "step": 6257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393860137347486e-05,
+      "loss": 0.6617,
+      "step": 6258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393040200372896e-05,
+      "loss": 0.6918,
+      "step": 6259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.392220179915361e-05,
+      "loss": 0.7057,
+      "step": 6260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.391400076003492e-05,
+      "loss": 21.6595,
+      "step": 6261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390579888665894e-05,
+      "loss": 18.8828,
+      "step": 6262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.389759617931182e-05,
+      "loss": 20.3654,
+      "step": 6263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.388939263827971e-05,
+      "loss": 19.8654,
+      "step": 6264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.388118826384883e-05,
+      "loss": 19.7856,
+      "step": 6265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.387298305630535e-05,
+      "loss": 19.3456,
+      "step": 6266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.386477701593553e-05,
+      "loss": 18.6059,
+      "step": 6267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.385657014302564e-05,
+      "loss": 19.385,
+      "step": 6268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3848362437862e-05,
+      "loss": 19.0585,
+      "step": 6269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.384015390073093e-05,
+      "loss": 0.7341,
+      "step": 6270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.383194453191879e-05,
+      "loss": 22.043,
+      "step": 6271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382373433171198e-05,
+      "loss": 19.8385,
+      "step": 6272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38155233003969e-05,
+      "loss": 0.7373,
+      "step": 6273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.380731143826e-05,
+      "loss": 0.6192,
+      "step": 6274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.379909874558777e-05,
+      "loss": 0.6543,
+      "step": 6275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37908852226667e-05,
+      "loss": 18.3753,
+      "step": 6276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.378267086978336e-05,
+      "loss": 19.9166,
+      "step": 6277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.377445568722427e-05,
+      "loss": 19.5227,
+      "step": 6278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.376623967527604e-05,
+      "loss": 0.7153,
+      "step": 6279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.375802283422531e-05,
+      "loss": 0.7614,
+      "step": 6280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37498051643587e-05,
+      "loss": 21.1673,
+      "step": 6281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.374158666596289e-05,
+      "loss": 0.7265,
+      "step": 6282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373336733932461e-05,
+      "loss": 0.6958,
+      "step": 6283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.372514718473062e-05,
+      "loss": 0.7123,
+      "step": 6284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.371692620246762e-05,
+      "loss": 0.7308,
+      "step": 6285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370870439282245e-05,
+      "loss": 0.7637,
+      "step": 6286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37004817560819e-05,
+      "loss": 0.7476,
+      "step": 6287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.369225829253288e-05,
+      "loss": 21.9789,
+      "step": 6288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.368403400246222e-05,
+      "loss": 19.8325,
+      "step": 6289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367580888615684e-05,
+      "loss": 20.1911,
+      "step": 6290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36675829439037e-05,
+      "loss": 19.5131,
+      "step": 6291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365935617598975e-05,
+      "loss": 0.6792,
+      "step": 6292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365112858270198e-05,
+      "loss": 19.3864,
+      "step": 6293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364290016432746e-05,
+      "loss": 17.0904,
+      "step": 6294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.363467092115317e-05,
+      "loss": 0.6225,
+      "step": 6295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362644085346626e-05,
+      "loss": 20.2049,
+      "step": 6296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.361820996155383e-05,
+      "loss": 0.7716,
+      "step": 6297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.360997824570297e-05,
+      "loss": 20.2673,
+      "step": 6298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.360174570620089e-05,
+      "loss": 20.2356,
+      "step": 6299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.35935123433348e-05,
+      "loss": 19.8281,
+      "step": 6300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.358527815739192e-05,
+      "loss": 18.9435,
+      "step": 6301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357704314865948e-05,
+      "loss": 19.5833,
+      "step": 6302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356880731742478e-05,
+      "loss": 18.6884,
+      "step": 6303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356057066397513e-05,
+      "loss": 19.2102,
+      "step": 6304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.355233318859789e-05,
+      "loss": 0.7116,
+      "step": 6305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.35440948915804e-05,
+      "loss": 19.4272,
+      "step": 6306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.353585577321009e-05,
+      "loss": 0.6357,
+      "step": 6307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.352761583377437e-05,
+      "loss": 0.727,
+      "step": 6308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351937507356071e-05,
+      "loss": 18.8635,
+      "step": 6309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351113349285655e-05,
+      "loss": 0.6896,
+      "step": 6310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.350289109194946e-05,
+      "loss": 0.7772,
+      "step": 6311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.349464787112694e-05,
+      "loss": 0.7483,
+      "step": 6312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.34864038306766e-05,
+      "loss": 0.7043,
+      "step": 6313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.347815897088602e-05,
+      "loss": 17.9641,
+      "step": 6314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346991329204282e-05,
+      "loss": 19.0761,
+      "step": 6315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346166679443467e-05,
+      "loss": 19.6898,
+      "step": 6316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.345341947834922e-05,
+      "loss": 18.9713,
+      "step": 6317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344517134407425e-05,
+      "loss": 19.6376,
+      "step": 6318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343692239189746e-05,
+      "loss": 17.8596,
+      "step": 6319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.34286726221066e-05,
+      "loss": 0.7079,
+      "step": 6320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.342042203498951e-05,
+      "loss": 0.7448,
+      "step": 6321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3412170630834e-05,
+      "loss": 20.6322,
+      "step": 6322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340391840992793e-05,
+      "loss": 18.6102,
+      "step": 6323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.339566537255919e-05,
+      "loss": 0.672,
+      "step": 6324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.338741151901568e-05,
+      "loss": 19.3103,
+      "step": 6325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.337915684958534e-05,
+      "loss": 20.5243,
+      "step": 6326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.337090136455615e-05,
+      "loss": 0.7106,
+      "step": 6327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.336264506421612e-05,
+      "loss": 0.6776,
+      "step": 6328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.335438794885325e-05,
+      "loss": 20.7727,
+      "step": 6329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.334613001875561e-05,
+      "loss": 19.8469,
+      "step": 6330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.333787127421129e-05,
+      "loss": 0.6781,
+      "step": 6331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.332961171550838e-05,
+      "loss": 20.1656,
+      "step": 6332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.332135134293505e-05,
+      "loss": 19.5982,
+      "step": 6333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.331309015677946e-05,
+      "loss": 20.9829,
+      "step": 6334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.33048281573298e-05,
+      "loss": 19.2552,
+      "step": 6335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.329656534487429e-05,
+      "loss": 0.7397,
+      "step": 6336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328830171970119e-05,
+      "loss": 18.3288,
+      "step": 6337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328003728209878e-05,
+      "loss": 0.6559,
+      "step": 6338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.327177203235539e-05,
+      "loss": 21.6311,
+      "step": 6339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.326350597075936e-05,
+      "loss": 17.3997,
+      "step": 6340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.325523909759902e-05,
+      "loss": 18.9409,
+      "step": 6341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.32469714131628e-05,
+      "loss": 0.7418,
+      "step": 6342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.32387029177391e-05,
+      "loss": 18.4416,
+      "step": 6343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323043361161638e-05,
+      "loss": 18.8508,
+      "step": 6344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.322216349508314e-05,
+      "loss": 0.6986,
+      "step": 6345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321389256842788e-05,
+      "loss": 0.7093,
+      "step": 6346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.320562083193911e-05,
+      "loss": 18.4656,
+      "step": 6347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319734828590544e-05,
+      "loss": 0.7052,
+      "step": 6348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.318907493061543e-05,
+      "loss": 20.5982,
+      "step": 6349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.318080076635772e-05,
+      "loss": 0.7129,
+      "step": 6350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.317252579342095e-05,
+      "loss": 20.477,
+      "step": 6351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.316425001209381e-05,
+      "loss": 18.912,
+      "step": 6352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3155973422665e-05,
+      "loss": 21.0325,
+      "step": 6353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.314769602542325e-05,
+      "loss": 17.0995,
+      "step": 6354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.313941782065732e-05,
+      "loss": 0.7394,
+      "step": 6355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3131138808656e-05,
+      "loss": 17.8951,
+      "step": 6356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.312285898970815e-05,
+      "loss": 20.3573,
+      "step": 6357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311457836410256e-05,
+      "loss": 0.6683,
+      "step": 6358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.310629693212812e-05,
+      "loss": 17.8966,
+      "step": 6359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309801469407377e-05,
+      "loss": 0.7514,
+      "step": 6360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.30897316502284e-05,
+      "loss": 19.5433,
+      "step": 6361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3081447800881e-05,
+      "loss": 0.6655,
+      "step": 6362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.307316314632053e-05,
+      "loss": 0.716,
+      "step": 6363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.306487768683601e-05,
+      "loss": 18.7057,
+      "step": 6364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.305659142271652e-05,
+      "loss": 19.3664,
+      "step": 6365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.30483043542511e-05,
+      "loss": 20.1237,
+      "step": 6366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.304001648172884e-05,
+      "loss": 0.687,
+      "step": 6367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.303172780543892e-05,
+      "loss": 19.9751,
+      "step": 6368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.302343832567043e-05,
+      "loss": 19.8134,
+      "step": 6369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.301514804271262e-05,
+      "loss": 20.925,
+      "step": 6370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.300685695685462e-05,
+      "loss": 19.1785,
+      "step": 6371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299856506838575e-05,
+      "loss": 0.7265,
+      "step": 6372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299027237759525e-05,
+      "loss": 0.7286,
+      "step": 6373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.298197888477242e-05,
+      "loss": 0.6472,
+      "step": 6374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297368459020657e-05,
+      "loss": 0.7508,
+      "step": 6375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.296538949418706e-05,
+      "loss": 0.73,
+      "step": 6376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295709359700328e-05,
+      "loss": 19.4454,
+      "step": 6377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.294879689894463e-05,
+      "loss": 20.027,
+      "step": 6378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.294049940030055e-05,
+      "loss": 19.9204,
+      "step": 6379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.29322011013605e-05,
+      "loss": 20.4521,
+      "step": 6380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.292390200241396e-05,
+      "loss": 0.7309,
+      "step": 6381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291560210375047e-05,
+      "loss": 19.0482,
+      "step": 6382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.290730140565957e-05,
+      "loss": 0.6479,
+      "step": 6383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289899990843084e-05,
+      "loss": 21.5235,
+      "step": 6384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289069761235389e-05,
+      "loss": 18.8899,
+      "step": 6385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.288239451771832e-05,
+      "loss": 18.8331,
+      "step": 6386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.287409062481384e-05,
+      "loss": 19.2255,
+      "step": 6387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286578593393009e-05,
+      "loss": 0.7698,
+      "step": 6388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.285748044535682e-05,
+      "loss": 18.8523,
+      "step": 6389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284917415938375e-05,
+      "loss": 0.7355,
+      "step": 6390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284086707630065e-05,
+      "loss": 0.7444,
+      "step": 6391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.283255919639734e-05,
+      "loss": 21.2649,
+      "step": 6392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.282425051996363e-05,
+      "loss": 20.1666,
+      "step": 6393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.281594104728938e-05,
+      "loss": 0.7371,
+      "step": 6394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280763077866447e-05,
+      "loss": 0.626,
+      "step": 6395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27993197143788e-05,
+      "loss": 20.3088,
+      "step": 6396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.279100785472233e-05,
+      "loss": 20.7299,
+      "step": 6397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.278269519998502e-05,
+      "loss": 0.6844,
+      "step": 6398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.277438175045683e-05,
+      "loss": 0.7173,
+      "step": 6399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.276606750642783e-05,
+      "loss": 17.3508,
+      "step": 6400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.275775246818803e-05,
+      "loss": 18.4494,
+      "step": 6401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.274943663602751e-05,
+      "loss": 0.6614,
+      "step": 6402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.274112001023641e-05,
+      "loss": 0.7622,
+      "step": 6403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.273280259110482e-05,
+      "loss": 0.6859,
+      "step": 6404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.272448437892292e-05,
+      "loss": 0.6965,
+      "step": 6405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27161653739809e-05,
+      "loss": 19.0473,
+      "step": 6406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270784557656896e-05,
+      "loss": 0.6949,
+      "step": 6407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.269952498697734e-05,
+      "loss": 18.187,
+      "step": 6408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.269120360549634e-05,
+      "loss": 0.7155,
+      "step": 6409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.268288143241621e-05,
+      "loss": 19.2009,
+      "step": 6410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.267455846802732e-05,
+      "loss": 0.7214,
+      "step": 6411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.266623471262e-05,
+      "loss": 0.694,
+      "step": 6412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.265791016648463e-05,
+      "loss": 23.2254,
+      "step": 6413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264958482991162e-05,
+      "loss": 18.8829,
+      "step": 6414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264125870319142e-05,
+      "loss": 20.4105,
+      "step": 6415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.263293178661447e-05,
+      "loss": 19.9438,
+      "step": 6416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262460408047127e-05,
+      "loss": 17.3849,
+      "step": 6417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.261627558505233e-05,
+      "loss": 0.6804,
+      "step": 6418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.260794630064822e-05,
+      "loss": 18.8989,
+      "step": 6419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259961622754951e-05,
+      "loss": 19.5068,
+      "step": 6420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259128536604676e-05,
+      "loss": 19.3935,
+      "step": 6421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.258295371643064e-05,
+      "loss": 20.0594,
+      "step": 6422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.257462127899178e-05,
+      "loss": 0.7492,
+      "step": 6423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256628805402089e-05,
+      "loss": 18.6733,
+      "step": 6424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.255795404180867e-05,
+      "loss": 18.5653,
+      "step": 6425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254961924264582e-05,
+      "loss": 19.6261,
+      "step": 6426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254128365682318e-05,
+      "loss": 0.7069,
+      "step": 6427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.253294728463148e-05,
+      "loss": 18.777,
+      "step": 6428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.252461012636158e-05,
+      "loss": 20.459,
+      "step": 6429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.25162721823043e-05,
+      "loss": 18.8342,
+      "step": 6430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.250793345275051e-05,
+      "loss": 20.542,
+      "step": 6431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249959393799115e-05,
+      "loss": 0.6894,
+      "step": 6432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249125363831713e-05,
+      "loss": 18.2966,
+      "step": 6433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24829125540194e-05,
+      "loss": 0.6873,
+      "step": 6434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.247457068538894e-05,
+      "loss": 19.0683,
+      "step": 6435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24662280327168e-05,
+      "loss": 18.964,
+      "step": 6436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.245788459629396e-05,
+      "loss": 18.0703,
+      "step": 6437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244954037641153e-05,
+      "loss": 22.934,
+      "step": 6438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24411953733606e-05,
+      "loss": 19.2963,
+      "step": 6439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243284958743227e-05,
+      "loss": 0.6917,
+      "step": 6440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.242450301891772e-05,
+      "loss": 0.6884,
+      "step": 6441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241615566810812e-05,
+      "loss": 18.7466,
+      "step": 6442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.240780753529463e-05,
+      "loss": 20.1045,
+      "step": 6443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239945862076853e-05,
+      "loss": 20.9317,
+      "step": 6444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239110892482106e-05,
+      "loss": 20.3965,
+      "step": 6445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.238275844774352e-05,
+      "loss": 0.7111,
+      "step": 6446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.23744071898272e-05,
+      "loss": 19.843,
+      "step": 6447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.236605515136343e-05,
+      "loss": 19.5332,
+      "step": 6448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.235770233264362e-05,
+      "loss": 0.6711,
+      "step": 6449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.234934873395912e-05,
+      "loss": 20.4751,
+      "step": 6450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.23409943556014e-05,
+      "loss": 19.3695,
+      "step": 6451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.233263919786186e-05,
+      "loss": 18.5362,
+      "step": 6452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2324283261032e-05,
+      "loss": 0.7207,
+      "step": 6453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231592654540331e-05,
+      "loss": 20.2244,
+      "step": 6454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.230756905126734e-05,
+      "loss": 0.7273,
+      "step": 6455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229921077891564e-05,
+      "loss": 18.4421,
+      "step": 6456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229085172863978e-05,
+      "loss": 0.7884,
+      "step": 6457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.228249190073138e-05,
+      "loss": 19.1075,
+      "step": 6458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227413129548209e-05,
+      "loss": 0.7134,
+      "step": 6459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.226576991318357e-05,
+      "loss": 20.7625,
+      "step": 6460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.225740775412751e-05,
+      "loss": 19.0595,
+      "step": 6461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.22490448186056e-05,
+      "loss": 20.4502,
+      "step": 6462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.224068110690964e-05,
+      "loss": 20.1915,
+      "step": 6463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.223231661933139e-05,
+      "loss": 0.7154,
+      "step": 6464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.222395135616261e-05,
+      "loss": 19.6359,
+      "step": 6465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.221558531769519e-05,
+      "loss": 19.1514,
+      "step": 6466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.220721850422092e-05,
+      "loss": 19.7599,
+      "step": 6467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.219885091603176e-05,
+      "loss": 0.709,
+      "step": 6468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.219048255341956e-05,
+      "loss": 0.6594,
+      "step": 6469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.218211341667627e-05,
+      "loss": 0.7417,
+      "step": 6470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.217374350609387e-05,
+      "loss": 22.4555,
+      "step": 6471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.216537282196431e-05,
+      "loss": 0.6505,
+      "step": 6472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.215700136457965e-05,
+      "loss": 18.2111,
+      "step": 6473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.214862913423192e-05,
+      "loss": 18.8481,
+      "step": 6474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.214025613121317e-05,
+      "loss": 0.697,
+      "step": 6475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.213188235581556e-05,
+      "loss": 0.6493,
+      "step": 6476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.212350780833115e-05,
+      "loss": 0.7428,
+      "step": 6477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.211513248905212e-05,
+      "loss": 18.3584,
+      "step": 6478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.210675639827064e-05,
+      "loss": 0.6267,
+      "step": 6479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209837953627891e-05,
+      "loss": 0.7127,
+      "step": 6480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.20900019033692e-05,
+      "loss": 0.7215,
+      "step": 6481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.208162349983372e-05,
+      "loss": 19.1809,
+      "step": 6482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.20732443259648e-05,
+      "loss": 19.5765,
+      "step": 6483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.206486438205474e-05,
+      "loss": 0.711,
+      "step": 6484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.205648366839584e-05,
+      "loss": 18.1616,
+      "step": 6485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.204810218528053e-05,
+      "loss": 18.3552,
+      "step": 6486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203971993300116e-05,
+      "loss": 19.8094,
+      "step": 6487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203133691185019e-05,
+      "loss": 19.9454,
+      "step": 6488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202295312212002e-05,
+      "loss": 0.7527,
+      "step": 6489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.201456856410315e-05,
+      "loss": 20.3422,
+      "step": 6490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.20061832380921e-05,
+      "loss": 18.1944,
+      "step": 6491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.199779714437936e-05,
+      "loss": 17.8109,
+      "step": 6492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.198941028325751e-05,
+      "loss": 19.6848,
+      "step": 6493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.198102265501912e-05,
+      "loss": 19.9586,
+      "step": 6494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.197263425995682e-05,
+      "loss": 17.849,
+      "step": 6495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196424509836323e-05,
+      "loss": 17.9801,
+      "step": 6496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1955855170531e-05,
+      "loss": 0.7238,
+      "step": 6497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.194746447675283e-05,
+      "loss": 21.0938,
+      "step": 6498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193907301732144e-05,
+      "loss": 0.7079,
+      "step": 6499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193068079252955e-05,
+      "loss": 21.659,
+      "step": 6500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.192228780266997e-05,
+      "loss": 19.5318,
+      "step": 6501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191389404803545e-05,
+      "loss": 18.6971,
+      "step": 6502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.190549952891885e-05,
+      "loss": 18.0437,
+      "step": 6503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1897104245613e-05,
+      "loss": 0.7556,
+      "step": 6504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188870819841078e-05,
+      "loss": 0.7019,
+      "step": 6505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188031138760508e-05,
+      "loss": 20.3842,
+      "step": 6506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.187191381348884e-05,
+      "loss": 18.6486,
+      "step": 6507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.186351547635504e-05,
+      "loss": 18.3614,
+      "step": 6508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.185511637649661e-05,
+      "loss": 20.2405,
+      "step": 6509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.184671651420659e-05,
+      "loss": 17.1886,
+      "step": 6510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.183831588977803e-05,
+      "loss": 17.5375,
+      "step": 6511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.182991450350395e-05,
+      "loss": 0.7343,
+      "step": 6512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.182151235567747e-05,
+      "loss": 18.0612,
+      "step": 6513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.18131094465917e-05,
+      "loss": 0.7175,
+      "step": 6514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.180470577653978e-05,
+      "loss": 19.6082,
+      "step": 6515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.179630134581488e-05,
+      "loss": 0.7386,
+      "step": 6516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.178789615471018e-05,
+      "loss": 19.8181,
+      "step": 6517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.177949020351891e-05,
+      "loss": 20.6367,
+      "step": 6518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.177108349253433e-05,
+      "loss": 19.1885,
+      "step": 6519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.17626760220497e-05,
+      "loss": 19.4855,
+      "step": 6520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.175426779235832e-05,
+      "loss": 16.9804,
+      "step": 6521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174585880375352e-05,
+      "loss": 0.7313,
+      "step": 6522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173744905652866e-05,
+      "loss": 18.2571,
+      "step": 6523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172903855097711e-05,
+      "loss": 19.1633,
+      "step": 6524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172062728739229e-05,
+      "loss": 16.6255,
+      "step": 6525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.171221526606763e-05,
+      "loss": 19.0871,
+      "step": 6526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.170380248729657e-05,
+      "loss": 0.6706,
+      "step": 6527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169538895137263e-05,
+      "loss": 19.69,
+      "step": 6528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.168697465858927e-05,
+      "loss": 0.6906,
+      "step": 6529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.167855960924009e-05,
+      "loss": 17.3693,
+      "step": 6530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.167014380361863e-05,
+      "loss": 0.6894,
+      "step": 6531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.166172724201847e-05,
+      "loss": 19.4676,
+      "step": 6532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.165330992473323e-05,
+      "loss": 21.1613,
+      "step": 6533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164489185205656e-05,
+      "loss": 19.2983,
+      "step": 6534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.163647302428214e-05,
+      "loss": 18.9788,
+      "step": 6535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.162805344170364e-05,
+      "loss": 17.1636,
+      "step": 6536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.161963310461481e-05,
+      "loss": 0.6898,
+      "step": 6537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.16112120133094e-05,
+      "loss": 19.6583,
+      "step": 6538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.160279016808118e-05,
+      "loss": 18.8964,
+      "step": 6539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.159436756922394e-05,
+      "loss": 19.9757,
+      "step": 6540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158594421703152e-05,
+      "loss": 19.781,
+      "step": 6541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.157752011179777e-05,
+      "loss": 0.7477,
+      "step": 6542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.156909525381657e-05,
+      "loss": 0.7329,
+      "step": 6543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.156066964338184e-05,
+      "loss": 19.8265,
+      "step": 6544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.155224328078749e-05,
+      "loss": 19.3684,
+      "step": 6545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.15438161663275e-05,
+      "loss": 0.7472,
+      "step": 6546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.153538830029585e-05,
+      "loss": 0.7916,
+      "step": 6547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.152695968298655e-05,
+      "loss": 19.3241,
+      "step": 6548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.151853031469365e-05,
+      "loss": 17.9719,
+      "step": 6549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.15101001957112e-05,
+      "loss": 20.9851,
+      "step": 6550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.150166932633328e-05,
+      "loss": 0.7436,
+      "step": 6551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149323770685404e-05,
+      "loss": 20.4663,
+      "step": 6552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14848053375676e-05,
+      "loss": 18.768,
+      "step": 6553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.147637221876813e-05,
+      "loss": 20.0905,
+      "step": 6554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.146793835074983e-05,
+      "loss": 19.4222,
+      "step": 6555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.145950373380693e-05,
+      "loss": 0.7499,
+      "step": 6556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.145106836823367e-05,
+      "loss": 0.7084,
+      "step": 6557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144263225432433e-05,
+      "loss": 17.5843,
+      "step": 6558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143419539237319e-05,
+      "loss": 19.1887,
+      "step": 6559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14257577826746e-05,
+      "loss": 19.412,
+      "step": 6560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.141731942552288e-05,
+      "loss": 18.0719,
+      "step": 6561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.140888032121246e-05,
+      "loss": 17.4743,
+      "step": 6562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14004404700377e-05,
+      "loss": 19.567,
+      "step": 6563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.139199987229304e-05,
+      "loss": 0.6471,
+      "step": 6564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.138355852827295e-05,
+      "loss": 0.6953,
+      "step": 6565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13751164382719e-05,
+      "loss": 0.6929,
+      "step": 6566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13666736025844e-05,
+      "loss": 18.6396,
+      "step": 6567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1358230021505e-05,
+      "loss": 18.4431,
+      "step": 6568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134978569532823e-05,
+      "loss": 0.6659,
+      "step": 6569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134134062434871e-05,
+      "loss": 0.7762,
+      "step": 6570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.133289480886101e-05,
+      "loss": 18.2518,
+      "step": 6571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.132444824915982e-05,
+      "loss": 17.5054,
+      "step": 6572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.131600094553977e-05,
+      "loss": 19.4196,
+      "step": 6573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.130755289829556e-05,
+      "loss": 19.0091,
+      "step": 6574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.12991041077219e-05,
+      "loss": 20.1504,
+      "step": 6575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.129065457411353e-05,
+      "loss": 20.873,
+      "step": 6576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.128220429776524e-05,
+      "loss": 19.6979,
+      "step": 6577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.12737532789718e-05,
+      "loss": 0.7053,
+      "step": 6578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.126530151802805e-05,
+      "loss": 0.7032,
+      "step": 6579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.125684901522882e-05,
+      "loss": 17.4755,
+      "step": 6580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.124839577086898e-05,
+      "loss": 18.5201,
+      "step": 6581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123994178524345e-05,
+      "loss": 19.9395,
+      "step": 6582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123148705864711e-05,
+      "loss": 18.6882,
+      "step": 6583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.122303159137496e-05,
+      "loss": 20.0781,
+      "step": 6584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121457538372196e-05,
+      "loss": 0.7067,
+      "step": 6585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.120611843598308e-05,
+      "loss": 0.7125,
+      "step": 6586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119766074845338e-05,
+      "loss": 17.6655,
+      "step": 6587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.11892023214279e-05,
+      "loss": 17.8466,
+      "step": 6588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.118074315520172e-05,
+      "loss": 0.6705,
+      "step": 6589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117228325006995e-05,
+      "loss": 22.4689,
+      "step": 6590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.116382260632771e-05,
+      "loss": 20.8361,
+      "step": 6591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.115536122427017e-05,
+      "loss": 0.6758,
+      "step": 6592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.11468991041925e-05,
+      "loss": 18.8648,
+      "step": 6593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113843624638989e-05,
+      "loss": 0.6946,
+      "step": 6594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.112997265115762e-05,
+      "loss": 19.5137,
+      "step": 6595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.112150831879091e-05,
+      "loss": 21.0286,
+      "step": 6596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.111304324958506e-05,
+      "loss": 16.4783,
+      "step": 6597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.110457744383536e-05,
+      "loss": 19.9995,
+      "step": 6598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109611090183718e-05,
+      "loss": 22.2685,
+      "step": 6599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.108764362388586e-05,
+      "loss": 20.324,
+      "step": 6600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107917561027677e-05,
+      "loss": 0.6907,
+      "step": 6601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107070686130536e-05,
+      "loss": 0.7201,
+      "step": 6602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106223737726703e-05,
+      "loss": 0.7024,
+      "step": 6603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.10537671584573e-05,
+      "loss": 0.678,
+      "step": 6604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.104529620517159e-05,
+      "loss": 18.1211,
+      "step": 6605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103682451770547e-05,
+      "loss": 19.4603,
+      "step": 6606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.102835209635447e-05,
+      "loss": 20.5984,
+      "step": 6607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101987894141412e-05,
+      "loss": 19.0981,
+      "step": 6608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101140505318006e-05,
+      "loss": 0.7271,
+      "step": 6609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.100293043194789e-05,
+      "loss": 19.9625,
+      "step": 6610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.099445507801323e-05,
+      "loss": 0.6577,
+      "step": 6611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.098597899167179e-05,
+      "loss": 20.1927,
+      "step": 6612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.097750217321924e-05,
+      "loss": 19.4187,
+      "step": 6613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096902462295129e-05,
+      "loss": 19.1323,
+      "step": 6614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096054634116371e-05,
+      "loss": 0.7212,
+      "step": 6615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.095206732815226e-05,
+      "loss": 0.702,
+      "step": 6616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.094358758421275e-05,
+      "loss": 19.2249,
+      "step": 6617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.093510710964096e-05,
+      "loss": 18.4958,
+      "step": 6618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.09266259047328e-05,
+      "loss": 18.699,
+      "step": 6619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.09181439697841e-05,
+      "loss": 0.6741,
+      "step": 6620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.090966130509075e-05,
+      "loss": 0.7405,
+      "step": 6621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.09011779109487e-05,
+      "loss": 0.769,
+      "step": 6622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.08926937876539e-05,
+      "loss": 19.2795,
+      "step": 6623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.088420893550232e-05,
+      "loss": 0.7646,
+      "step": 6624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.087572335478994e-05,
+      "loss": 18.8734,
+      "step": 6625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086723704581281e-05,
+      "loss": 19.791,
+      "step": 6626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.085875000886697e-05,
+      "loss": 18.4329,
+      "step": 6627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.085026224424851e-05,
+      "loss": 0.7939,
+      "step": 6628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084177375225352e-05,
+      "loss": 0.7314,
+      "step": 6629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.083328453317811e-05,
+      "loss": 19.7455,
+      "step": 6630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.082479458731848e-05,
+      "loss": 19.1279,
+      "step": 6631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081630391497077e-05,
+      "loss": 0.7611,
+      "step": 6632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.080781251643117e-05,
+      "loss": 0.6762,
+      "step": 6633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079932039199594e-05,
+      "loss": 0.7327,
+      "step": 6634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079082754196134e-05,
+      "loss": 18.4921,
+      "step": 6635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.078233396662363e-05,
+      "loss": 18.3123,
+      "step": 6636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07738396662791e-05,
+      "loss": 19.4094,
+      "step": 6637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.076534464122412e-05,
+      "loss": 20.8654,
+      "step": 6638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075684889175502e-05,
+      "loss": 18.581,
+      "step": 6639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.074835241816817e-05,
+      "loss": 20.3818,
+      "step": 6640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.073985522076002e-05,
+      "loss": 0.6963,
+      "step": 6641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.073135729982694e-05,
+      "loss": 19.1914,
+      "step": 6642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.072285865566543e-05,
+      "loss": 0.743,
+      "step": 6643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.071435928857196e-05,
+      "loss": 0.66,
+      "step": 6644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.070585919884302e-05,
+      "loss": 0.7138,
+      "step": 6645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.069735838677516e-05,
+      "loss": 18.9832,
+      "step": 6646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.068885685266495e-05,
+      "loss": 0.6986,
+      "step": 6647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.068035459680894e-05,
+      "loss": 0.7374,
+      "step": 6648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.067185161950376e-05,
+      "loss": 18.8535,
+      "step": 6649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.066334792104603e-05,
+      "loss": 0.6562,
+      "step": 6650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.065484350173241e-05,
+      "loss": 17.688,
+      "step": 6651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.064633836185961e-05,
+      "loss": 18.587,
+      "step": 6652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.063783250172429e-05,
+      "loss": 18.112,
+      "step": 6653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062932592162322e-05,
+      "loss": 18.8016,
+      "step": 6654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062081862185314e-05,
+      "loss": 19.6384,
+      "step": 6655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.061231060271084e-05,
+      "loss": 0.6773,
+      "step": 6656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.060380186449311e-05,
+      "loss": 0.7412,
+      "step": 6657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.059529240749683e-05,
+      "loss": 0.7058,
+      "step": 6658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.058678223201881e-05,
+      "loss": 19.3969,
+      "step": 6659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057827133835595e-05,
+      "loss": 21.2724,
+      "step": 6660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.056975972680517e-05,
+      "loss": 0.6668,
+      "step": 6661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05612473976634e-05,
+      "loss": 18.7193,
+      "step": 6662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055273435122758e-05,
+      "loss": 0.7306,
+      "step": 6663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.054422058779473e-05,
+      "loss": 18.7307,
+      "step": 6664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.053570610766182e-05,
+      "loss": 0.6544,
+      "step": 6665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05271909111259e-05,
+      "loss": 17.8479,
+      "step": 6666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.051867499848402e-05,
+      "loss": 0.6896,
+      "step": 6667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05101583700333e-05,
+      "loss": 0.6834,
+      "step": 6668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05016410260708e-05,
+      "loss": 18.3238,
+      "step": 6669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.049312296689368e-05,
+      "loss": 0.7157,
+      "step": 6670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.04846041927991e-05,
+      "loss": 18.3455,
+      "step": 6671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.047608470408424e-05,
+      "loss": 19.5801,
+      "step": 6672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.04675645010463e-05,
+      "loss": 19.9573,
+      "step": 6673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.045904358398253e-05,
+      "loss": 19.5688,
+      "step": 6674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.045052195319016e-05,
+      "loss": 18.113,
+      "step": 6675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.044199960896651e-05,
+      "loss": 0.6785,
+      "step": 6676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.043347655160885e-05,
+      "loss": 0.7829,
+      "step": 6677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.042495278141454e-05,
+      "loss": 0.6974,
+      "step": 6678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.041642829868094e-05,
+      "loss": 0.6982,
+      "step": 6679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.040790310370543e-05,
+      "loss": 0.7197,
+      "step": 6680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.03993771967854e-05,
+      "loss": 19.5695,
+      "step": 6681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.039085057821829e-05,
+      "loss": 0.6611,
+      "step": 6682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.038232324830156e-05,
+      "loss": 0.6944,
+      "step": 6683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037379520733269e-05,
+      "loss": 18.3575,
+      "step": 6684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.03652664556092e-05,
+      "loss": 17.7279,
+      "step": 6685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.035673699342862e-05,
+      "loss": 0.7046,
+      "step": 6686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.034820682108847e-05,
+      "loss": 0.7723,
+      "step": 6687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033967593888638e-05,
+      "loss": 19.2047,
+      "step": 6688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033114434711994e-05,
+      "loss": 0.7613,
+      "step": 6689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.032261204608675e-05,
+      "loss": 0.6607,
+      "step": 6690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.031407903608451e-05,
+      "loss": 19.1981,
+      "step": 6691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.030554531741087e-05,
+      "loss": 17.529,
+      "step": 6692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029701089036357e-05,
+      "loss": 0.6562,
+      "step": 6693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.02884757552403e-05,
+      "loss": 16.7985,
+      "step": 6694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027993991233883e-05,
+      "loss": 18.5764,
+      "step": 6695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027140336195695e-05,
+      "loss": 19.4223,
+      "step": 6696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.026286610439243e-05,
+      "loss": 0.7394,
+      "step": 6697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.025432813994315e-05,
+      "loss": 21.2313,
+      "step": 6698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.024578946890692e-05,
+      "loss": 19.0829,
+      "step": 6699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023725009158162e-05,
+      "loss": 20.0687,
+      "step": 6700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022871000826519e-05,
+      "loss": 21.2338,
+      "step": 6701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022016921925551e-05,
+      "loss": 17.1574,
+      "step": 6702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.021162772485058e-05,
+      "loss": 19.1681,
+      "step": 6703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.020308552534834e-05,
+      "loss": 0.668,
+      "step": 6704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.01945426210468e-05,
+      "loss": 18.7323,
+      "step": 6705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.018599901224399e-05,
+      "loss": 0.7092,
+      "step": 6706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.017745469923794e-05,
+      "loss": 0.6917,
+      "step": 6707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016890968232675e-05,
+      "loss": 20.4558,
+      "step": 6708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016036396180852e-05,
+      "loss": 20.8678,
+      "step": 6709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.015181753798135e-05,
+      "loss": 0.731,
+      "step": 6710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.014327041114341e-05,
+      "loss": 0.6929,
+      "step": 6711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.013472258159286e-05,
+      "loss": 0.6819,
+      "step": 6712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.01261740496279e-05,
+      "loss": 20.0857,
+      "step": 6713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.011762481554675e-05,
+      "loss": 0.7345,
+      "step": 6714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010907487964766e-05,
+      "loss": 18.8842,
+      "step": 6715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010052424222892e-05,
+      "loss": 20.8649,
+      "step": 6716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009197290358879e-05,
+      "loss": 20.0005,
+      "step": 6717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.008342086402561e-05,
+      "loss": 20.5409,
+      "step": 6718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00748681238377e-05,
+      "loss": 20.1476,
+      "step": 6719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.006631468332347e-05,
+      "loss": 0.7769,
+      "step": 6720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00577605427813e-05,
+      "loss": 18.8053,
+      "step": 6721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004920570250957e-05,
+      "loss": 0.6566,
+      "step": 6722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004065016280677e-05,
+      "loss": 0.7187,
+      "step": 6723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.003209392397133e-05,
+      "loss": 18.0361,
+      "step": 6724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002353698630177e-05,
+      "loss": 0.7174,
+      "step": 6725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00149793500966e-05,
+      "loss": 0.6841,
+      "step": 6726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.000642101565434e-05,
+      "loss": 18.7226,
+      "step": 6727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.999786198327356e-05,
+      "loss": 0.7354,
+      "step": 6728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.998930225325286e-05,
+      "loss": 0.7282,
+      "step": 6729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.998074182589084e-05,
+      "loss": 18.9562,
+      "step": 6730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997218070148615e-05,
+      "loss": 19.7351,
+      "step": 6731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.996361888033743e-05,
+      "loss": 19.0347,
+      "step": 6732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99550563627434e-05,
+      "loss": 18.0239,
+      "step": 6733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.994649314900274e-05,
+      "loss": 18.3513,
+      "step": 6734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993792923941418e-05,
+      "loss": 0.7155,
+      "step": 6735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99293646342765e-05,
+      "loss": 20.7381,
+      "step": 6736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.992079933388848e-05,
+      "loss": 0.6957,
+      "step": 6737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99122333385489e-05,
+      "loss": 0.685,
+      "step": 6738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.990366664855661e-05,
+      "loss": 0.6152,
+      "step": 6739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.989509926421049e-05,
+      "loss": 0.7179,
+      "step": 6740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.988653118580938e-05,
+      "loss": 0.7369,
+      "step": 6741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.987796241365221e-05,
+      "loss": 0.7162,
+      "step": 6742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986939294803788e-05,
+      "loss": 0.6234,
+      "step": 6743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986082278926536e-05,
+      "loss": 0.6992,
+      "step": 6744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.985225193763365e-05,
+      "loss": 20.0814,
+      "step": 6745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.984368039344169e-05,
+      "loss": 0.6997,
+      "step": 6746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983510815698856e-05,
+      "loss": 17.2175,
+      "step": 6747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.982653522857329e-05,
+      "loss": 19.4556,
+      "step": 6748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.981796160849492e-05,
+      "loss": 18.9224,
+      "step": 6749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.980938729705262e-05,
+      "loss": 0.7414,
+      "step": 6750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.980081229454545e-05,
+      "loss": 20.606,
+      "step": 6751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979223660127258e-05,
+      "loss": 19.3471,
+      "step": 6752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.978366021753317e-05,
+      "loss": 18.5391,
+      "step": 6753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.977508314362641e-05,
+      "loss": 20.2508,
+      "step": 6754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.976650537985154e-05,
+      "loss": 0.6304,
+      "step": 6755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.975792692650777e-05,
+      "loss": 0.6838,
+      "step": 6756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974934778389439e-05,
+      "loss": 20.2781,
+      "step": 6757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974076795231069e-05,
+      "loss": 0.7836,
+      "step": 6758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.973218743205596e-05,
+      "loss": 0.7007,
+      "step": 6759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.972360622342957e-05,
+      "loss": 0.7147,
+      "step": 6760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971502432673085e-05,
+      "loss": 18.8255,
+      "step": 6761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.97064417422592e-05,
+      "loss": 17.6788,
+      "step": 6762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969785847031402e-05,
+      "loss": 0.7626,
+      "step": 6763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.968927451119478e-05,
+      "loss": 20.5382,
+      "step": 6764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.96806898652009e-05,
+      "loss": 0.7152,
+      "step": 6765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967210453263187e-05,
+      "loss": 0.733,
+      "step": 6766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.966351851378719e-05,
+      "loss": 19.8341,
+      "step": 6767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.96549318089664e-05,
+      "loss": 19.9353,
+      "step": 6768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.964634441846906e-05,
+      "loss": 18.9901,
+      "step": 6769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.963775634259474e-05,
+      "loss": 0.7082,
+      "step": 6770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962916758164303e-05,
+      "loss": 0.7076,
+      "step": 6771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962057813591357e-05,
+      "loss": 18.3646,
+      "step": 6772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9611988005706e-05,
+      "loss": 20.2237,
+      "step": 6773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.960339719132e-05,
+      "loss": 0.7324,
+      "step": 6774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.959480569305527e-05,
+      "loss": 19.4269,
+      "step": 6775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.958621351121151e-05,
+      "loss": 19.9648,
+      "step": 6776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.957762064608848e-05,
+      "loss": 0.7464,
+      "step": 6777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.956902709798596e-05,
+      "loss": 0.716,
+      "step": 6778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.956043286720372e-05,
+      "loss": 0.6991,
+      "step": 6779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955183795404159e-05,
+      "loss": 0.706,
+      "step": 6780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.954324235879939e-05,
+      "loss": 18.4746,
+      "step": 6781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953464608177702e-05,
+      "loss": 19.5079,
+      "step": 6782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.952604912327433e-05,
+      "loss": 20.1251,
+      "step": 6783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.951745148359125e-05,
+      "loss": 0.6915,
+      "step": 6784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.950885316302773e-05,
+      "loss": 0.7051,
+      "step": 6785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.950025416188369e-05,
+      "loss": 17.9025,
+      "step": 6786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.949165448045914e-05,
+      "loss": 19.2062,
+      "step": 6787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.948305411905409e-05,
+      "loss": 0.7017,
+      "step": 6788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947445307796855e-05,
+      "loss": 0.696,
+      "step": 6789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.946585135750259e-05,
+      "loss": 17.9988,
+      "step": 6790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.945724895795626e-05,
+      "loss": 0.7129,
+      "step": 6791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.944864587962969e-05,
+      "loss": 0.7263,
+      "step": 6792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9440042122823e-05,
+      "loss": 19.3905,
+      "step": 6793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.943143768783633e-05,
+      "loss": 18.4977,
+      "step": 6794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.942283257496985e-05,
+      "loss": 21.4558,
+      "step": 6795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.941422678452376e-05,
+      "loss": 18.2624,
+      "step": 6796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.94056203167983e-05,
+      "loss": 0.6738,
+      "step": 6797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939701317209366e-05,
+      "loss": 20.228,
+      "step": 6798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.938840535071017e-05,
+      "loss": 18.7358,
+      "step": 6799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.937979685294809e-05,
+      "loss": 19.4971,
+      "step": 6800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.937118767910771e-05,
+      "loss": 0.6351,
+      "step": 6801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.936257782948941e-05,
+      "loss": 19.3799,
+      "step": 6802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935396730439352e-05,
+      "loss": 0.6792,
+      "step": 6803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.934535610412046e-05,
+      "loss": 20.5526,
+      "step": 6804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933674422897058e-05,
+      "loss": 17.6384,
+      "step": 6805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.932813167924437e-05,
+      "loss": 0.6953,
+      "step": 6806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.931951845524224e-05,
+      "loss": 20.0485,
+      "step": 6807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.93109045572647e-05,
+      "loss": 18.9559,
+      "step": 6808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.930228998561226e-05,
+      "loss": 0.6774,
+      "step": 6809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.929367474058541e-05,
+      "loss": 18.7106,
+      "step": 6810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.928505882248471e-05,
+      "loss": 20.6817,
+      "step": 6811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927644223161074e-05,
+      "loss": 19.8269,
+      "step": 6812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92678249682641e-05,
+      "loss": 19.2911,
+      "step": 6813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.925920703274541e-05,
+      "loss": 18.9649,
+      "step": 6814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92505884253553e-05,
+      "loss": 19.3246,
+      "step": 6815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.924196914639445e-05,
+      "loss": 0.7765,
+      "step": 6816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.923334919616354e-05,
+      "loss": 0.6949,
+      "step": 6817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.922472857496327e-05,
+      "loss": 18.1596,
+      "step": 6818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.921610728309442e-05,
+      "loss": 0.7352,
+      "step": 6819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92074853208577e-05,
+      "loss": 0.7546,
+      "step": 6820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.919886268855393e-05,
+      "loss": 18.9068,
+      "step": 6821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.91902393864839e-05,
+      "loss": 19.9398,
+      "step": 6822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918161541494844e-05,
+      "loss": 0.7331,
+      "step": 6823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.917299077424842e-05,
+      "loss": 0.6727,
+      "step": 6824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916436546468468e-05,
+      "loss": 18.6296,
+      "step": 6825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.915573948655817e-05,
+      "loss": 18.449,
+      "step": 6826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914711284016976e-05,
+      "loss": 19.0261,
+      "step": 6827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.913848552582045e-05,
+      "loss": 18.8759,
+      "step": 6828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.912985754381117e-05,
+      "loss": 0.7054,
+      "step": 6829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.912122889444293e-05,
+      "loss": 17.212,
+      "step": 6830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911259957801675e-05,
+      "loss": 18.7786,
+      "step": 6831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910396959483367e-05,
+      "loss": 0.7606,
+      "step": 6832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.909533894519473e-05,
+      "loss": 0.6972,
+      "step": 6833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.908670762940103e-05,
+      "loss": 0.7593,
+      "step": 6834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.90780756477537e-05,
+      "loss": 20.43,
+      "step": 6835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906944300055386e-05,
+      "loss": 19.3859,
+      "step": 6836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906080968810265e-05,
+      "loss": 0.7676,
+      "step": 6837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.905217571070127e-05,
+      "loss": 0.6812,
+      "step": 6838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904354106865093e-05,
+      "loss": 18.6841,
+      "step": 6839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.90349057622528e-05,
+      "loss": 20.8214,
+      "step": 6840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.902626979180821e-05,
+      "loss": 0.7579,
+      "step": 6841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.901763315761838e-05,
+      "loss": 17.7198,
+      "step": 6842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.90089958599846e-05,
+      "loss": 0.758,
+      "step": 6843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.900035789920822e-05,
+      "loss": 0.6999,
+      "step": 6844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.899171927559058e-05,
+      "loss": 20.1049,
+      "step": 6845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.8983079989433e-05,
+      "loss": 18.9839,
+      "step": 6846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.897444004103691e-05,
+      "loss": 20.4318,
+      "step": 6847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.896579943070372e-05,
+      "loss": 18.5779,
+      "step": 6848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.895715815873486e-05,
+      "loss": 0.7107,
+      "step": 6849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.894851622543176e-05,
+      "loss": 18.2966,
+      "step": 6850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893987363109594e-05,
+      "loss": 17.3592,
+      "step": 6851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893123037602886e-05,
+      "loss": 19.2154,
+      "step": 6852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.89225864605321e-05,
+      "loss": 21.2177,
+      "step": 6853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.891394188490717e-05,
+      "loss": 19.4437,
+      "step": 6854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.890529664945564e-05,
+      "loss": 19.3518,
+      "step": 6855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.889665075447912e-05,
+      "loss": 19.8027,
+      "step": 6856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888800420027924e-05,
+      "loss": 17.5308,
+      "step": 6857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.887935698715762e-05,
+      "loss": 17.8813,
+      "step": 6858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.887070911541595e-05,
+      "loss": 19.6027,
+      "step": 6859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.886206058535587e-05,
+      "loss": 17.7918,
+      "step": 6860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.885341139727912e-05,
+      "loss": 18.6423,
+      "step": 6861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884476155148745e-05,
+      "loss": 16.6318,
+      "step": 6862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.883611104828258e-05,
+      "loss": 19.2264,
+      "step": 6863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.882745988796633e-05,
+      "loss": 21.314,
+      "step": 6864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881880807084046e-05,
+      "loss": 18.6439,
+      "step": 6865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881015559720683e-05,
+      "loss": 0.6447,
+      "step": 6866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.880150246736725e-05,
+      "loss": 19.5018,
+      "step": 6867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.879284868162364e-05,
+      "loss": 0.6725,
+      "step": 6868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.878419424027784e-05,
+      "loss": 18.2344,
+      "step": 6869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877553914363182e-05,
+      "loss": 18.2137,
+      "step": 6870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.876688339198748e-05,
+      "loss": 0.7908,
+      "step": 6871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.875822698564679e-05,
+      "loss": 19.0762,
+      "step": 6872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.874956992491174e-05,
+      "loss": 0.6856,
+      "step": 6873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.874091221008433e-05,
+      "loss": 0.6742,
+      "step": 6874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.873225384146662e-05,
+      "loss": 0.743,
+      "step": 6875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.872359481936062e-05,
+      "loss": 0.7535,
+      "step": 6876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.871493514406844e-05,
+      "loss": 0.6907,
+      "step": 6877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.870627481589216e-05,
+      "loss": 17.9451,
+      "step": 6878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.869761383513392e-05,
+      "loss": 20.2006,
+      "step": 6879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868895220209584e-05,
+      "loss": 16.4049,
+      "step": 6880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868028991708012e-05,
+      "loss": 0.7446,
+      "step": 6881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.867162698038889e-05,
+      "loss": 19.7425,
+      "step": 6882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.866296339232444e-05,
+      "loss": 18.0185,
+      "step": 6883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.865429915318897e-05,
+      "loss": 21.4885,
+      "step": 6884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864563426328472e-05,
+      "loss": 0.7087,
+      "step": 6885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.863696872291399e-05,
+      "loss": 19.7853,
+      "step": 6886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.862830253237908e-05,
+      "loss": 19.0429,
+      "step": 6887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.861963569198233e-05,
+      "loss": 19.0921,
+      "step": 6888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.861096820202609e-05,
+      "loss": 19.6669,
+      "step": 6889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.860230006281268e-05,
+      "loss": 19.7514,
+      "step": 6890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859363127464456e-05,
+      "loss": 0.7456,
+      "step": 6891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.858496183782411e-05,
+      "loss": 19.9555,
+      "step": 6892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.857629175265379e-05,
+      "loss": 0.7553,
+      "step": 6893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856762101943605e-05,
+      "loss": 18.914,
+      "step": 6894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855894963847337e-05,
+      "loss": 16.5779,
+      "step": 6895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855027761006827e-05,
+      "loss": 0.6859,
+      "step": 6896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.854160493452328e-05,
+      "loss": 18.3769,
+      "step": 6897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.853293161214093e-05,
+      "loss": 0.6729,
+      "step": 6898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.852425764322384e-05,
+      "loss": 18.8505,
+      "step": 6899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.851558302807455e-05,
+      "loss": 19.008,
+      "step": 6900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850690776699573e-05,
+      "loss": 0.719,
+      "step": 6901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.849823186028998e-05,
+      "loss": 0.7209,
+      "step": 6902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848955530826002e-05,
+      "loss": 19.8181,
+      "step": 6903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848087811120848e-05,
+      "loss": 0.671,
+      "step": 6904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.847220026943811e-05,
+      "loss": 20.0233,
+      "step": 6905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.846352178325163e-05,
+      "loss": 18.2135,
+      "step": 6906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.845484265295178e-05,
+      "loss": 17.5636,
+      "step": 6907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.844616287884136e-05,
+      "loss": 0.7131,
+      "step": 6908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.843748246122316e-05,
+      "loss": 20.2767,
+      "step": 6909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.842880140040001e-05,
+      "loss": 0.6684,
+      "step": 6910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.842011969667474e-05,
+      "loss": 18.4697,
+      "step": 6911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841143735035024e-05,
+      "loss": 19.8868,
+      "step": 6912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.840275436172939e-05,
+      "loss": 20.9796,
+      "step": 6913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839407073111508e-05,
+      "loss": 19.2644,
+      "step": 6914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.838538645881028e-05,
+      "loss": 17.7761,
+      "step": 6915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.837670154511793e-05,
+      "loss": 18.1378,
+      "step": 6916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.836801599034102e-05,
+      "loss": 0.6624,
+      "step": 6917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.835932979478252e-05,
+      "loss": 0.6858,
+      "step": 6918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.83506429587455e-05,
+      "loss": 18.926,
+      "step": 6919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.834195548253297e-05,
+      "loss": 18.8655,
+      "step": 6920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833326736644803e-05,
+      "loss": 20.5172,
+      "step": 6921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.832457861079374e-05,
+      "loss": 19.6471,
+      "step": 6922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.831588921587323e-05,
+      "loss": 0.6895,
+      "step": 6923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830719918198962e-05,
+      "loss": 22.2147,
+      "step": 6924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.829850850944611e-05,
+      "loss": 0.6712,
+      "step": 6925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.828981719854584e-05,
+      "loss": 20.1548,
+      "step": 6926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.828112524959202e-05,
+      "loss": 19.0786,
+      "step": 6927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.827243266288788e-05,
+      "loss": 21.1852,
+      "step": 6928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.826373943873667e-05,
+      "loss": 0.7531,
+      "step": 6929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.825504557744167e-05,
+      "loss": 0.7479,
+      "step": 6930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.824635107930613e-05,
+      "loss": 20.6465,
+      "step": 6931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82376559446334e-05,
+      "loss": 0.6858,
+      "step": 6932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.822896017372682e-05,
+      "loss": 19.3166,
+      "step": 6933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.822026376688973e-05,
+      "loss": 0.7083,
+      "step": 6934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.821156672442553e-05,
+      "loss": 0.7258,
+      "step": 6935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.820286904663757e-05,
+      "loss": 0.6703,
+      "step": 6936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819417073382936e-05,
+      "loss": 0.6921,
+      "step": 6937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.818547178630428e-05,
+      "loss": 0.7406,
+      "step": 6938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.817677220436582e-05,
+      "loss": 0.7078,
+      "step": 6939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.816807198831748e-05,
+      "loss": 0.6468,
+      "step": 6940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815937113846275e-05,
+      "loss": 18.3232,
+      "step": 6941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815066965510519e-05,
+      "loss": 0.6874,
+      "step": 6942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.814196753854834e-05,
+      "loss": 0.6922,
+      "step": 6943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.81332647890958e-05,
+      "loss": 18.0533,
+      "step": 6944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.812456140705115e-05,
+      "loss": 0.7026,
+      "step": 6945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811585739271802e-05,
+      "loss": 19.4474,
+      "step": 6946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.810715274640007e-05,
+      "loss": 20.1928,
+      "step": 6947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809844746840095e-05,
+      "loss": 17.9579,
+      "step": 6948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.808974155902436e-05,
+      "loss": 19.5031,
+      "step": 6949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.808103501857402e-05,
+      "loss": 19.3317,
+      "step": 6950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.807232784735362e-05,
+      "loss": 0.7343,
+      "step": 6951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8063620045667e-05,
+      "loss": 20.0346,
+      "step": 6952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.805491161381782e-05,
+      "loss": 19.0743,
+      "step": 6953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.804620255211e-05,
+      "loss": 19.7923,
+      "step": 6954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80374928608473e-05,
+      "loss": 17.8783,
+      "step": 6955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.802878254033355e-05,
+      "loss": 20.9997,
+      "step": 6956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.802007159087266e-05,
+      "loss": 0.682,
+      "step": 6957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.801136001276848e-05,
+      "loss": 19.4427,
+      "step": 6958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.800264780632494e-05,
+      "loss": 18.7969,
+      "step": 6959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.799393497184598e-05,
+      "loss": 17.8084,
+      "step": 6960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.798522150963552e-05,
+      "loss": 18.4377,
+      "step": 6961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.797650741999757e-05,
+      "loss": 0.6689,
+      "step": 6962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796779270323611e-05,
+      "loss": 19.2323,
+      "step": 6963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795907735965517e-05,
+      "loss": 20.153,
+      "step": 6964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795036138955878e-05,
+      "loss": 18.4969,
+      "step": 6965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7941644793251e-05,
+      "loss": 19.2131,
+      "step": 6966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.793292757103593e-05,
+      "loss": 17.9414,
+      "step": 6967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792420972321767e-05,
+      "loss": 18.7742,
+      "step": 6968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.791549125010033e-05,
+      "loss": 19.9534,
+      "step": 6969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.790677215198811e-05,
+      "loss": 18.9683,
+      "step": 6970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.789805242918512e-05,
+      "loss": 18.6528,
+      "step": 6971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78893320819956e-05,
+      "loss": 0.7331,
+      "step": 6972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.788061111072375e-05,
+      "loss": 0.6974,
+      "step": 6973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.787188951567378e-05,
+      "loss": 20.5456,
+      "step": 6974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.786316729715e-05,
+      "loss": 19.0704,
+      "step": 6975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.785444445545666e-05,
+      "loss": 0.7141,
+      "step": 6976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.784572099089806e-05,
+      "loss": 20.3526,
+      "step": 6977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783699690377855e-05,
+      "loss": 19.358,
+      "step": 6978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.782827219440245e-05,
+      "loss": 18.5496,
+      "step": 6979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781954686307415e-05,
+      "loss": 0.706,
+      "step": 6980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7810820910098e-05,
+      "loss": 0.6557,
+      "step": 6981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.780209433577846e-05,
+      "loss": 18.9924,
+      "step": 6982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.779336714041993e-05,
+      "loss": 18.3895,
+      "step": 6983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778463932432688e-05,
+      "loss": 18.6604,
+      "step": 6984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.777591088780379e-05,
+      "loss": 0.71,
+      "step": 6985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776718183115513e-05,
+      "loss": 17.6846,
+      "step": 6986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.775845215468545e-05,
+      "loss": 19.4334,
+      "step": 6987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774972185869927e-05,
+      "loss": 18.3915,
+      "step": 6988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774099094350118e-05,
+      "loss": 18.7737,
+      "step": 6989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.773225940939573e-05,
+      "loss": 18.2261,
+      "step": 6990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772352725668756e-05,
+      "loss": 19.9738,
+      "step": 6991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.771479448568127e-05,
+      "loss": 18.492,
+      "step": 6992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770606109668151e-05,
+      "loss": 0.7187,
+      "step": 6993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.769732708999298e-05,
+      "loss": 19.9287,
+      "step": 6994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768859246592035e-05,
+      "loss": 20.8065,
+      "step": 6995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767985722476833e-05,
+      "loss": 20.918,
+      "step": 6996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767112136684166e-05,
+      "loss": 19.3776,
+      "step": 6997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.766238489244508e-05,
+      "loss": 0.718,
+      "step": 6998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.765364780188341e-05,
+      "loss": 0.708,
+      "step": 6999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.764491009546143e-05,
+      "loss": 19.3553,
+      "step": 7000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.763617177348393e-05,
+      "loss": 20.3203,
+      "step": 7001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76274328362558e-05,
+      "loss": 17.9265,
+      "step": 7002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.761869328408189e-05,
+      "loss": 19.2124,
+      "step": 7003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.760995311726707e-05,
+      "loss": 18.9549,
+      "step": 7004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.760121233611627e-05,
+      "loss": 19.6537,
+      "step": 7005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.759247094093439e-05,
+      "loss": 0.7267,
+      "step": 7006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75837289320264e-05,
+      "loss": 20.3608,
+      "step": 7007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757498630969726e-05,
+      "loss": 0.6579,
+      "step": 7008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756624307425198e-05,
+      "loss": 19.4978,
+      "step": 7009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.755749922599555e-05,
+      "loss": 19.0739,
+      "step": 7010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.754875476523303e-05,
+      "loss": 19.2053,
+      "step": 7011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.754000969226946e-05,
+      "loss": 18.1911,
+      "step": 7012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.753126400740993e-05,
+      "loss": 0.7162,
+      "step": 7013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752251771095952e-05,
+      "loss": 0.7234,
+      "step": 7014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.751377080322336e-05,
+      "loss": 19.148,
+      "step": 7015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750502328450662e-05,
+      "loss": 19.6564,
+      "step": 7016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.749627515511442e-05,
+      "loss": 20.5839,
+      "step": 7017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.748752641535198e-05,
+      "loss": 18.6528,
+      "step": 7018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.747877706552448e-05,
+      "loss": 0.7374,
+      "step": 7019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.747002710593717e-05,
+      "loss": 0.717,
+      "step": 7020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.746127653689528e-05,
+      "loss": 0.6926,
+      "step": 7021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.74525253587041e-05,
+      "loss": 0.7037,
+      "step": 7022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.744377357166888e-05,
+      "loss": 0.7419,
+      "step": 7023
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743502117609497e-05,
+      "loss": 19.6549,
+      "step": 7024
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.74262681722877e-05,
+      "loss": 0.7778,
+      "step": 7025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.741751456055242e-05,
+      "loss": 21.2894,
+      "step": 7026
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740876034119451e-05,
+      "loss": 0.6413,
+      "step": 7027
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740000551451935e-05,
+      "loss": 19.4506,
+      "step": 7028
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739125008083238e-05,
+      "loss": 0.6936,
+      "step": 7029
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.738249404043902e-05,
+      "loss": 18.8624,
+      "step": 7030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.737373739364477e-05,
+      "loss": 18.2916,
+      "step": 7031
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.736498014075506e-05,
+      "loss": 18.3103,
+      "step": 7032
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.73562222820754e-05,
+      "loss": 0.705,
+      "step": 7033
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.734746381791138e-05,
+      "loss": 19.8562,
+      "step": 7034
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.733870474856847e-05,
+      "loss": 17.2762,
+      "step": 7035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732994507435226e-05,
+      "loss": 0.6345,
+      "step": 7036
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732118479556833e-05,
+      "loss": 18.1262,
+      "step": 7037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.731242391252232e-05,
+      "loss": 0.7161,
+      "step": 7038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.730366242551984e-05,
+      "loss": 18.8393,
+      "step": 7039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.729490033486651e-05,
+      "loss": 19.5144,
+      "step": 7040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.728613764086807e-05,
+      "loss": 0.6983,
+      "step": 7041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.727737434383014e-05,
+      "loss": 0.7172,
+      "step": 7042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.726861044405848e-05,
+      "loss": 19.7576,
+      "step": 7043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725984594185884e-05,
+      "loss": 19.8501,
+      "step": 7044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725108083753691e-05,
+      "loss": 19.2736,
+      "step": 7045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.724231513139852e-05,
+      "loss": 20.9691,
+      "step": 7046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.723354882374947e-05,
+      "loss": 17.7782,
+      "step": 7047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.722478191489553e-05,
+      "loss": 20.7806,
+      "step": 7048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.721601440514261e-05,
+      "loss": 0.6743,
+      "step": 7049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.720724629479652e-05,
+      "loss": 18.0287,
+      "step": 7050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719847758416315e-05,
+      "loss": 0.7527,
+      "step": 7051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718970827354842e-05,
+      "loss": 19.7224,
+      "step": 7052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718093836325823e-05,
+      "loss": 0.7393,
+      "step": 7053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.717216785359856e-05,
+      "loss": 20.8845,
+      "step": 7054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.716339674487532e-05,
+      "loss": 0.7163,
+      "step": 7055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.715462503739456e-05,
+      "loss": 18.428,
+      "step": 7056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.714585273146225e-05,
+      "loss": 0.6652,
+      "step": 7057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713707982738441e-05,
+      "loss": 0.6787,
+      "step": 7058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.712830632546713e-05,
+      "loss": 20.321,
+      "step": 7059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.711953222601644e-05,
+      "loss": 0.7568,
+      "step": 7060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.711075752933847e-05,
+      "loss": 19.382,
+      "step": 7061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.710198223573927e-05,
+      "loss": 20.2777,
+      "step": 7062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.709320634552502e-05,
+      "loss": 19.5162,
+      "step": 7063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708442985900187e-05,
+      "loss": 0.7555,
+      "step": 7064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.707565277647597e-05,
+      "loss": 0.728,
+      "step": 7065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706687509825356e-05,
+      "loss": 0.7081,
+      "step": 7066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.705809682464079e-05,
+      "loss": 0.6924,
+      "step": 7067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.704931795594396e-05,
+      "loss": 0.7311,
+      "step": 7068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.70405384924693e-05,
+      "loss": 20.8565,
+      "step": 7069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.703175843452309e-05,
+      "loss": 17.8132,
+      "step": 7070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702297778241162e-05,
+      "loss": 19.0762,
+      "step": 7071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.701419653644122e-05,
+      "loss": 19.8965,
+      "step": 7072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700541469691823e-05,
+      "loss": 0.6729,
+      "step": 7073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6996632264149e-05,
+      "loss": 20.0324,
+      "step": 7074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.698784923843992e-05,
+      "loss": 19.2842,
+      "step": 7075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.697906562009742e-05,
+      "loss": 0.7273,
+      "step": 7076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.697028140942784e-05,
+      "loss": 0.7257,
+      "step": 7077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.696149660673773e-05,
+      "loss": 18.2762,
+      "step": 7078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.695271121233348e-05,
+      "loss": 21.1939,
+      "step": 7079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.69439252265216e-05,
+      "loss": 0.6605,
+      "step": 7080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.69351386496086e-05,
+      "loss": 18.3126,
+      "step": 7081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.692635148190099e-05,
+      "loss": 0.7467,
+      "step": 7082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691756372370533e-05,
+      "loss": 0.7441,
+      "step": 7083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.690877537532818e-05,
+      "loss": 0.7049,
+      "step": 7084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.689998643707612e-05,
+      "loss": 20.6667,
+      "step": 7085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68911969092558e-05,
+      "loss": 0.7158,
+      "step": 7086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.688240679217379e-05,
+      "loss": 19.1865,
+      "step": 7087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687361608613679e-05,
+      "loss": 0.6622,
+      "step": 7088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.686482479145143e-05,
+      "loss": 18.1081,
+      "step": 7089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.685603290842442e-05,
+      "loss": 20.1597,
+      "step": 7090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.684724043736249e-05,
+      "loss": 19.1684,
+      "step": 7091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.683844737857234e-05,
+      "loss": 18.1586,
+      "step": 7092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682965373236072e-05,
+      "loss": 18.5665,
+      "step": 7093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682085949903443e-05,
+      "loss": 0.6638,
+      "step": 7094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.681206467890025e-05,
+      "loss": 19.2028,
+      "step": 7095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6803269272265e-05,
+      "loss": 20.4506,
+      "step": 7096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.679447327943548e-05,
+      "loss": 18.9066,
+      "step": 7097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.678567670071862e-05,
+      "loss": 0.7243,
+      "step": 7098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67768795364212e-05,
+      "loss": 21.5637,
+      "step": 7099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67680817868502e-05,
+      "loss": 17.9864,
+      "step": 7100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.675928345231249e-05,
+      "loss": 0.7457,
+      "step": 7101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.675048453311499e-05,
+      "loss": 19.609,
+      "step": 7102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67416850295647e-05,
+      "loss": 16.7663,
+      "step": 7103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.673288494196858e-05,
+      "loss": 0.636,
+      "step": 7104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.672408427063364e-05,
+      "loss": 0.6854,
+      "step": 7105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.671528301586687e-05,
+      "loss": 20.384,
+      "step": 7106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670648117797531e-05,
+      "loss": 0.7214,
+      "step": 7107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.669767875726605e-05,
+      "loss": 18.1585,
+      "step": 7108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668887575404613e-05,
+      "loss": 21.6953,
+      "step": 7109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668007216862269e-05,
+      "loss": 20.2008,
+      "step": 7110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667126800130282e-05,
+      "loss": 0.7171,
+      "step": 7111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.666246325239367e-05,
+      "loss": 18.5735,
+      "step": 7112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.66536579222024e-05,
+      "loss": 17.3001,
+      "step": 7113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.664485201103618e-05,
+      "loss": 18.6659,
+      "step": 7114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663604551920222e-05,
+      "loss": 19.5248,
+      "step": 7115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.662723844700774e-05,
+      "loss": 0.6942,
+      "step": 7116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661843079476e-05,
+      "loss": 18.1017,
+      "step": 7117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.660962256276623e-05,
+      "loss": 0.7618,
+      "step": 7118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.660081375133371e-05,
+      "loss": 0.6804,
+      "step": 7119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.659200436076978e-05,
+      "loss": 19.2636,
+      "step": 7120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65831943913817e-05,
+      "loss": 0.7683,
+      "step": 7121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657438384347689e-05,
+      "loss": 0.7477,
+      "step": 7122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.656557271736265e-05,
+      "loss": 19.8637,
+      "step": 7123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.655676101334639e-05,
+      "loss": 0.7107,
+      "step": 7124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.654794873173552e-05,
+      "loss": 19.2317,
+      "step": 7125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653913587283743e-05,
+      "loss": 20.7728,
+      "step": 7126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653032243695961e-05,
+      "loss": 0.6711,
+      "step": 7127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.652150842440948e-05,
+      "loss": 17.0579,
+      "step": 7128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.651269383549453e-05,
+      "loss": 17.5415,
+      "step": 7129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650387867052228e-05,
+      "loss": 16.6958,
+      "step": 7130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649506292980024e-05,
+      "loss": 20.0931,
+      "step": 7131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.648624661363597e-05,
+      "loss": 0.7493,
+      "step": 7132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.647742972233703e-05,
+      "loss": 18.2535,
+      "step": 7133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.646861225621097e-05,
+      "loss": 0.6979,
+      "step": 7134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.645979421556545e-05,
+      "loss": 18.1762,
+      "step": 7135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.645097560070804e-05,
+      "loss": 0.6521,
+      "step": 7136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644215641194644e-05,
+      "loss": 18.9755,
+      "step": 7137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.643333664958826e-05,
+      "loss": 17.8632,
+      "step": 7138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.64245163139412e-05,
+      "loss": 19.2864,
+      "step": 7139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.641569540531298e-05,
+      "loss": 0.6848,
+      "step": 7140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640687392401131e-05,
+      "loss": 0.7168,
+      "step": 7141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.639805187034394e-05,
+      "loss": 0.7101,
+      "step": 7142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638922924461864e-05,
+      "loss": 17.7233,
+      "step": 7143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638040604714319e-05,
+      "loss": 0.7421,
+      "step": 7144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637158227822537e-05,
+      "loss": 17.7156,
+      "step": 7145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636275793817302e-05,
+      "loss": 17.1829,
+      "step": 7146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6353933027294e-05,
+      "loss": 0.6774,
+      "step": 7147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.634510754589614e-05,
+      "loss": 0.6897,
+      "step": 7148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.633628149428736e-05,
+      "loss": 0.7066,
+      "step": 7149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.632745487277554e-05,
+      "loss": 0.7232,
+      "step": 7150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.631862768166862e-05,
+      "loss": 0.681,
+      "step": 7151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.63097999212745e-05,
+      "loss": 20.0077,
+      "step": 7152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.63009715919012e-05,
+      "loss": 20.3588,
+      "step": 7153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.629214269385667e-05,
+      "loss": 20.9702,
+      "step": 7154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628331322744893e-05,
+      "loss": 0.7197,
+      "step": 7155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.627448319298597e-05,
+      "loss": 0.707,
+      "step": 7156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.626565259077587e-05,
+      "loss": 20.2191,
+      "step": 7157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.625682142112667e-05,
+      "loss": 0.7349,
+      "step": 7158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.624798968434647e-05,
+      "loss": 0.7107,
+      "step": 7159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623915738074334e-05,
+      "loss": 0.6775,
+      "step": 7160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623032451062542e-05,
+      "loss": 17.6747,
+      "step": 7161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622149107430088e-05,
+      "loss": 19.4856,
+      "step": 7162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621265707207783e-05,
+      "loss": 18.705,
+      "step": 7163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.620382250426447e-05,
+      "loss": 17.791,
+      "step": 7164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.619498737116902e-05,
+      "loss": 0.68,
+      "step": 7165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618615167309968e-05,
+      "loss": 20.1019,
+      "step": 7166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.617731541036468e-05,
+      "loss": 0.7763,
+      "step": 7167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.61684785832723e-05,
+      "loss": 19.5391,
+      "step": 7168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.61596411921308e-05,
+      "loss": 0.7352,
+      "step": 7169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.615080323724853e-05,
+      "loss": 20.6625,
+      "step": 7170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614196471893374e-05,
+      "loss": 19.4283,
+      "step": 7171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.613312563749479e-05,
+      "loss": 0.6829,
+      "step": 7172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.612428599324006e-05,
+      "loss": 0.7397,
+      "step": 7173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.611544578647791e-05,
+      "loss": 18.7941,
+      "step": 7174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.610660501751674e-05,
+      "loss": 18.6514,
+      "step": 7175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.609776368666495e-05,
+      "loss": 17.58,
+      "step": 7176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6088921794231e-05,
+      "loss": 17.9347,
+      "step": 7177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.608007934052333e-05,
+      "loss": 0.6952,
+      "step": 7178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.607123632585045e-05,
+      "loss": 20.5041,
+      "step": 7179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.606239275052082e-05,
+      "loss": 19.146,
+      "step": 7180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.605354861484295e-05,
+      "loss": 18.5195,
+      "step": 7181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.604470391912539e-05,
+      "loss": 19.1418,
+      "step": 7182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.60358586636767e-05,
+      "loss": 0.7291,
+      "step": 7183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.602701284880543e-05,
+      "loss": 0.6566,
+      "step": 7184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601816647482021e-05,
+      "loss": 18.9958,
+      "step": 7185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.600931954202961e-05,
+      "loss": 19.1578,
+      "step": 7186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.600047205074229e-05,
+      "loss": 18.8602,
+      "step": 7187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599162400126689e-05,
+      "loss": 20.576,
+      "step": 7188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.598277539391206e-05,
+      "loss": 17.0616,
+      "step": 7189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597392622898655e-05,
+      "loss": 0.7723,
+      "step": 7190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5965076506799e-05,
+      "loss": 0.7393,
+      "step": 7191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.595622622765819e-05,
+      "loss": 0.7317,
+      "step": 7192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.594737539187283e-05,
+      "loss": 20.7228,
+      "step": 7193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.593852399975173e-05,
+      "loss": 18.6808,
+      "step": 7194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592967205160366e-05,
+      "loss": 20.4084,
+      "step": 7195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592081954773739e-05,
+      "loss": 0.6834,
+      "step": 7196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.591196648846179e-05,
+      "loss": 0.7043,
+      "step": 7197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.59031128740857e-05,
+      "loss": 0.7168,
+      "step": 7198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.589425870491798e-05,
+      "loss": 19.1393,
+      "step": 7199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.588540398126749e-05,
+      "loss": 0.7138,
+      "step": 7200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.587654870344317e-05,
+      "loss": 20.5662,
+      "step": 7201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.586769287175393e-05,
+      "loss": 0.701,
+      "step": 7202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.58588364865087e-05,
+      "loss": 18.7977,
+      "step": 7203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584997954801646e-05,
+      "loss": 0.6959,
+      "step": 7204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584112205658617e-05,
+      "loss": 19.2543,
+      "step": 7205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.583226401252685e-05,
+      "loss": 18.4529,
+      "step": 7206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.582340541614752e-05,
+      "loss": 18.2752,
+      "step": 7207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.581454626775719e-05,
+      "loss": 0.7077,
+      "step": 7208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.580568656766495e-05,
+      "loss": 17.4767,
+      "step": 7209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.579682631617986e-05,
+      "loss": 18.9001,
+      "step": 7210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578796551361102e-05,
+      "loss": 20.4385,
+      "step": 7211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.577910416026755e-05,
+      "loss": 19.1116,
+      "step": 7212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.577024225645858e-05,
+      "loss": 18.9283,
+      "step": 7213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.576137980249327e-05,
+      "loss": 0.6878,
+      "step": 7214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.575251679868077e-05,
+      "loss": 0.7608,
+      "step": 7215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.574365324533032e-05,
+      "loss": 17.8992,
+      "step": 7216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.573478914275108e-05,
+      "loss": 17.8455,
+      "step": 7217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.572592449125231e-05,
+      "loss": 17.176,
+      "step": 7218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.571705929114325e-05,
+      "loss": 0.6976,
+      "step": 7219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.570819354273317e-05,
+      "loss": 18.64,
+      "step": 7220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.569932724633136e-05,
+      "loss": 20.2957,
+      "step": 7221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.569046040224713e-05,
+      "loss": 0.7539,
+      "step": 7222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.56815930107898e-05,
+      "loss": 19.9136,
+      "step": 7223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.567272507226872e-05,
+      "loss": 19.1865,
+      "step": 7224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.566385658699327e-05,
+      "loss": 0.7261,
+      "step": 7225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.565498755527279e-05,
+      "loss": 21.0617,
+      "step": 7226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.564611797741672e-05,
+      "loss": 0.6782,
+      "step": 7227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.563724785373448e-05,
+      "loss": 18.9038,
+      "step": 7228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562837718453548e-05,
+      "loss": 0.6415,
+      "step": 7229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.561950597012921e-05,
+      "loss": 17.9327,
+      "step": 7230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.561063421082514e-05,
+      "loss": 19.5412,
+      "step": 7231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.560176190693277e-05,
+      "loss": 20.9705,
+      "step": 7232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55928890587616e-05,
+      "loss": 18.9686,
+      "step": 7233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55840156666212e-05,
+      "loss": 20.1873,
+      "step": 7234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55751417308211e-05,
+      "loss": 18.5784,
+      "step": 7235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.556626725167088e-05,
+      "loss": 19.2018,
+      "step": 7236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.555739222948012e-05,
+      "loss": 19.4327,
+      "step": 7237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.554851666455846e-05,
+      "loss": 17.8253,
+      "step": 7238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55396405572155e-05,
+      "loss": 0.7247,
+      "step": 7239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553076390776089e-05,
+      "loss": 0.6496,
+      "step": 7240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.552188671650434e-05,
+      "loss": 19.4649,
+      "step": 7241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551300898375549e-05,
+      "loss": 18.6075,
+      "step": 7242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.550413070982407e-05,
+      "loss": 19.8387,
+      "step": 7243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.54952518950198e-05,
+      "loss": 20.109,
+      "step": 7244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.548637253965242e-05,
+      "loss": 0.6394,
+      "step": 7245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.547749264403169e-05,
+      "loss": 0.7471,
+      "step": 7246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.546861220846741e-05,
+      "loss": 0.7003,
+      "step": 7247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545973123326937e-05,
+      "loss": 0.7613,
+      "step": 7248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545084971874738e-05,
+      "loss": 18.464,
+      "step": 7249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.544196766521128e-05,
+      "loss": 19.3034,
+      "step": 7250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.543308507297095e-05,
+      "loss": 0.7388,
+      "step": 7251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.542420194233623e-05,
+      "loss": 19.7061,
+      "step": 7252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541531827361704e-05,
+      "loss": 0.7049,
+      "step": 7253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.540643406712327e-05,
+      "loss": 19.5826,
+      "step": 7254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53975493231649e-05,
+      "loss": 19.4896,
+      "step": 7255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.538866404205183e-05,
+      "loss": 18.523,
+      "step": 7256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537977822409405e-05,
+      "loss": 18.9344,
+      "step": 7257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537089186960155e-05,
+      "loss": 0.7245,
+      "step": 7258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.536200497888432e-05,
+      "loss": 0.7109,
+      "step": 7259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53531175522524e-05,
+      "loss": 0.6798,
+      "step": 7260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.534422959001585e-05,
+      "loss": 20.9973,
+      "step": 7261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53353410924847e-05,
+      "loss": 0.6946,
+      "step": 7262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.532645205996905e-05,
+      "loss": 20.3726,
+      "step": 7263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.531756249277901e-05,
+      "loss": 17.6172,
+      "step": 7264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.530867239122468e-05,
+      "loss": 0.7214,
+      "step": 7265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.52997817556162e-05,
+      "loss": 19.2154,
+      "step": 7266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.529089058626375e-05,
+      "loss": 0.7639,
+      "step": 7267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.528199888347745e-05,
+      "loss": 17.2526,
+      "step": 7268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.527310664756756e-05,
+      "loss": 0.7254,
+      "step": 7269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526421387884425e-05,
+      "loss": 18.9402,
+      "step": 7270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.525532057761777e-05,
+      "loss": 18.5854,
+      "step": 7271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.524642674419835e-05,
+      "loss": 19.538,
+      "step": 7272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.523753237889627e-05,
+      "loss": 0.6859,
+      "step": 7273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522863748202183e-05,
+      "loss": 18.789,
+      "step": 7274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.52197420538853e-05,
+      "loss": 18.1385,
+      "step": 7275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.521084609479704e-05,
+      "loss": 0.7385,
+      "step": 7276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.520194960506736e-05,
+      "loss": 18.249,
+      "step": 7277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.519305258500666e-05,
+      "loss": 19.8793,
+      "step": 7278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.518415503492529e-05,
+      "loss": 19.8861,
+      "step": 7279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.517525695513364e-05,
+      "loss": 19.1296,
+      "step": 7280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.516635834594216e-05,
+      "loss": 18.4149,
+      "step": 7281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.515745920766125e-05,
+      "loss": 20.1706,
+      "step": 7282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.514855954060137e-05,
+      "loss": 0.6817,
+      "step": 7283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5139659345073e-05,
+      "loss": 20.4472,
+      "step": 7284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.513075862138664e-05,
+      "loss": 19.2201,
+      "step": 7285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.512185736985277e-05,
+      "loss": 16.4387,
+      "step": 7286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.511295559078193e-05,
+      "loss": 0.7254,
+      "step": 7287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.510405328448467e-05,
+      "loss": 18.1757,
+      "step": 7288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.509515045127157e-05,
+      "loss": 0.7027,
+      "step": 7289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.508624709145318e-05,
+      "loss": 0.6789,
+      "step": 7290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.507734320534011e-05,
+      "loss": 20.7636,
+      "step": 7291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.506843879324299e-05,
+      "loss": 20.3466,
+      "step": 7292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505953385547243e-05,
+      "loss": 18.4019,
+      "step": 7293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505062839233911e-05,
+      "loss": 0.6737,
+      "step": 7294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50417224041537e-05,
+      "loss": 17.5516,
+      "step": 7295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.503281589122689e-05,
+      "loss": 18.1369,
+      "step": 7296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.502390885386939e-05,
+      "loss": 0.6276,
+      "step": 7297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.501500129239193e-05,
+      "loss": 0.7213,
+      "step": 7298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.500609320710525e-05,
+      "loss": 19.9029,
+      "step": 7299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499718459832011e-05,
+      "loss": 19.515,
+      "step": 7300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.498827546634733e-05,
+      "loss": 19.1536,
+      "step": 7301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497936581149766e-05,
+      "loss": 19.2126,
+      "step": 7302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497045563408195e-05,
+      "loss": 17.7946,
+      "step": 7303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.496154493441102e-05,
+      "loss": 19.2531,
+      "step": 7304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.495263371279576e-05,
+      "loss": 17.2852,
+      "step": 7305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.494372196954702e-05,
+      "loss": 0.6617,
+      "step": 7306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.493480970497569e-05,
+      "loss": 18.5289,
+      "step": 7307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.492589691939269e-05,
+      "loss": 18.3621,
+      "step": 7308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.491698361310893e-05,
+      "loss": 0.7009,
+      "step": 7309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.490806978643537e-05,
+      "loss": 0.7088,
+      "step": 7310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.489915543968299e-05,
+      "loss": 20.1183,
+      "step": 7311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.489024057316276e-05,
+      "loss": 18.3882,
+      "step": 7312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.488132518718568e-05,
+      "loss": 0.705,
+      "step": 7313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.487240928206277e-05,
+      "loss": 19.1789,
+      "step": 7314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.486349285810508e-05,
+      "loss": 0.715,
+      "step": 7315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.485457591562363e-05,
+      "loss": 19.1196,
+      "step": 7316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.484565845492954e-05,
+      "loss": 0.6926,
+      "step": 7317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.48367404763339e-05,
+      "loss": 19.4909,
+      "step": 7318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.482782198014776e-05,
+      "loss": 18.737,
+      "step": 7319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.481890296668232e-05,
+      "loss": 0.6757,
+      "step": 7320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.480998343624869e-05,
+      "loss": 19.089,
+      "step": 7321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.480106338915804e-05,
+      "loss": 17.2837,
+      "step": 7322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.479214282572154e-05,
+      "loss": 19.3179,
+      "step": 7323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478322174625042e-05,
+      "loss": 0.713,
+      "step": 7324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.477430015105588e-05,
+      "loss": 0.7541,
+      "step": 7325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.476537804044916e-05,
+      "loss": 0.648,
+      "step": 7326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.47564554147415e-05,
+      "loss": 19.5159,
+      "step": 7327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.474753227424421e-05,
+      "loss": 0.6993,
+      "step": 7328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.473860861926852e-05,
+      "loss": 19.5659,
+      "step": 7329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472968445012579e-05,
+      "loss": 0.6474,
+      "step": 7330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472075976712732e-05,
+      "loss": 0.7096,
+      "step": 7331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.471183457058449e-05,
+      "loss": 0.7146,
+      "step": 7332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470290886080863e-05,
+      "loss": 0.7216,
+      "step": 7333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.469398263811111e-05,
+      "loss": 0.6545,
+      "step": 7334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468505590280335e-05,
+      "loss": 18.285,
+      "step": 7335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.467612865519674e-05,
+      "loss": 17.6215,
+      "step": 7336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466720089560275e-05,
+      "loss": 0.7295,
+      "step": 7337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.465827262433281e-05,
+      "loss": 20.61,
+      "step": 7338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.46493438416984e-05,
+      "loss": 0.7202,
+      "step": 7339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4640414548011e-05,
+      "loss": 0.7091,
+      "step": 7340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.463148474358209e-05,
+      "loss": 0.7237,
+      "step": 7341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462255442872323e-05,
+      "loss": 0.6799,
+      "step": 7342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.461362360374595e-05,
+      "loss": 0.7339,
+      "step": 7343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.460469226896182e-05,
+      "loss": 17.3645,
+      "step": 7344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.459576042468238e-05,
+      "loss": 21.3183,
+      "step": 7345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.458682807121925e-05,
+      "loss": 17.8941,
+      "step": 7346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.457789520888405e-05,
+      "loss": 0.7554,
+      "step": 7347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.45689618379884e-05,
+      "loss": 0.6475,
+      "step": 7348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.456002795884393e-05,
+      "loss": 17.777,
+      "step": 7349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.455109357176232e-05,
+      "loss": 0.7148,
+      "step": 7350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.454215867705525e-05,
+      "loss": 18.1758,
+      "step": 7351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.453322327503443e-05,
+      "loss": 20.1543,
+      "step": 7352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.452428736601157e-05,
+      "loss": 17.6529,
+      "step": 7353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.45153509502984e-05,
+      "loss": 19.1955,
+      "step": 7354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.450641402820668e-05,
+      "loss": 18.997,
+      "step": 7355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.449747660004818e-05,
+      "loss": 0.6885,
+      "step": 7356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.448853866613468e-05,
+      "loss": 0.6727,
+      "step": 7357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447960022677802e-05,
+      "loss": 0.7135,
+      "step": 7358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447066128228998e-05,
+      "loss": 0.7245,
+      "step": 7359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.446172183298243e-05,
+      "loss": 0.7007,
+      "step": 7360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.445278187916722e-05,
+      "loss": 20.6745,
+      "step": 7361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.444384142115623e-05,
+      "loss": 0.7833,
+      "step": 7362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.443490045926133e-05,
+      "loss": 19.2262,
+      "step": 7363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.442595899379447e-05,
+      "loss": 21.1642,
+      "step": 7364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.441701702506754e-05,
+      "loss": 17.9176,
+      "step": 7365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.440807455339252e-05,
+      "loss": 19.7917,
+      "step": 7366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.439913157908139e-05,
+      "loss": 0.7348,
+      "step": 7367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.439018810244607e-05,
+      "loss": 20.5912,
+      "step": 7368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.438124412379862e-05,
+      "loss": 18.2164,
+      "step": 7369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.437229964345102e-05,
+      "loss": 0.6411,
+      "step": 7370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436335466171531e-05,
+      "loss": 19.778,
+      "step": 7371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.435440917890356e-05,
+      "loss": 21.0689,
+      "step": 7372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434546319532783e-05,
+      "loss": 19.7593,
+      "step": 7373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.433651671130021e-05,
+      "loss": 0.6951,
+      "step": 7374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.43275697271328e-05,
+      "loss": 0.679,
+      "step": 7375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.431862224313772e-05,
+      "loss": 0.6797,
+      "step": 7376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.43096742596271e-05,
+      "loss": 18.8137,
+      "step": 7377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430072577691313e-05,
+      "loss": 19.4347,
+      "step": 7378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.429177679530795e-05,
+      "loss": 17.8335,
+      "step": 7379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.428282731512379e-05,
+      "loss": 17.4327,
+      "step": 7380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42738773366728e-05,
+      "loss": 19.6009,
+      "step": 7381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.426492686026725e-05,
+      "loss": 0.6801,
+      "step": 7382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.425597588621938e-05,
+      "loss": 0.8014,
+      "step": 7383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.424702441484146e-05,
+      "loss": 17.7835,
+      "step": 7384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.423807244644575e-05,
+      "loss": 0.6675,
+      "step": 7385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422911998134453e-05,
+      "loss": 20.5891,
+      "step": 7386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422016701985016e-05,
+      "loss": 18.7018,
+      "step": 7387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.421121356227493e-05,
+      "loss": 19.7226,
+      "step": 7388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.420225960893121e-05,
+      "loss": 18.1901,
+      "step": 7389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.419330516013135e-05,
+      "loss": 18.7215,
+      "step": 7390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.418435021618774e-05,
+      "loss": 20.9376,
+      "step": 7391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.417539477741279e-05,
+      "loss": 18.7652,
+      "step": 7392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.416643884411889e-05,
+      "loss": 18.1146,
+      "step": 7393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.415748241661851e-05,
+      "loss": 0.6722,
+      "step": 7394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.414852549522406e-05,
+      "loss": 19.5262,
+      "step": 7395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413956808024805e-05,
+      "loss": 0.6992,
+      "step": 7396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413061017200294e-05,
+      "loss": 17.7179,
+      "step": 7397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.412165177080123e-05,
+      "loss": 19.8805,
+      "step": 7398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.411269287695548e-05,
+      "loss": 0.7278,
+      "step": 7399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.410373349077816e-05,
+      "loss": 19.9575,
+      "step": 7400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.409477361258189e-05,
+      "loss": 0.6984,
+      "step": 7401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40858132426792e-05,
+      "loss": 19.5292,
+      "step": 7402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40768523813827e-05,
+      "loss": 18.0794,
+      "step": 7403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4067891029005e-05,
+      "loss": 0.6945,
+      "step": 7404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405892918585868e-05,
+      "loss": 18.6278,
+      "step": 7405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.404996685225644e-05,
+      "loss": 0.7222,
+      "step": 7406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.404100402851092e-05,
+      "loss": 0.6945,
+      "step": 7407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.403204071493476e-05,
+      "loss": 0.6691,
+      "step": 7408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.402307691184069e-05,
+      "loss": 0.7616,
+      "step": 7409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40141126195414e-05,
+      "loss": 21.1629,
+      "step": 7410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.400514783834963e-05,
+      "loss": 16.9348,
+      "step": 7411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.399618256857811e-05,
+      "loss": 0.6876,
+      "step": 7412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.398721681053961e-05,
+      "loss": 18.815,
+      "step": 7413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.39782505645469e-05,
+      "loss": 18.2313,
+      "step": 7414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.396928383091276e-05,
+      "loss": 0.7191,
+      "step": 7415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.396031660995004e-05,
+      "loss": 17.8833,
+      "step": 7416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.395134890197153e-05,
+      "loss": 0.7067,
+      "step": 7417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.39423807072901e-05,
+      "loss": 19.1122,
+      "step": 7418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.39334120262186e-05,
+      "loss": 18.3957,
+      "step": 7419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.39244428590699e-05,
+      "loss": 19.7017,
+      "step": 7420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.391547320615692e-05,
+      "loss": 0.7205,
+      "step": 7421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.390650306779254e-05,
+      "loss": 17.6099,
+      "step": 7422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.389753244428972e-05,
+      "loss": 18.3812,
+      "step": 7423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38885613359614e-05,
+      "loss": 0.73,
+      "step": 7424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387958974312054e-05,
+      "loss": 17.9909,
+      "step": 7425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387061766608012e-05,
+      "loss": 0.7707,
+      "step": 7426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.386164510515311e-05,
+      "loss": 17.4177,
+      "step": 7427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.385267206065258e-05,
+      "loss": 16.9297,
+      "step": 7428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384369853289154e-05,
+      "loss": 0.6514,
+      "step": 7429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.383472452218301e-05,
+      "loss": 0.6905,
+      "step": 7430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.382575002884008e-05,
+      "loss": 0.7066,
+      "step": 7431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.381677505317583e-05,
+      "loss": 19.3647,
+      "step": 7432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.380779959550336e-05,
+      "loss": 18.3133,
+      "step": 7433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.379882365613578e-05,
+      "loss": 19.0176,
+      "step": 7434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.378984723538623e-05,
+      "loss": 20.3358,
+      "step": 7435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.378087033356785e-05,
+      "loss": 18.2397,
+      "step": 7436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.377189295099378e-05,
+      "loss": 17.4174,
+      "step": 7437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.376291508797725e-05,
+      "loss": 0.7464,
+      "step": 7438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375393674483145e-05,
+      "loss": 17.479,
+      "step": 7439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.374495792186958e-05,
+      "loss": 0.7091,
+      "step": 7440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.373597861940488e-05,
+      "loss": 19.2835,
+      "step": 7441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.37269988377506e-05,
+      "loss": 18.4892,
+      "step": 7442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371801857722002e-05,
+      "loss": 18.8906,
+      "step": 7443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.370903783812637e-05,
+      "loss": 21.2609,
+      "step": 7444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.370005662078302e-05,
+      "loss": 0.7019,
+      "step": 7445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369107492550323e-05,
+      "loss": 18.4803,
+      "step": 7446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.368209275260036e-05,
+      "loss": 18.0896,
+      "step": 7447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.367311010238776e-05,
+      "loss": 0.6525,
+      "step": 7448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.366412697517879e-05,
+      "loss": 0.6896,
+      "step": 7449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.365514337128683e-05,
+      "loss": 17.4869,
+      "step": 7450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.364615929102531e-05,
+      "loss": 18.0143,
+      "step": 7451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.363717473470759e-05,
+      "loss": 0.6488,
+      "step": 7452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.362818970264715e-05,
+      "loss": 18.422,
+      "step": 7453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361920419515742e-05,
+      "loss": 0.7187,
+      "step": 7454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361021821255186e-05,
+      "loss": 0.7126,
+      "step": 7455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.360123175514397e-05,
+      "loss": 20.0,
+      "step": 7456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.359224482324724e-05,
+      "loss": 0.7144,
+      "step": 7457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.358325741717519e-05,
+      "loss": 20.7314,
+      "step": 7458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357426953724135e-05,
+      "loss": 0.6921,
+      "step": 7459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.356528118375925e-05,
+      "loss": 17.816,
+      "step": 7460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.355629235704248e-05,
+      "loss": 0.6542,
+      "step": 7461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.354730305740462e-05,
+      "loss": 19.7093,
+      "step": 7462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353831328515926e-05,
+      "loss": 0.7004,
+      "step": 7463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.352932304062002e-05,
+      "loss": 17.7317,
+      "step": 7464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.35203323241005e-05,
+      "loss": 0.7643,
+      "step": 7465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351134113591442e-05,
+      "loss": 18.2469,
+      "step": 7466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.350234947637537e-05,
+      "loss": 0.6992,
+      "step": 7467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.349335734579706e-05,
+      "loss": 0.72,
+      "step": 7468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.348436474449321e-05,
+      "loss": 0.7207,
+      "step": 7469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.347537167277749e-05,
+      "loss": 20.4286,
+      "step": 7470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.346637813096367e-05,
+      "loss": 17.1042,
+      "step": 7471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.345738411936546e-05,
+      "loss": 0.6374,
+      "step": 7472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.344838963829665e-05,
+      "loss": 0.6343,
+      "step": 7473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.343939468807099e-05,
+      "loss": 18.3857,
+      "step": 7474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.343039926900232e-05,
+      "loss": 21.2144,
+      "step": 7475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.342140338140441e-05,
+      "loss": 18.9943,
+      "step": 7476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.34124070255911e-05,
+      "loss": 0.6833,
+      "step": 7477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.340341020187624e-05,
+      "loss": 17.9139,
+      "step": 7478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.339441291057369e-05,
+      "loss": 0.7746,
+      "step": 7479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.338541515199732e-05,
+      "loss": 17.8632,
+      "step": 7480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337641692646106e-05,
+      "loss": 0.7248,
+      "step": 7481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.336741823427876e-05,
+      "loss": 17.675,
+      "step": 7482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.335841907576438e-05,
+      "loss": 19.4691,
+      "step": 7483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.334941945123185e-05,
+      "loss": 0.6921,
+      "step": 7484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.334041936099513e-05,
+      "loss": 19.8362,
+      "step": 7485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333141880536824e-05,
+      "loss": 19.3655,
+      "step": 7486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.332241778466509e-05,
+      "loss": 0.7173,
+      "step": 7487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331341629919975e-05,
+      "loss": 19.7078,
+      "step": 7488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.33044143492862e-05,
+      "loss": 0.6542,
+      "step": 7489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329541193523851e-05,
+      "loss": 0.6807,
+      "step": 7490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.328640905737074e-05,
+      "loss": 20.0341,
+      "step": 7491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327740571599694e-05,
+      "loss": 19.6059,
+      "step": 7492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.326840191143121e-05,
+      "loss": 0.6967,
+      "step": 7493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325939764398764e-05,
+      "loss": 19.4562,
+      "step": 7494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325039291398037e-05,
+      "loss": 0.7249,
+      "step": 7495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.324138772172353e-05,
+      "loss": 0.6759,
+      "step": 7496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323238206753127e-05,
+      "loss": 16.7206,
+      "step": 7497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.322337595171777e-05,
+      "loss": 19.2956,
+      "step": 7498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321436937459717e-05,
+      "loss": 18.6248,
+      "step": 7499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.320536233648374e-05,
+      "loss": 0.6803,
+      "step": 7500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.319635483769164e-05,
+      "loss": 19.9796,
+      "step": 7501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.318734687853514e-05,
+      "loss": 0.6713,
+      "step": 7502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317833845932847e-05,
+      "loss": 19.267,
+      "step": 7503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.316932958038588e-05,
+      "loss": 0.6759,
+      "step": 7504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.31603202420217e-05,
+      "loss": 18.3074,
+      "step": 7505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.315131044455019e-05,
+      "loss": 0.7273,
+      "step": 7506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.314230018828566e-05,
+      "loss": 19.7903,
+      "step": 7507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.313328947354247e-05,
+      "loss": 0.703,
+      "step": 7508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.312427830063493e-05,
+      "loss": 18.1178,
+      "step": 7509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.311526666987743e-05,
+      "loss": 18.5272,
+      "step": 7510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.310625458158432e-05,
+      "loss": 18.9134,
+      "step": 7511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.309724203607001e-05,
+      "loss": 17.9722,
+      "step": 7512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308822903364893e-05,
+      "loss": 17.3822,
+      "step": 7513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.307921557463546e-05,
+      "loss": 17.5812,
+      "step": 7514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.307020165934408e-05,
+      "loss": 0.6519,
+      "step": 7515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306118728808921e-05,
+      "loss": 18.226,
+      "step": 7516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.305217246118535e-05,
+      "loss": 0.7517,
+      "step": 7517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3043157178947e-05,
+      "loss": 18.3741,
+      "step": 7518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.303414144168864e-05,
+      "loss": 0.7332,
+      "step": 7519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302512524972478e-05,
+      "loss": 0.7836,
+      "step": 7520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.301610860336998e-05,
+      "loss": 19.4312,
+      "step": 7521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.30070915029388e-05,
+      "loss": 0.6912,
+      "step": 7522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.299807394874579e-05,
+      "loss": 0.7087,
+      "step": 7523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.298905594110552e-05,
+      "loss": 19.7506,
+      "step": 7524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.298003748033265e-05,
+      "loss": 20.0684,
+      "step": 7525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.297101856674171e-05,
+      "loss": 18.6824,
+      "step": 7526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.296199920064739e-05,
+      "loss": 0.6186,
+      "step": 7527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.295297938236431e-05,
+      "loss": 20.556,
+      "step": 7528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.294395911220717e-05,
+      "loss": 18.4594,
+      "step": 7529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293493839049062e-05,
+      "loss": 0.7072,
+      "step": 7530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.292591721752934e-05,
+      "loss": 0.7026,
+      "step": 7531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.291689559363807e-05,
+      "loss": 0.6313,
+      "step": 7532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.290787351913151e-05,
+      "loss": 20.6419,
+      "step": 7533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289885099432443e-05,
+      "loss": 0.6899,
+      "step": 7534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.288982801953158e-05,
+      "loss": 0.6735,
+      "step": 7535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.28808045950677e-05,
+      "loss": 17.8328,
+      "step": 7536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287178072124762e-05,
+      "loss": 0.719,
+      "step": 7537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.286275639838613e-05,
+      "loss": 0.6349,
+      "step": 7538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.285373162679803e-05,
+      "loss": 0.7235,
+      "step": 7539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.284470640679822e-05,
+      "loss": 17.6301,
+      "step": 7540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283568073870147e-05,
+      "loss": 18.1586,
+      "step": 7541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.282665462282269e-05,
+      "loss": 18.7729,
+      "step": 7542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.281762805947675e-05,
+      "loss": 17.3624,
+      "step": 7543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.280860104897857e-05,
+      "loss": 17.3119,
+      "step": 7544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279957359164303e-05,
+      "loss": 17.9717,
+      "step": 7545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279054568778508e-05,
+      "loss": 19.5515,
+      "step": 7546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.278151733771968e-05,
+      "loss": 17.0322,
+      "step": 7547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.277248854176175e-05,
+      "loss": 0.7408,
+      "step": 7548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27634593002263e-05,
+      "loss": 20.1092,
+      "step": 7549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.275442961342832e-05,
+      "loss": 19.2543,
+      "step": 7550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.274539948168278e-05,
+      "loss": 0.6912,
+      "step": 7551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.273636890530476e-05,
+      "loss": 0.7419,
+      "step": 7552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.272733788460925e-05,
+      "loss": 18.0825,
+      "step": 7553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.271830641991133e-05,
+      "loss": 19.5243,
+      "step": 7554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270927451152606e-05,
+      "loss": 0.7141,
+      "step": 7555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270024215976853e-05,
+      "loss": 18.625,
+      "step": 7556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.269120936495383e-05,
+      "loss": 0.6538,
+      "step": 7557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.268217612739707e-05,
+      "loss": 17.7428,
+      "step": 7558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267314244741342e-05,
+      "loss": 19.6065,
+      "step": 7559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.266410832531798e-05,
+      "loss": 18.1833,
+      "step": 7560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.265507376142594e-05,
+      "loss": 18.2484,
+      "step": 7561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.264603875605245e-05,
+      "loss": 20.3093,
+      "step": 7562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263700330951274e-05,
+      "loss": 16.8405,
+      "step": 7563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.262796742212201e-05,
+      "loss": 0.6416,
+      "step": 7564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.261893109419545e-05,
+      "loss": 0.6737,
+      "step": 7565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.260989432604834e-05,
+      "loss": 21.6703,
+      "step": 7566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26008571179959e-05,
+      "loss": 17.1607,
+      "step": 7567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.259181947035342e-05,
+      "loss": 18.677,
+      "step": 7568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.258278138343619e-05,
+      "loss": 18.5681,
+      "step": 7569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.257374285755948e-05,
+      "loss": 0.7188,
+      "step": 7570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.256470389303863e-05,
+      "loss": 0.6847,
+      "step": 7571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255566449018899e-05,
+      "loss": 18.8923,
+      "step": 7572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.254662464932586e-05,
+      "loss": 0.7142,
+      "step": 7573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.253758437076463e-05,
+      "loss": 18.0131,
+      "step": 7574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.252854365482069e-05,
+      "loss": 21.4704,
+      "step": 7575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251950250180938e-05,
+      "loss": 20.3378,
+      "step": 7576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251046091204617e-05,
+      "loss": 0.6624,
+      "step": 7577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.250141888584644e-05,
+      "loss": 0.7041,
+      "step": 7578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.249237642352563e-05,
+      "loss": 0.6981,
+      "step": 7579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.248333352539923e-05,
+      "loss": 18.3412,
+      "step": 7580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.247429019178266e-05,
+      "loss": 20.8966,
+      "step": 7581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.246524642299145e-05,
+      "loss": 18.5588,
+      "step": 7582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245620221934103e-05,
+      "loss": 0.7091,
+      "step": 7583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.244715758114699e-05,
+      "loss": 0.7542,
+      "step": 7584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.243811250872482e-05,
+      "loss": 17.7374,
+      "step": 7585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242906700239005e-05,
+      "loss": 20.0369,
+      "step": 7586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242002106245828e-05,
+      "loss": 20.5434,
+      "step": 7587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.241097468924505e-05,
+      "loss": 19.0771,
+      "step": 7588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.240192788306597e-05,
+      "loss": 0.7001,
+      "step": 7589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.239288064423663e-05,
+      "loss": 18.0064,
+      "step": 7590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.238383297307267e-05,
+      "loss": 0.7228,
+      "step": 7591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.237478486988968e-05,
+      "loss": 18.8254,
+      "step": 7592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.236573633500336e-05,
+      "loss": 0.6984,
+      "step": 7593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.235668736872936e-05,
+      "loss": 0.7018,
+      "step": 7594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.234763797138334e-05,
+      "loss": 0.5766,
+      "step": 7595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.233858814328102e-05,
+      "loss": 18.7715,
+      "step": 7596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.232953788473811e-05,
+      "loss": 0.6497,
+      "step": 7597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23204871960703e-05,
+      "loss": 17.2624,
+      "step": 7598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.231143607759337e-05,
+      "loss": 0.7021,
+      "step": 7599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230238452962304e-05,
+      "loss": 0.7505,
+      "step": 7600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.229333255247511e-05,
+      "loss": 20.2979,
+      "step": 7601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.228428014646537e-05,
+      "loss": 18.7782,
+      "step": 7602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.227522731190958e-05,
+      "loss": 0.702,
+      "step": 7603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.226617404912359e-05,
+      "loss": 0.7109,
+      "step": 7604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.225712035842319e-05,
+      "loss": 18.716,
+      "step": 7605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22480662401243e-05,
+      "loss": 0.7133,
+      "step": 7606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22390116945427e-05,
+      "loss": 17.9831,
+      "step": 7607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22299567219943e-05,
+      "loss": 17.9034,
+      "step": 7608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.222090132279499e-05,
+      "loss": 17.7507,
+      "step": 7609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.221184549726066e-05,
+      "loss": 19.2062,
+      "step": 7610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.220278924570724e-05,
+      "loss": 19.117,
+      "step": 7611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.219373256845066e-05,
+      "loss": 0.7441,
+      "step": 7612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.218467546580688e-05,
+      "loss": 0.7128,
+      "step": 7613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.217561793809186e-05,
+      "loss": 19.1434,
+      "step": 7614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.216655998562155e-05,
+      "loss": 0.7306,
+      "step": 7615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.215750160871197e-05,
+      "loss": 19.7964,
+      "step": 7616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214844280767914e-05,
+      "loss": 19.5931,
+      "step": 7617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213938358283906e-05,
+      "loss": 0.6826,
+      "step": 7618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213032393450779e-05,
+      "loss": 17.895,
+      "step": 7619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.212126386300134e-05,
+      "loss": 16.4672,
+      "step": 7620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.211220336863582e-05,
+      "loss": 0.7126,
+      "step": 7621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.210314245172728e-05,
+      "loss": 19.2211,
+      "step": 7622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.209408111259185e-05,
+      "loss": 18.843,
+      "step": 7623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.208501935154562e-05,
+      "loss": 0.6431,
+      "step": 7624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.207595716890471e-05,
+      "loss": 0.7094,
+      "step": 7625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206689456498529e-05,
+      "loss": 17.5777,
+      "step": 7626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.205783154010348e-05,
+      "loss": 0.7352,
+      "step": 7627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204876809457548e-05,
+      "loss": 19.5669,
+      "step": 7628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203970422871745e-05,
+      "loss": 19.3166,
+      "step": 7629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203063994284561e-05,
+      "loss": 0.6974,
+      "step": 7630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.202157523727617e-05,
+      "loss": 0.7368,
+      "step": 7631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.201251011232533e-05,
+      "loss": 20.7588,
+      "step": 7632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.200344456830938e-05,
+      "loss": 0.6464,
+      "step": 7633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.199437860554456e-05,
+      "loss": 20.529,
+      "step": 7634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.198531222434712e-05,
+      "loss": 18.7479,
+      "step": 7635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.197624542503338e-05,
+      "loss": 0.71,
+      "step": 7636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196717820791962e-05,
+      "loss": 19.9663,
+      "step": 7637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.195811057332216e-05,
+      "loss": 0.6825,
+      "step": 7638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.194904252155735e-05,
+      "loss": 0.6738,
+      "step": 7639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193997405294151e-05,
+      "loss": 19.61,
+      "step": 7640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193090516779102e-05,
+      "loss": 0.7034,
+      "step": 7641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.192183586642223e-05,
+      "loss": 18.5942,
+      "step": 7642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.191276614915155e-05,
+      "loss": 20.3889,
+      "step": 7643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.190369601629539e-05,
+      "loss": 19.3268,
+      "step": 7644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.189462546817014e-05,
+      "loss": 19.7021,
+      "step": 7645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.188555450509226e-05,
+      "loss": 19.1724,
+      "step": 7646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.187648312737817e-05,
+      "loss": 0.6776,
+      "step": 7647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.186741133534438e-05,
+      "loss": 0.7096,
+      "step": 7648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.185833912930731e-05,
+      "loss": 18.7901,
+      "step": 7649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184926650958347e-05,
+      "loss": 0.6711,
+      "step": 7650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184019347648938e-05,
+      "loss": 17.7799,
+      "step": 7651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.183112003034154e-05,
+      "loss": 19.8034,
+      "step": 7652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.18220461714565e-05,
+      "loss": 0.6159,
+      "step": 7653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.181297190015081e-05,
+      "loss": 18.425,
+      "step": 7654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1803897216741e-05,
+      "loss": 17.8756,
+      "step": 7655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.179482212154369e-05,
+      "loss": 17.0201,
+      "step": 7656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178574661487544e-05,
+      "loss": 19.0407,
+      "step": 7657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.177667069705287e-05,
+      "loss": 0.695,
+      "step": 7658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.17675943683926e-05,
+      "loss": 0.6798,
+      "step": 7659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.175851762921125e-05,
+      "loss": 19.3542,
+      "step": 7660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174944047982549e-05,
+      "loss": 17.8192,
+      "step": 7661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174036292055198e-05,
+      "loss": 0.7001,
+      "step": 7662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.17312849517074e-05,
+      "loss": 0.6986,
+      "step": 7663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.172220657360839e-05,
+      "loss": 18.1907,
+      "step": 7664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.171312778657173e-05,
+      "loss": 18.1101,
+      "step": 7665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.170404859091409e-05,
+      "loss": 0.7075,
+      "step": 7666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.169496898695222e-05,
+      "loss": 17.5269,
+      "step": 7667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.16858889750029e-05,
+      "loss": 0.6705,
+      "step": 7668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.167680855538283e-05,
+      "loss": 18.7396,
+      "step": 7669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.166772772840883e-05,
+      "loss": 0.7484,
+      "step": 7670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.16586464943977e-05,
+      "loss": 19.0139,
+      "step": 7671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164956485366619e-05,
+      "loss": 20.173,
+      "step": 7672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164048280653119e-05,
+      "loss": 18.2154,
+      "step": 7673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.163140035330947e-05,
+      "loss": 16.3585,
+      "step": 7674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.162231749431793e-05,
+      "loss": 17.0689,
+      "step": 7675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.161323422987342e-05,
+      "loss": 20.6828,
+      "step": 7676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.160415056029278e-05,
+      "loss": 18.8552,
+      "step": 7677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.159506648589294e-05,
+      "loss": 19.9573,
+      "step": 7678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.15859820069908e-05,
+      "loss": 0.6889,
+      "step": 7679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157689712390326e-05,
+      "loss": 19.6002,
+      "step": 7680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.156781183694727e-05,
+      "loss": 16.8472,
+      "step": 7681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155872614643975e-05,
+      "loss": 18.5048,
+      "step": 7682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.154964005269771e-05,
+      "loss": 0.6653,
+      "step": 7683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.154055355603807e-05,
+      "loss": 19.6865,
+      "step": 7684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153146665677785e-05,
+      "loss": 16.6791,
+      "step": 7685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.152237935523406e-05,
+      "loss": 0.6794,
+      "step": 7686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151329165172369e-05,
+      "loss": 0.7083,
+      "step": 7687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.150420354656381e-05,
+      "loss": 17.751,
+      "step": 7688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.149511504007143e-05,
+      "loss": 18.9449,
+      "step": 7689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.148602613256362e-05,
+      "loss": 19.1947,
+      "step": 7690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.147693682435743e-05,
+      "loss": 0.6877,
+      "step": 7691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.146784711577e-05,
+      "loss": 18.9488,
+      "step": 7692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14587570071184e-05,
+      "loss": 0.6604,
+      "step": 7693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.144966649871973e-05,
+      "loss": 0.6986,
+      "step": 7694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.144057559089113e-05,
+      "loss": 0.644,
+      "step": 7695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.143148428394976e-05,
+      "loss": 0.7424,
+      "step": 7696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.142239257821277e-05,
+      "loss": 19.0295,
+      "step": 7697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.141330047399734e-05,
+      "loss": 0.6747,
+      "step": 7698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.140420797162061e-05,
+      "loss": 16.4364,
+      "step": 7699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.139511507139983e-05,
+      "loss": 0.6798,
+      "step": 7700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138602177365217e-05,
+      "loss": 17.792,
+      "step": 7701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.137692807869491e-05,
+      "loss": 18.8831,
+      "step": 7702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.136783398684524e-05,
+      "loss": 17.5262,
+      "step": 7703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.135873949842043e-05,
+      "loss": 0.6996,
+      "step": 7704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134964461373775e-05,
+      "loss": 0.7233,
+      "step": 7705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134054933311448e-05,
+      "loss": 17.8838,
+      "step": 7706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.133145365686792e-05,
+      "loss": 18.196,
+      "step": 7707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.13223575853154e-05,
+      "loss": 17.5753,
+      "step": 7708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.131326111877417e-05,
+      "loss": 0.6832,
+      "step": 7709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.130416425756166e-05,
+      "loss": 0.6423,
+      "step": 7710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.129506700199515e-05,
+      "loss": 18.3225,
+      "step": 7711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128596935239205e-05,
+      "loss": 0.6872,
+      "step": 7712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.127687130906972e-05,
+      "loss": 18.5524,
+      "step": 7713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.126777287234553e-05,
+      "loss": 19.7578,
+      "step": 7714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.125867404253693e-05,
+      "loss": 17.6697,
+      "step": 7715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.12495748199613e-05,
+      "loss": 20.5959,
+      "step": 7716
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.12404752049361e-05,
+      "loss": 20.3741,
+      "step": 7717
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.123137519777877e-05,
+      "loss": 0.6814,
+      "step": 7718
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.122227479880675e-05,
+      "loss": 0.6875,
+      "step": 7719
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.121317400833754e-05,
+      "loss": 0.6677,
+      "step": 7720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.120407282668861e-05,
+      "loss": 16.8702,
+      "step": 7721
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.119497125417748e-05,
+      "loss": 20.74,
+      "step": 7722
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.118586929112163e-05,
+      "loss": 18.8705,
+      "step": 7723
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117676693783863e-05,
+      "loss": 0.7242,
+      "step": 7724
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.116766419464601e-05,
+      "loss": 0.6838,
+      "step": 7725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.115856106186132e-05,
+      "loss": 0.7358,
+      "step": 7726
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.114945753980213e-05,
+      "loss": 0.6796,
+      "step": 7727
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1140353628786e-05,
+      "loss": 0.6818,
+      "step": 7728
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113124932913058e-05,
+      "loss": 19.2891,
+      "step": 7729
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.112214464115344e-05,
+      "loss": 20.8051,
+      "step": 7730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.11130395651722e-05,
+      "loss": 0.6428,
+      "step": 7731
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.110393410150453e-05,
+      "loss": 20.3196,
+      "step": 7732
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.109482825046806e-05,
+      "loss": 19.3356,
+      "step": 7733
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.108572201238044e-05,
+      "loss": 18.5224,
+      "step": 7734
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.107661538755937e-05,
+      "loss": 18.7657,
+      "step": 7735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.106750837632254e-05,
+      "loss": 18.3107,
+      "step": 7736
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105840097898764e-05,
+      "loss": 19.4763,
+      "step": 7737
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104929319587241e-05,
+      "loss": 0.64,
+      "step": 7738
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104018502729455e-05,
+      "loss": 0.7487,
+      "step": 7739
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1031076473571845e-05,
+      "loss": 0.6837,
+      "step": 7740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1021967535022014e-05,
+      "loss": 0.7122,
+      "step": 7741
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.101285821196285e-05,
+      "loss": 0.7084,
+      "step": 7742
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1003748504712155e-05,
+      "loss": 0.6965,
+      "step": 7743
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.09946384135877e-05,
+      "loss": 20.0053,
+      "step": 7744
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0985527938907306e-05,
+      "loss": 21.7712,
+      "step": 7745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.097641708098878e-05,
+      "loss": 18.2717,
+      "step": 7746
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.096730584014999e-05,
+      "loss": 18.9816,
+      "step": 7747
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.095819421670878e-05,
+      "loss": 0.6846,
+      "step": 7748
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.094908221098301e-05,
+      "loss": 19.1418,
+      "step": 7749
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0939969823290564e-05,
+      "loss": 18.7403,
+      "step": 7750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.093085705394933e-05,
+      "loss": 0.6712,
+      "step": 7751
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0921743903277216e-05,
+      "loss": 18.367,
+      "step": 7752
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.091263037159214e-05,
+      "loss": 19.7,
+      "step": 7753
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.090351645921204e-05,
+      "loss": 18.5117,
+      "step": 7754
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.089440216645486e-05,
+      "loss": 20.0801,
+      "step": 7755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.088528749363853e-05,
+      "loss": 0.6861,
+      "step": 7756
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.087617244108107e-05,
+      "loss": 18.746,
+      "step": 7757
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.086705700910043e-05,
+      "loss": 0.7102,
+      "step": 7758
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0857941198014623e-05,
+      "loss": 0.7163,
+      "step": 7759
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0848825008141675e-05,
+      "loss": 0.737,
+      "step": 7760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083970843979957e-05,
+      "loss": 0.6318,
+      "step": 7761
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083059149330638e-05,
+      "loss": 0.7175,
+      "step": 7762
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.082147416898014e-05,
+      "loss": 18.879,
+      "step": 7763
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081235646713893e-05,
+      "loss": 20.3761,
+      "step": 7764
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0803238388100825e-05,
+      "loss": 19.8884,
+      "step": 7765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.079411993218389e-05,
+      "loss": 0.7206,
+      "step": 7766
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0785001099706265e-05,
+      "loss": 20.8484,
+      "step": 7767
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.077588189098603e-05,
+      "loss": 17.3764,
+      "step": 7768
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.076676230634136e-05,
+      "loss": 0.7715,
+      "step": 7769
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.075764234609036e-05,
+      "loss": 0.6947,
+      "step": 7770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0748522010551215e-05,
+      "loss": 21.0185,
+      "step": 7771
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0739401300042075e-05,
+      "loss": 19.8913,
+      "step": 7772
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.073028021488111e-05,
+      "loss": 0.6947,
+      "step": 7773
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072115875538655e-05,
+      "loss": 20.7151,
+      "step": 7774
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.071203692187657e-05,
+      "loss": 0.649,
+      "step": 7775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.070291471466942e-05,
+      "loss": 17.4318,
+      "step": 7776
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.069379213408334e-05,
+      "loss": 19.4377,
+      "step": 7777
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.068466918043654e-05,
+      "loss": 0.7175,
+      "step": 7778
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.06755458540473e-05,
+      "loss": 0.611,
+      "step": 7779
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0666422155233905e-05,
+      "loss": 18.663,
+      "step": 7780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.065729808431463e-05,
+      "loss": 15.5933,
+      "step": 7781
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064817364160777e-05,
+      "loss": 19.9504,
+      "step": 7782
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0639048827431654e-05,
+      "loss": 19.7848,
+      "step": 7783
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0629923642104605e-05,
+      "loss": 16.7692,
+      "step": 7784
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.062079808594494e-05,
+      "loss": 19.4286,
+      "step": 7785
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.061167215927105e-05,
+      "loss": 0.7002,
+      "step": 7786
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.060254586240126e-05,
+      "loss": 19.7366,
+      "step": 7787
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0593419195653966e-05,
+      "loss": 18.1392,
+      "step": 7788
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058429215934756e-05,
+      "loss": 18.3552,
+      "step": 7789
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.057516475380043e-05,
+      "loss": 0.6993,
+      "step": 7790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.056603697933102e-05,
+      "loss": 20.3203,
+      "step": 7791
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.055690883625773e-05,
+      "loss": 18.0909,
+      "step": 7792
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.054778032489902e-05,
+      "loss": 17.9469,
+      "step": 7793
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053865144557335e-05,
+      "loss": 22.2197,
+      "step": 7794
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.052952219859915e-05,
+      "loss": 18.1099,
+      "step": 7795
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.052039258429495e-05,
+      "loss": 0.651,
+      "step": 7796
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0511262602979215e-05,
+      "loss": 21.9223,
+      "step": 7797
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0502132254970455e-05,
+      "loss": 19.307,
+      "step": 7798
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04930015405872e-05,
+      "loss": 18.3503,
+      "step": 7799
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.048387046014795e-05,
+      "loss": 0.6038,
+      "step": 7800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.047473901397129e-05,
+      "loss": 0.6752,
+      "step": 7801
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.046560720237574e-05,
+      "loss": 18.1741,
+      "step": 7802
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04564750256799e-05,
+      "loss": 17.7232,
+      "step": 7803
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0447342484202354e-05,
+      "loss": 17.6846,
+      "step": 7804
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.043820957826166e-05,
+      "loss": 17.5391,
+      "step": 7805
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0429076308176466e-05,
+      "loss": 0.6978,
+      "step": 7806
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041994267426536e-05,
+      "loss": 0.7156,
+      "step": 7807
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041080867684702e-05,
+      "loss": 0.7032,
+      "step": 7808
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.040167431624005e-05,
+      "loss": 19.7557,
+      "step": 7809
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039253959276312e-05,
+      "loss": 0.6632,
+      "step": 7810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0383404506734906e-05,
+      "loss": 18.6361,
+      "step": 7811
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0374269058474083e-05,
+      "loss": 18.6834,
+      "step": 7812
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.036513324829938e-05,
+      "loss": 18.221,
+      "step": 7813
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0355997076529455e-05,
+      "loss": 0.6847,
+      "step": 7814
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.034686054348307e-05,
+      "loss": 17.848,
+      "step": 7815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.033772364947895e-05,
+      "loss": 19.939,
+      "step": 7816
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.032858639483583e-05,
+      "loss": 18.4742,
+      "step": 7817
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.031944877987249e-05,
+      "loss": 19.3848,
+      "step": 7818
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.031031080490768e-05,
+      "loss": 16.242,
+      "step": 7819
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.030117247026019e-05,
+      "loss": 19.4777,
+      "step": 7820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0292033776248845e-05,
+      "loss": 17.5668,
+      "step": 7821
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0282894723192416e-05,
+      "loss": 20.5559,
+      "step": 7822
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0273755311409755e-05,
+      "loss": 20.6687,
+      "step": 7823
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.026461554121967e-05,
+      "loss": 18.4752,
+      "step": 7824
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0255475412941034e-05,
+      "loss": 0.7181,
+      "step": 7825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0246334926892703e-05,
+      "loss": 19.9138,
+      "step": 7826
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.023719408339352e-05,
+      "loss": 0.7356,
+      "step": 7827
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.022805288276242e-05,
+      "loss": 0.6309,
+      "step": 7828
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.021891132531825e-05,
+      "loss": 0.6736,
+      "step": 7829
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020976941137996e-05,
+      "loss": 0.7769,
+      "step": 7830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020062714126644e-05,
+      "loss": 18.4239,
+      "step": 7831
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0191484515296645e-05,
+      "loss": 0.68,
+      "step": 7832
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.018234153378952e-05,
+      "loss": 20.179,
+      "step": 7833
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.017319819706403e-05,
+      "loss": 0.6786,
+      "step": 7834
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.016405450543912e-05,
+      "loss": 0.7156,
+      "step": 7835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01549104592338e-05,
+      "loss": 0.6402,
+      "step": 7836
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.014576605876706e-05,
+      "loss": 18.5245,
+      "step": 7837
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.013662130435791e-05,
+      "loss": 0.6803,
+      "step": 7838
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.012747619632538e-05,
+      "loss": 19.3064,
+      "step": 7839
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.011833073498847e-05,
+      "loss": 0.7343,
+      "step": 7840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0109184920666273e-05,
+      "loss": 21.2489,
+      "step": 7841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0100038753677815e-05,
+      "loss": 17.9463,
+      "step": 7842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.009089223434218e-05,
+      "loss": 0.6668,
+      "step": 7843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0081745362978433e-05,
+      "loss": 0.6974,
+      "step": 7844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.007259813990569e-05,
+      "loss": 18.9932,
+      "step": 7845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0063450565443066e-05,
+      "loss": 18.6875,
+      "step": 7846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0054302639909645e-05,
+      "loss": 19.929,
+      "step": 7847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.004515436362459e-05,
+      "loss": 18.3283,
+      "step": 7848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.003600573690704e-05,
+      "loss": 19.7438,
+      "step": 7849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0026856760076144e-05,
+      "loss": 20.9327,
+      "step": 7850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.001770743345108e-05,
+      "loss": 20.3544,
+      "step": 7851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0008557757351005e-05,
+      "loss": 18.7886,
+      "step": 7852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9999407732095156e-05,
+      "loss": 17.8434,
+      "step": 7853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9990257358002686e-05,
+      "loss": 0.7655,
+      "step": 7854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.998110663539286e-05,
+      "loss": 18.5648,
+      "step": 7855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997195556458488e-05,
+      "loss": 18.1751,
+      "step": 7856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9962804145897975e-05,
+      "loss": 17.1176,
+      "step": 7857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.995365237965144e-05,
+      "loss": 0.6572,
+      "step": 7858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9944500266164507e-05,
+      "loss": 18.5479,
+      "step": 7859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.993534780575647e-05,
+      "loss": 0.7796,
+      "step": 7860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9926194998746624e-05,
+      "loss": 16.7899,
+      "step": 7861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9917041845454235e-05,
+      "loss": 0.6834,
+      "step": 7862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.990788834619867e-05,
+      "loss": 0.6895,
+      "step": 7863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9898734501299214e-05,
+      "loss": 18.8111,
+      "step": 7864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.988958031107523e-05,
+      "loss": 0.7235,
+      "step": 7865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9880425775846074e-05,
+      "loss": 17.0707,
+      "step": 7866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.987127089593108e-05,
+      "loss": 0.6769,
+      "step": 7867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9862115671649644e-05,
+      "loss": 20.0028,
+      "step": 7868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9852960103321135e-05,
+      "loss": 18.4752,
+      "step": 7869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9843804191264975e-05,
+      "loss": 0.6991,
+      "step": 7870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.983464793580056e-05,
+      "loss": 0.7054,
+      "step": 7871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.982549133724731e-05,
+      "loss": 21.318,
+      "step": 7872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9816334395924665e-05,
+      "loss": 0.7212,
+      "step": 7873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.980717711215207e-05,
+      "loss": 19.3213,
+      "step": 7874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.979801948624899e-05,
+      "loss": 18.1239,
+      "step": 7875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.978886151853489e-05,
+      "loss": 20.7921,
+      "step": 7876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.977970320932924e-05,
+      "loss": 0.6101,
+      "step": 7877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.977054455895157e-05,
+      "loss": 0.718,
+      "step": 7878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.976138556772134e-05,
+      "loss": 0.7445,
+      "step": 7879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.975222623595812e-05,
+      "loss": 20.2734,
+      "step": 7880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9743066563981385e-05,
+      "loss": 20.0843,
+      "step": 7881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9733906552110716e-05,
+      "loss": 0.7292,
+      "step": 7882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.972474620066565e-05,
+      "loss": 19.704,
+      "step": 7883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.971558550996574e-05,
+      "loss": 0.6928,
+      "step": 7884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9706424480330606e-05,
+      "loss": 0.7527,
+      "step": 7885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.969726311207979e-05,
+      "loss": 18.7067,
+      "step": 7886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9688101405532925e-05,
+      "loss": 0.7178,
+      "step": 7887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.967893936100961e-05,
+      "loss": 0.6816,
+      "step": 7888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9669776978829475e-05,
+      "loss": 19.2879,
+      "step": 7889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9660614259312145e-05,
+      "loss": 17.199,
+      "step": 7890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.965145120277728e-05,
+      "loss": 0.6963,
+      "step": 7891
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.964228780954453e-05,
+      "loss": 18.0759,
+      "step": 7892
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.963312407993358e-05,
+      "loss": 0.7319,
+      "step": 7893
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9623960014264104e-05,
+      "loss": 0.6746,
+      "step": 7894
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.96147956128558e-05,
+      "loss": 17.556,
+      "step": 7895
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9605630876028375e-05,
+      "loss": 0.6241,
+      "step": 7896
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9596465804101545e-05,
+      "loss": 0.7304,
+      "step": 7897
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.958730039739504e-05,
+      "loss": 18.9268,
+      "step": 7898
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.95781346562286e-05,
+      "loss": 18.8442,
+      "step": 7899
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.956896858092199e-05,
+      "loss": 0.7239,
+      "step": 7900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9559802171794955e-05,
+      "loss": 0.7222,
+      "step": 7901
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.95506354291673e-05,
+      "loss": 20.6217,
+      "step": 7902
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.95414683533588e-05,
+      "loss": 18.806,
+      "step": 7903
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.953230094468923e-05,
+      "loss": 17.7206,
+      "step": 7904
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9523133203478445e-05,
+      "loss": 0.6942,
+      "step": 7905
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.951396513004623e-05,
+      "loss": 0.6899,
+      "step": 7906
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9504796724712454e-05,
+      "loss": 0.743,
+      "step": 7907
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9495627987796956e-05,
+      "loss": 18.6597,
+      "step": 7908
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.948645891961957e-05,
+      "loss": 18.7854,
+      "step": 7909
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.947728952050018e-05,
+      "loss": 0.7405,
+      "step": 7910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.946811979075866e-05,
+      "loss": 0.65,
+      "step": 7911
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.945894973071493e-05,
+      "loss": 19.1757,
+      "step": 7912
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944977934068885e-05,
+      "loss": 0.6688,
+      "step": 7913
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944060862100038e-05,
+      "loss": 18.4562,
+      "step": 7914
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.943143757196943e-05,
+      "loss": 0.7505,
+      "step": 7915
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9422266193915924e-05,
+      "loss": 0.6332,
+      "step": 7916
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9413094487159825e-05,
+      "loss": 17.6886,
+      "step": 7917
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9403922452021086e-05,
+      "loss": 18.6781,
+      "step": 7918
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9394750088819705e-05,
+      "loss": 17.0229,
+      "step": 7919
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.938557739787566e-05,
+      "loss": 18.9544,
+      "step": 7920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.937640437950891e-05,
+      "loss": 0.6487,
+      "step": 7921
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9367231034039504e-05,
+      "loss": 18.4787,
+      "step": 7922
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9358057361787436e-05,
+      "loss": 17.6357,
+      "step": 7923
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.934888336307276e-05,
+      "loss": 20.8312,
+      "step": 7924
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93397090382155e-05,
+      "loss": 21.0277,
+      "step": 7925
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93305343875357e-05,
+      "loss": 0.7693,
+      "step": 7926
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.932135941135345e-05,
+      "loss": 19.5533,
+      "step": 7927
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.931218410998881e-05,
+      "loss": 18.8321,
+      "step": 7928
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.930300848376187e-05,
+      "loss": 19.929,
+      "step": 7929
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.929383253299272e-05,
+      "loss": 0.7334,
+      "step": 7930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9284656258001484e-05,
+      "loss": 0.679,
+      "step": 7931
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.927547965910828e-05,
+      "loss": 19.8829,
+      "step": 7932
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.926630273663324e-05,
+      "loss": 18.0146,
+      "step": 7933
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9257125490896495e-05,
+      "loss": 20.8931,
+      "step": 7934
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.924794792221823e-05,
+      "loss": 20.7762,
+      "step": 7935
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.923877003091856e-05,
+      "loss": 20.0198,
+      "step": 7936
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922959181731772e-05,
+      "loss": 0.7296,
+      "step": 7937
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922041328173585e-05,
+      "loss": 18.1395,
+      "step": 7938
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.921123442449319e-05,
+      "loss": 18.4329,
+      "step": 7939
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9202055245909925e-05,
+      "loss": 0.6484,
+      "step": 7940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9192875746306275e-05,
+      "loss": 19.0693,
+      "step": 7941
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.91836959260025e-05,
+      "loss": 0.7211,
+      "step": 7942
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.917451578531882e-05,
+      "loss": 0.7334,
+      "step": 7943
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.91653353245755e-05,
+      "loss": 0.7726,
+      "step": 7944
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9156154544092815e-05,
+      "loss": 0.6707,
+      "step": 7945
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9146973444191024e-05,
+      "loss": 18.4872,
+      "step": 7946
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.913779202519043e-05,
+      "loss": 20.1472,
+      "step": 7947
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9128610287411326e-05,
+      "loss": 17.9863,
+      "step": 7948
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.911942823117403e-05,
+      "loss": 0.6819,
+      "step": 7949
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9110245856798876e-05,
+      "loss": 18.0706,
+      "step": 7950
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9101063164606164e-05,
+      "loss": 20.5113,
+      "step": 7951
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.909188015491627e-05,
+      "loss": 18.5996,
+      "step": 7952
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9082696828049535e-05,
+      "loss": 0.6618,
+      "step": 7953
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9073513184326335e-05,
+      "loss": 0.7414,
+      "step": 7954
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.906432922406705e-05,
+      "loss": 0.7007,
+      "step": 7955
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9055144947592045e-05,
+      "loss": 19.1641,
+      "step": 7956
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.904596035522175e-05,
+      "loss": 0.6829,
+      "step": 7957
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.903677544727655e-05,
+      "loss": 0.6623,
+      "step": 7958
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90275902240769e-05,
+      "loss": 0.6613,
+      "step": 7959
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9018404685943206e-05,
+      "loss": 0.727,
+      "step": 7960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.900921883319591e-05,
+      "loss": 0.7039,
+      "step": 7961
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9000032666155504e-05,
+      "loss": 19.3178,
+      "step": 7962
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.89908461851424e-05,
+      "loss": 20.5938,
+      "step": 7963
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.898165939047712e-05,
+      "loss": 19.7118,
+      "step": 7964
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8972472282480116e-05,
+      "loss": 18.5378,
+      "step": 7965
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8963284861471926e-05,
+      "loss": 0.6876,
+      "step": 7966
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.895409712777303e-05,
+      "loss": 17.9689,
+      "step": 7967
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.894490908170395e-05,
+      "loss": 19.8263,
+      "step": 7968
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.893572072358524e-05,
+      "loss": 16.9565,
+      "step": 7969
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.892653205373742e-05,
+      "loss": 17.7723,
+      "step": 7970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8917343072481047e-05,
+      "loss": 18.5909,
+      "step": 7971
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.89081537801367e-05,
+      "loss": 0.7278,
+      "step": 7972
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.889896417702494e-05,
+      "loss": 0.7401,
+      "step": 7973
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8889774263466355e-05,
+      "loss": 18.6235,
+      "step": 7974
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.888058403978154e-05,
+      "loss": 17.9467,
+      "step": 7975
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8871393506291116e-05,
+      "loss": 18.2281,
+      "step": 7976
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.88622026633157e-05,
+      "loss": 0.6612,
+      "step": 7977
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.885301151117589e-05,
+      "loss": 0.6884,
+      "step": 7978
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.884382005019237e-05,
+      "loss": 0.665,
+      "step": 7979
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.883462828068576e-05,
+      "loss": 0.7329,
+      "step": 7980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.882543620297673e-05,
+      "loss": 0.7409,
+      "step": 7981
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.881624381738595e-05,
+      "loss": 0.7379,
+      "step": 7982
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.880705112423411e-05,
+      "loss": 0.6874,
+      "step": 7983
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.879785812384191e-05,
+      "loss": 18.7957,
+      "step": 7984
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.878866481653003e-05,
+      "loss": 18.8253,
+      "step": 7985
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.877947120261922e-05,
+      "loss": 0.6585,
+      "step": 7986
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.877027728243017e-05,
+      "loss": 18.7347,
+      "step": 7987
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.876108305628364e-05,
+      "loss": 0.6722,
+      "step": 7988
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8751888524500375e-05,
+      "loss": 18.7847,
+      "step": 7989
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8742693687401115e-05,
+      "loss": 18.5234,
+      "step": 7990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8733498545306656e-05,
+      "loss": 0.6957,
+      "step": 7991
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8724303098537746e-05,
+      "loss": 19.3958,
+      "step": 7992
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.87151073474152e-05,
+      "loss": 18.812,
+      "step": 7993
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.870591129225982e-05,
+      "loss": 0.6915,
+      "step": 7994
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.869671493339238e-05,
+      "loss": 0.6647,
+      "step": 7995
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8687518271133755e-05,
+      "loss": 0.7393,
+      "step": 7996
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8678321305804726e-05,
+      "loss": 0.6462,
+      "step": 7997
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8669124037726186e-05,
+      "loss": 18.405,
+      "step": 7998
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8659926467218964e-05,
+      "loss": 19.4485,
+      "step": 7999
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8650728594603896e-05,
+      "loss": 20.3737,
+      "step": 8000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8641530420201905e-05,
+      "loss": 21.243,
+      "step": 8001
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.863233194433384e-05,
+      "loss": 20.2716,
+      "step": 8002
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.862313316732063e-05,
+      "loss": 19.3982,
+      "step": 8003
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.861393408948316e-05,
+      "loss": 0.6935,
+      "step": 8004
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860473471114233e-05,
+      "loss": 0.7251,
+      "step": 8005
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.85955350326191e-05,
+      "loss": 18.5634,
+      "step": 8006
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8586335054234375e-05,
+      "loss": 18.1238,
+      "step": 8007
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857713477630914e-05,
+      "loss": 0.6412,
+      "step": 8008
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.856793419916433e-05,
+      "loss": 0.6783,
+      "step": 8009
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.855873332312092e-05,
+      "loss": 18.7251,
+      "step": 8010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8549532148499885e-05,
+      "loss": 0.7143,
+      "step": 8011
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.854033067562221e-05,
+      "loss": 17.428,
+      "step": 8012
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8531128904808905e-05,
+      "loss": 17.5582,
+      "step": 8013
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.852192683638099e-05,
+      "loss": 19.2372,
+      "step": 8014
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851272447065945e-05,
+      "loss": 0.7282,
+      "step": 8015
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8503521807965366e-05,
+      "loss": 0.7266,
+      "step": 8016
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.849431884861975e-05,
+      "loss": 0.6791,
+      "step": 8017
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.848511559294365e-05,
+      "loss": 17.7794,
+      "step": 8018
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.847591204125812e-05,
+      "loss": 0.7348,
+      "step": 8019
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8466708193884265e-05,
+      "loss": 17.4032,
+      "step": 8020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.845750405114315e-05,
+      "loss": 0.7325,
+      "step": 8021
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.844829961335586e-05,
+      "loss": 18.9797,
+      "step": 8022
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.843909488084351e-05,
+      "loss": 0.7315,
+      "step": 8023
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.842988985392721e-05,
+      "loss": 18.3408,
+      "step": 8024
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8420684532928096e-05,
+      "loss": 18.2609,
+      "step": 8025
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8411478918167284e-05,
+      "loss": 18.1634,
+      "step": 8026
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8402273009965926e-05,
+      "loss": 0.7012,
+      "step": 8027
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.839306680864518e-05,
+      "loss": 0.676,
+      "step": 8028
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838386031452621e-05,
+      "loss": 19.4823,
+      "step": 8029
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.837465352793018e-05,
+      "loss": 0.6865,
+      "step": 8030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8365446449178287e-05,
+      "loss": 17.5245,
+      "step": 8031
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8356239078591724e-05,
+      "loss": 0.7001,
+      "step": 8032
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8347031416491705e-05,
+      "loss": 19.0208,
+      "step": 8033
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8337823463199424e-05,
+      "loss": 0.7089,
+      "step": 8034
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.832861521903613e-05,
+      "loss": 17.6144,
+      "step": 8035
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8319406684323054e-05,
+      "loss": 18.2511,
+      "step": 8036
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831019785938142e-05,
+      "loss": 21.086,
+      "step": 8037
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8300988744532525e-05,
+      "loss": 0.7514,
+      "step": 8038
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8291779340097605e-05,
+      "loss": 0.7097,
+      "step": 8039
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8282569646397946e-05,
+      "loss": 19.4285,
+      "step": 8040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.827335966375484e-05,
+      "loss": 18.6124,
+      "step": 8041
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8264149392489566e-05,
+      "loss": 0.6879,
+      "step": 8042
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825493883292345e-05,
+      "loss": 18.7253,
+      "step": 8043
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.824572798537782e-05,
+      "loss": 18.7583,
+      "step": 8044
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823651685017397e-05,
+      "loss": 18.6666,
+      "step": 8045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.822730542763325e-05,
+      "loss": 0.6807,
+      "step": 8046
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8218093718077004e-05,
+      "loss": 18.7983,
+      "step": 8047
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8208881721826616e-05,
+      "loss": 17.0564,
+      "step": 8048
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.819966943920341e-05,
+      "loss": 18.7785,
+      "step": 8049
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81904568705288e-05,
+      "loss": 17.0608,
+      "step": 8050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.818124401612416e-05,
+      "loss": 0.7386,
+      "step": 8051
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8172030876310865e-05,
+      "loss": 0.7254,
+      "step": 8052
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8162817451410364e-05,
+      "loss": 0.6529,
+      "step": 8053
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8153603741744035e-05,
+      "loss": 19.803,
+      "step": 8054
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.814438974763334e-05,
+      "loss": 19.6551,
+      "step": 8055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8135175469399696e-05,
+      "loss": 18.9269,
+      "step": 8056
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.812596090736454e-05,
+      "loss": 17.0922,
+      "step": 8057
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.811674606184936e-05,
+      "loss": 19.0099,
+      "step": 8058
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.810753093317558e-05,
+      "loss": 17.9546,
+      "step": 8059
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.809831552166472e-05,
+      "loss": 19.1682,
+      "step": 8060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.808909982763825e-05,
+      "loss": 16.7837,
+      "step": 8061
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8079883851417637e-05,
+      "loss": 0.6906,
+      "step": 8062
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.807066759332444e-05,
+      "loss": 0.7047,
+      "step": 8063
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.806145105368013e-05,
+      "loss": 17.6623,
+      "step": 8064
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.805223423280626e-05,
+      "loss": 0.6571,
+      "step": 8065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.804301713102435e-05,
+      "loss": 18.1231,
+      "step": 8066
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803379974865595e-05,
+      "loss": 18.5637,
+      "step": 8067
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.802458208602262e-05,
+      "loss": 0.6581,
+      "step": 8068
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.8015364143445906e-05,
+      "loss": 17.9079,
+      "step": 8069
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.800614592124741e-05,
+      "loss": 0.6986,
+      "step": 8070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7996927419748716e-05,
+      "loss": 0.6682,
+      "step": 8071
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.79877086392714e-05,
+      "loss": 18.6174,
+      "step": 8072
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7978489580137064e-05,
+      "loss": 0.6493,
+      "step": 8073
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.796927024266733e-05,
+      "loss": 19.135,
+      "step": 8074
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7960050627183815e-05,
+      "loss": 18.357,
+      "step": 8075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795083073400817e-05,
+      "loss": 0.7101,
+      "step": 8076
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.794161056346201e-05,
+      "loss": 0.6956,
+      "step": 8077
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.793239011586702e-05,
+      "loss": 17.7705,
+      "step": 8078
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.792316939154483e-05,
+      "loss": 20.5968,
+      "step": 8079
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.791394839081714e-05,
+      "loss": 17.4246,
+      "step": 8080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7904727114005595e-05,
+      "loss": 17.4555,
+      "step": 8081
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.789550556143193e-05,
+      "loss": 21.3058,
+      "step": 8082
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7886283733417825e-05,
+      "loss": 19.588,
+      "step": 8083
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.787706163028498e-05,
+      "loss": 18.6241,
+      "step": 8084
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7867839252355125e-05,
+      "loss": 18.635,
+      "step": 8085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.785861659994998e-05,
+      "loss": 18.6,
+      "step": 8086
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7849393673391305e-05,
+      "loss": 17.5294,
+      "step": 8087
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.784017047300084e-05,
+      "loss": 18.2595,
+      "step": 8088
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.783094699910032e-05,
+      "loss": 0.7323,
+      "step": 8089
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782172325201155e-05,
+      "loss": 18.0541,
+      "step": 8090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.781249923205627e-05,
+      "loss": 17.6548,
+      "step": 8091
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7803274939556306e-05,
+      "loss": 19.0461,
+      "step": 8092
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.779405037483342e-05,
+      "loss": 0.7218,
+      "step": 8093
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.778482553820943e-05,
+      "loss": 17.8093,
+      "step": 8094
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.777560043000616e-05,
+      "loss": 19.9312,
+      "step": 8095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.776637505054542e-05,
+      "loss": 19.1968,
+      "step": 8096
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.775714940014905e-05,
+      "loss": 0.702,
+      "step": 8097
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.77479234791389e-05,
+      "loss": 0.7281,
+      "step": 8098
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.773869728783682e-05,
+      "loss": 17.8716,
+      "step": 8099
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.772947082656467e-05,
+      "loss": 19.2091,
+      "step": 8100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.772024409564431e-05,
+      "loss": 0.746,
+      "step": 8101
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.771101709539764e-05,
+      "loss": 18.8736,
+      "step": 8102
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.770178982614654e-05,
+      "loss": 19.641,
+      "step": 8103
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.76925622882129e-05,
+      "loss": 17.3768,
+      "step": 8104
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7683334481918674e-05,
+      "loss": 0.6815,
+      "step": 8105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7674106407585715e-05,
+      "loss": 0.6963,
+      "step": 8106
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.766487806553601e-05,
+      "loss": 18.0121,
+      "step": 8107
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7655649456091455e-05,
+      "loss": 0.7579,
+      "step": 8108
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7646420579574026e-05,
+      "loss": 0.7468,
+      "step": 8109
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.763719143630566e-05,
+      "loss": 0.6363,
+      "step": 8110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.762796202660833e-05,
+      "loss": 18.0229,
+      "step": 8111
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7618732350804006e-05,
+      "loss": 18.9356,
+      "step": 8112
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.760950240921468e-05,
+      "loss": 19.7448,
+      "step": 8113
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7600272202162344e-05,
+      "loss": 19.1586,
+      "step": 8114
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7591041729968986e-05,
+      "loss": 18.9901,
+      "step": 8115
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.758181099295663e-05,
+      "loss": 18.08,
+      "step": 8116
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.757257999144731e-05,
+      "loss": 0.6983,
+      "step": 8117
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.756334872576303e-05,
+      "loss": 0.6427,
+      "step": 8118
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7554117196225846e-05,
+      "loss": 19.3548,
+      "step": 8119
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.75448854031578e-05,
+      "loss": 18.1909,
+      "step": 8120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7535653346880934e-05,
+      "loss": 0.6855,
+      "step": 8121
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752642102771736e-05,
+      "loss": 18.9037,
+      "step": 8122
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7517188445989124e-05,
+      "loss": 0.7098,
+      "step": 8123
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7507955602018295e-05,
+      "loss": 18.8107,
+      "step": 8124
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7498722496126997e-05,
+      "loss": 18.6602,
+      "step": 8125
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7489489128637344e-05,
+      "loss": 18.3082,
+      "step": 8126
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.748025549987141e-05,
+      "loss": 20.6374,
+      "step": 8127
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7471021610151324e-05,
+      "loss": 0.7285,
+      "step": 8128
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.746178745979924e-05,
+      "loss": 19.199,
+      "step": 8129
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.745255304913728e-05,
+      "loss": 18.3787,
+      "step": 8130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.74433183784876e-05,
+      "loss": 20.6316,
+      "step": 8131
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743408344817236e-05,
+      "loss": 0.6865,
+      "step": 8132
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.742484825851372e-05,
+      "loss": 0.7291,
+      "step": 8133
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7415612809833875e-05,
+      "loss": 0.7141,
+      "step": 8134
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.740637710245498e-05,
+      "loss": 20.4637,
+      "step": 8135
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739714113669925e-05,
+      "loss": 18.1163,
+      "step": 8136
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.738790491288888e-05,
+      "loss": 0.6954,
+      "step": 8137
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7378668431346086e-05,
+      "loss": 0.6708,
+      "step": 8138
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7369431692393095e-05,
+      "loss": 0.6943,
+      "step": 8139
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.736019469635213e-05,
+      "loss": 0.7136,
+      "step": 8140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.735095744354543e-05,
+      "loss": 0.7182,
+      "step": 8141
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.734171993429525e-05,
+      "loss": 18.3975,
+      "step": 8142
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7332482168923826e-05,
+      "loss": 19.495,
+      "step": 8143
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732324414775346e-05,
+      "loss": 0.7044,
+      "step": 8144
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.731400587110641e-05,
+      "loss": 0.7558,
+      "step": 8145
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730476733930494e-05,
+      "loss": 0.7515,
+      "step": 8146
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.729552855267138e-05,
+      "loss": 18.3293,
+      "step": 8147
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.728628951152799e-05,
+      "loss": 19.3348,
+      "step": 8148
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7277050216197114e-05,
+      "loss": 19.2943,
+      "step": 8149
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7267810667001076e-05,
+      "loss": 0.729,
+      "step": 8150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7258570864262165e-05,
+      "loss": 19.1625,
+      "step": 8151
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724933080830276e-05,
+      "loss": 19.5094,
+      "step": 8152
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724009049944518e-05,
+      "loss": 0.7243,
+      "step": 8153
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.72308499380118e-05,
+      "loss": 0.6843,
+      "step": 8154
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.722160912432495e-05,
+      "loss": 0.689,
+      "step": 8155
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721236805870705e-05,
+      "loss": 17.9762,
+      "step": 8156
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.720312674148044e-05,
+      "loss": 0.7388,
+      "step": 8157
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719388517296753e-05,
+      "loss": 0.6734,
+      "step": 8158
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.718464335349072e-05,
+      "loss": 18.8024,
+      "step": 8159
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.71754012833724e-05,
+      "loss": 18.9756,
+      "step": 8160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.716615896293501e-05,
+      "loss": 18.8101,
+      "step": 8161
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.715691639250097e-05,
+      "loss": 18.4629,
+      "step": 8162
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.71476735723927e-05,
+      "loss": 0.6682,
+      "step": 8163
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713843050293265e-05,
+      "loss": 0.7155,
+      "step": 8164
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.712918718444327e-05,
+      "loss": 19.6347,
+      "step": 8165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.711994361724703e-05,
+      "loss": 18.1224,
+      "step": 8166
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.711069980166639e-05,
+      "loss": 18.6118,
+      "step": 8167
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710145573802381e-05,
+      "loss": 0.6807,
+      "step": 8168
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.709221142664182e-05,
+      "loss": 18.868,
+      "step": 8169
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708296686784287e-05,
+      "loss": 18.3184,
+      "step": 8170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7073722061949485e-05,
+      "loss": 0.6648,
+      "step": 8171
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.706447700928418e-05,
+      "loss": 0.7436,
+      "step": 8172
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.705523171016945e-05,
+      "loss": 20.066,
+      "step": 8173
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7045986164927856e-05,
+      "loss": 17.3041,
+      "step": 8174
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.703674037388191e-05,
+      "loss": 0.7369,
+      "step": 8175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7027494337354194e-05,
+      "loss": 0.737,
+      "step": 8176
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.701824805566722e-05,
+      "loss": 0.6772,
+      "step": 8177
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.700900152914357e-05,
+      "loss": 18.1031,
+      "step": 8178
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6999754758105814e-05,
+      "loss": 19.3312,
+      "step": 8179
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.699050774287654e-05,
+      "loss": 0.7249,
+      "step": 8180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.698126048377834e-05,
+      "loss": 0.6989,
+      "step": 8181
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697201298113378e-05,
+      "loss": 0.6956,
+      "step": 8182
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6962765235265516e-05,
+      "loss": 18.6814,
+      "step": 8183
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695351724649611e-05,
+      "loss": 20.2618,
+      "step": 8184
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.694426901514821e-05,
+      "loss": 18.247,
+      "step": 8185
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.693502054154446e-05,
+      "loss": 19.4437,
+      "step": 8186
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6925771826007466e-05,
+      "loss": 18.2091,
+      "step": 8187
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6916522868859926e-05,
+      "loss": 16.8597,
+      "step": 8188
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6907273670424446e-05,
+      "loss": 20.5609,
+      "step": 8189
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6898024231023705e-05,
+      "loss": 17.7071,
+      "step": 8190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6888774550980396e-05,
+      "loss": 0.7277,
+      "step": 8191
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.687952463061719e-05,
+      "loss": 18.6753,
+      "step": 8192
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.687027447025677e-05,
+      "loss": 19.3334,
+      "step": 8193
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.686102407022183e-05,
+      "loss": 18.8006,
+      "step": 8194
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.68517734308351e-05,
+      "loss": 0.6411,
+      "step": 8195
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.684252255241927e-05,
+      "loss": 16.3394,
+      "step": 8196
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6833271435297087e-05,
+      "loss": 19.0303,
+      "step": 8197
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6824020079791274e-05,
+      "loss": 18.703,
+      "step": 8198
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6814768486224555e-05,
+      "loss": 17.8773,
+      "step": 8199
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.68055166549197e-05,
+      "loss": 0.6934,
+      "step": 8200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.679626458619946e-05,
+      "loss": 18.1731,
+      "step": 8201
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6787012280386584e-05,
+      "loss": 19.3665,
+      "step": 8202
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.677775973780388e-05,
+      "loss": 17.6463,
+      "step": 8203
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.67685069587741e-05,
+      "loss": 19.2706,
+      "step": 8204
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675925394362004e-05,
+      "loss": 19.3091,
+      "step": 8205
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675000069266451e-05,
+      "loss": 0.6933,
+      "step": 8206
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6740747206230295e-05,
+      "loss": 0.6982,
+      "step": 8207
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.673149348464024e-05,
+      "loss": 18.2192,
+      "step": 8208
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6722239528217136e-05,
+      "loss": 18.3401,
+      "step": 8209
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6712985337283833e-05,
+      "loss": 0.6747,
+      "step": 8210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6703730912163167e-05,
+      "loss": 19.433,
+      "step": 8211
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669447625317798e-05,
+      "loss": 19.6422,
+      "step": 8212
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.668522136065113e-05,
+      "loss": 0.7078,
+      "step": 8213
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.667596623490548e-05,
+      "loss": 18.7992,
+      "step": 8214
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66667108762639e-05,
+      "loss": 0.7456,
+      "step": 8215
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665745528504929e-05,
+      "loss": 18.1625,
+      "step": 8216
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.664819946158451e-05,
+      "loss": 18.0041,
+      "step": 8217
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.663894340619247e-05,
+      "loss": 19.2196,
+      "step": 8218
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.662968711919607e-05,
+      "loss": 17.8733,
+      "step": 8219
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.662043060091822e-05,
+      "loss": 0.6769,
+      "step": 8220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.661117385168184e-05,
+      "loss": 0.6798,
+      "step": 8221
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.660191687180987e-05,
+      "loss": 18.7284,
+      "step": 8222
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.659265966162525e-05,
+      "loss": 17.6899,
+      "step": 8223
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65834022214509e-05,
+      "loss": 0.7213,
+      "step": 8224
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.657414455160979e-05,
+      "loss": 18.0898,
+      "step": 8225
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.656488665242488e-05,
+      "loss": 0.6826,
+      "step": 8226
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.655562852421912e-05,
+      "loss": 18.5418,
+      "step": 8227
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.654637016731552e-05,
+      "loss": 18.3114,
+      "step": 8228
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.653711158203704e-05,
+      "loss": 19.3834,
+      "step": 8229
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6527852768706665e-05,
+      "loss": 0.7537,
+      "step": 8230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6518593727647426e-05,
+      "loss": 0.7281,
+      "step": 8231
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.650933445918232e-05,
+      "loss": 18.9844,
+      "step": 8232
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6500074963634344e-05,
+      "loss": 17.3572,
+      "step": 8233
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.649081524132654e-05,
+      "loss": 0.6864,
+      "step": 8234
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6481555292581946e-05,
+      "loss": 17.8243,
+      "step": 8235
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6472295117723584e-05,
+      "loss": 0.7316,
+      "step": 8236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.64630347170745e-05,
+      "loss": 18.9615,
+      "step": 8237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.645377409095778e-05,
+      "loss": 0.6729,
+      "step": 8238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.644451323969645e-05,
+      "loss": 18.19,
+      "step": 8239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.64352521636136e-05,
+      "loss": 19.4515,
+      "step": 8240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6425990863032326e-05,
+      "loss": 21.0077,
+      "step": 8241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641672933827569e-05,
+      "loss": 18.3231,
+      "step": 8242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6407467589666785e-05,
+      "loss": 18.9563,
+      "step": 8243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.639820561752872e-05,
+      "loss": 0.6858,
+      "step": 8244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.638894342218463e-05,
+      "loss": 18.4387,
+      "step": 8245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.637968100395761e-05,
+      "loss": 20.1971,
+      "step": 8246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6370418363170764e-05,
+      "loss": 0.7208,
+      "step": 8247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.636115550014727e-05,
+      "loss": 0.7187,
+      "step": 8248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635189241521024e-05,
+      "loss": 19.7364,
+      "step": 8249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.634262910868284e-05,
+      "loss": 18.3736,
+      "step": 8250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.633336558088823e-05,
+      "loss": 0.696,
+      "step": 8251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6324101832149556e-05,
+      "loss": 19.1358,
+      "step": 8252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.631483786279e-05,
+      "loss": 0.7633,
+      "step": 8253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630557367313274e-05,
+      "loss": 17.6121,
+      "step": 8254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.629630926350098e-05,
+      "loss": 18.4611,
+      "step": 8255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628704463421789e-05,
+      "loss": 0.7212,
+      "step": 8256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.627777978560669e-05,
+      "loss": 18.4833,
+      "step": 8257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6268514717990585e-05,
+      "loss": 18.0743,
+      "step": 8258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.625924943169279e-05,
+      "loss": 19.4664,
+      "step": 8259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624998392703654e-05,
+      "loss": 16.9664,
+      "step": 8260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624071820434508e-05,
+      "loss": 19.4483,
+      "step": 8261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6231452263941606e-05,
+      "loss": 17.4324,
+      "step": 8262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6222186106149444e-05,
+      "loss": 17.4101,
+      "step": 8263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.621291973129177e-05,
+      "loss": 18.3299,
+      "step": 8264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.620365313969189e-05,
+      "loss": 0.6809,
+      "step": 8265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.619438633167306e-05,
+      "loss": 0.6934,
+      "step": 8266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6185119307558575e-05,
+      "loss": 19.7722,
+      "step": 8267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.617585206767172e-05,
+      "loss": 0.6563,
+      "step": 8268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.616658461233578e-05,
+      "loss": 0.7436,
+      "step": 8269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615731694187406e-05,
+      "loss": 0.7195,
+      "step": 8270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.614804905660986e-05,
+      "loss": 20.516,
+      "step": 8271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6138780956866534e-05,
+      "loss": 19.0055,
+      "step": 8272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.612951264296737e-05,
+      "loss": 18.795,
+      "step": 8273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.612024411523569e-05,
+      "loss": 19.0946,
+      "step": 8274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6110975373994865e-05,
+      "loss": 18.0359,
+      "step": 8275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6101706419568224e-05,
+      "loss": 17.9344,
+      "step": 8276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.609243725227914e-05,
+      "loss": 0.7002,
+      "step": 8277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.608316787245096e-05,
+      "loss": 18.0798,
+      "step": 8278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6073898280407045e-05,
+      "loss": 16.572,
+      "step": 8279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.606462847647078e-05,
+      "loss": 18.8934,
+      "step": 8280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.605535846096556e-05,
+      "loss": 18.2256,
+      "step": 8281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.604608823421475e-05,
+      "loss": 0.6982,
+      "step": 8282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6036817796541786e-05,
+      "loss": 16.9852,
+      "step": 8283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.602754714827004e-05,
+      "loss": 19.3704,
+      "step": 8284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.601827628972294e-05,
+      "loss": 17.9289,
+      "step": 8285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.60090052212239e-05,
+      "loss": 18.508,
+      "step": 8286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5999733943096365e-05,
+      "loss": 18.8998,
+      "step": 8287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599046245566375e-05,
+      "loss": 20.5684,
+      "step": 8288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.59811907592495e-05,
+      "loss": 0.6602,
+      "step": 8289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597191885417709e-05,
+      "loss": 18.4674,
+      "step": 8290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.596264674076993e-05,
+      "loss": 17.8241,
+      "step": 8291
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5953374419351525e-05,
+      "loss": 17.9783,
+      "step": 8292
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5944101890245324e-05,
+      "loss": 0.7063,
+      "step": 8293
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5934829153774825e-05,
+      "loss": 0.6637,
+      "step": 8294
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5925556210263507e-05,
+      "loss": 0.6752,
+      "step": 8295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5916283060034855e-05,
+      "loss": 18.2105,
+      "step": 8296
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.590700970341238e-05,
+      "loss": 17.1977,
+      "step": 8297
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.589773614071957e-05,
+      "loss": 0.7024,
+      "step": 8298
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5888462372279974e-05,
+      "loss": 0.6772,
+      "step": 8299
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.587918839841709e-05,
+      "loss": 0.6956,
+      "step": 8300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.586991421945444e-05,
+      "loss": 0.6444,
+      "step": 8301
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.586063983571559e-05,
+      "loss": 20.2534,
+      "step": 8302
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.585136524752404e-05,
+      "loss": 19.1226,
+      "step": 8303
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.58420904552034e-05,
+      "loss": 0.7272,
+      "step": 8304
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.583281545907718e-05,
+      "loss": 18.2201,
+      "step": 8305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5823540259468954e-05,
+      "loss": 17.3431,
+      "step": 8306
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5814264856702316e-05,
+      "loss": 0.681,
+      "step": 8307
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.580498925110081e-05,
+      "loss": 20.7367,
+      "step": 8308
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.579571344298806e-05,
+      "loss": 0.6833,
+      "step": 8309
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.578643743268763e-05,
+      "loss": 0.6769,
+      "step": 8310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5777161220523136e-05,
+      "loss": 0.7458,
+      "step": 8311
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.576788480681818e-05,
+      "loss": 19.3176,
+      "step": 8312
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.575860819189638e-05,
+      "loss": 0.7327,
+      "step": 8313
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574933137608134e-05,
+      "loss": 19.447,
+      "step": 8314
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574005435969673e-05,
+      "loss": 0.6911,
+      "step": 8315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5730777143066124e-05,
+      "loss": 19.9076,
+      "step": 8316
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5721499726513226e-05,
+      "loss": 0.6811,
+      "step": 8317
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.571222211036164e-05,
+      "loss": 18.5116,
+      "step": 8318
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5702944294935055e-05,
+      "loss": 0.6651,
+      "step": 8319
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5693666280557125e-05,
+      "loss": 17.9045,
+      "step": 8320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.568438806755149e-05,
+      "loss": 20.6084,
+      "step": 8321
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5675109656241876e-05,
+      "loss": 0.6642,
+      "step": 8322
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.566583104695192e-05,
+      "loss": 17.9687,
+      "step": 8323
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5656552240005356e-05,
+      "loss": 16.962,
+      "step": 8324
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.564727323572586e-05,
+      "loss": 19.7944,
+      "step": 8325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5637994034437136e-05,
+      "loss": 17.6118,
+      "step": 8326
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.56287146364629e-05,
+      "loss": 17.2088,
+      "step": 8327
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561943504212685e-05,
+      "loss": 0.641,
+      "step": 8328
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561015525175276e-05,
+      "loss": 0.7046,
+      "step": 8329
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.560087526566433e-05,
+      "loss": 17.9857,
+      "step": 8330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559159508418529e-05,
+      "loss": 19.9516,
+      "step": 8331
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5582314707639395e-05,
+      "loss": 17.2327,
+      "step": 8332
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.55730341363504e-05,
+      "loss": 17.9822,
+      "step": 8333
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.556375337064208e-05,
+      "loss": 0.6846,
+      "step": 8334
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.555447241083819e-05,
+      "loss": 19.7608,
+      "step": 8335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.554519125726247e-05,
+      "loss": 18.7294,
+      "step": 8336
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.553590991023876e-05,
+      "loss": 0.7541,
+      "step": 8337
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552662837009079e-05,
+      "loss": 17.6819,
+      "step": 8338
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5517346637142395e-05,
+      "loss": 0.6407,
+      "step": 8339
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550806471171735e-05,
+      "loss": 19.9138,
+      "step": 8340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.549878259413948e-05,
+      "loss": 0.6593,
+      "step": 8341
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.548950028473259e-05,
+      "loss": 18.2536,
+      "step": 8342
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5480217783820475e-05,
+      "loss": 16.9152,
+      "step": 8343
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.547093509172703e-05,
+      "loss": 16.7349,
+      "step": 8344
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5461652208776e-05,
+      "loss": 17.8701,
+      "step": 8345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.545236913529131e-05,
+      "loss": 0.6843,
+      "step": 8346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.544308587159674e-05,
+      "loss": 0.7186,
+      "step": 8347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.543380241801617e-05,
+      "loss": 0.6411,
+      "step": 8348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.542451877487346e-05,
+      "loss": 18.8783,
+      "step": 8349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.541523494249248e-05,
+      "loss": 18.7259,
+      "step": 8350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.540595092119709e-05,
+      "loss": 16.4699,
+      "step": 8351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5396666711311196e-05,
+      "loss": 19.8046,
+      "step": 8352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.538738231315865e-05,
+      "loss": 17.0105,
+      "step": 8353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5378097727063384e-05,
+      "loss": 17.328,
+      "step": 8354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.536881295334926e-05,
+      "loss": 18.3174,
+      "step": 8355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.53595279923402e-05,
+      "loss": 18.8833,
+      "step": 8356
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5350242844360124e-05,
+      "loss": 19.7203,
+      "step": 8357
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5340957509732935e-05,
+      "loss": 16.764,
+      "step": 8358
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.533167198878259e-05,
+      "loss": 17.2561,
+      "step": 8359
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5322386281832973e-05,
+      "loss": 19.0385,
+      "step": 8360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.531310038920805e-05,
+      "loss": 0.6895,
+      "step": 8361
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.530381431123178e-05,
+      "loss": 0.6883,
+      "step": 8362
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.529452804822809e-05,
+      "loss": 19.8535,
+      "step": 8363
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5285241600520945e-05,
+      "loss": 19.691,
+      "step": 8364
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527595496843431e-05,
+      "loss": 0.7437,
+      "step": 8365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.526666815229217e-05,
+      "loss": 19.5818,
+      "step": 8366
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.525738115241849e-05,
+      "loss": 0.6377,
+      "step": 8367
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5248093969137224e-05,
+      "loss": 17.2874,
+      "step": 8368
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5238806602772417e-05,
+      "loss": 17.2058,
+      "step": 8369
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.522951905364804e-05,
+      "loss": 17.4735,
+      "step": 8370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.522023132208809e-05,
+      "loss": 18.6452,
+      "step": 8371
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5210943408416595e-05,
+      "loss": 0.6903,
+      "step": 8372
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5201655312957546e-05,
+      "loss": 0.6611,
+      "step": 8373
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.519236703603498e-05,
+      "loss": 19.4909,
+      "step": 8374
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5183078577972915e-05,
+      "loss": 19.6631,
+      "step": 8375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51737899390954e-05,
+      "loss": 18.101,
+      "step": 8376
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.516450111972645e-05,
+      "loss": 18.9913,
+      "step": 8377
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5155212120190156e-05,
+      "loss": 16.0419,
+      "step": 8378
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5145922940810543e-05,
+      "loss": 0.6882,
+      "step": 8379
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5136633581911655e-05,
+      "loss": 0.6742,
+      "step": 8380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5127344043817594e-05,
+      "loss": 0.7341,
+      "step": 8381
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.511805432685241e-05,
+      "loss": 0.7024,
+      "step": 8382
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.510876443134019e-05,
+      "loss": 16.6616,
+      "step": 8383
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509947435760502e-05,
+      "loss": 19.5024,
+      "step": 8384
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509018410597098e-05,
+      "loss": 19.1623,
+      "step": 8385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.508089367676217e-05,
+      "loss": 17.4666,
+      "step": 8386
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.507160307030269e-05,
+      "loss": 17.7926,
+      "step": 8387
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.506231228691667e-05,
+      "loss": 17.1063,
+      "step": 8388
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.505302132692822e-05,
+      "loss": 0.7122,
+      "step": 8389
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5043730190661436e-05,
+      "loss": 17.7085,
+      "step": 8390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5034438878440475e-05,
+      "loss": 20.0091,
+      "step": 8391
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.502514739058945e-05,
+      "loss": 0.6748,
+      "step": 8392
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.501585572743253e-05,
+      "loss": 0.6842,
+      "step": 8393
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.500656388929384e-05,
+      "loss": 19.4269,
+      "step": 8394
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4997271876497514e-05,
+      "loss": 17.5602,
+      "step": 8395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.498797968936775e-05,
+      "loss": 18.8224,
+      "step": 8396
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.497868732822868e-05,
+      "loss": 17.1328,
+      "step": 8397
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496939479340452e-05,
+      "loss": 0.6829,
+      "step": 8398
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496010208521939e-05,
+      "loss": 0.7434,
+      "step": 8399
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.495080920399751e-05,
+      "loss": 19.3756,
+      "step": 8400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.494151615006307e-05,
+      "loss": 0.6694,
+      "step": 8401
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4932222923740226e-05,
+      "loss": 0.6827,
+      "step": 8402
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.492292952535323e-05,
+      "loss": 20.0905,
+      "step": 8403
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491363595522624e-05,
+      "loss": 0.6016,
+      "step": 8404
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.490434221368351e-05,
+      "loss": 19.1429,
+      "step": 8405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4895048301049245e-05,
+      "loss": 17.5783,
+      "step": 8406
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.488575421764765e-05,
+      "loss": 0.6904,
+      "step": 8407
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.487645996380301e-05,
+      "loss": 19.3088,
+      "step": 8408
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.486716553983951e-05,
+      "loss": 21.1018,
+      "step": 8409
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.485787094608138e-05,
+      "loss": 17.9954,
+      "step": 8410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.484857618285293e-05,
+      "loss": 0.7052,
+      "step": 8411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.483928125047836e-05,
+      "loss": 0.6814,
+      "step": 8412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4829986149281964e-05,
+      "loss": 19.2583,
+      "step": 8413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4820690879588e-05,
+      "loss": 0.652,
+      "step": 8414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4811395441720724e-05,
+      "loss": 16.2767,
+      "step": 8415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480209983600444e-05,
+      "loss": 18.7453,
+      "step": 8416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4792804062763406e-05,
+      "loss": 0.7427,
+      "step": 8417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478350812232192e-05,
+      "loss": 0.6606,
+      "step": 8418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.47742120150043e-05,
+      "loss": 16.8568,
+      "step": 8419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476491574113482e-05,
+      "loss": 20.6544,
+      "step": 8420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.47556193010378e-05,
+      "loss": 0.6619,
+      "step": 8421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.474632269503754e-05,
+      "loss": 19.1649,
+      "step": 8422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.473702592345838e-05,
+      "loss": 19.2551,
+      "step": 8423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.472772898662464e-05,
+      "loss": 17.3516,
+      "step": 8424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4718431884860625e-05,
+      "loss": 0.7219,
+      "step": 8425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.470913461849071e-05,
+      "loss": 17.8231,
+      "step": 8426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.46998371878392e-05,
+      "loss": 18.0914,
+      "step": 8427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469053959323046e-05,
+      "loss": 17.4297,
+      "step": 8428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4681241834988836e-05,
+      "loss": 19.5474,
+      "step": 8429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4671943913438716e-05,
+      "loss": 17.3659,
+      "step": 8430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4662645828904434e-05,
+      "loss": 0.6876,
+      "step": 8431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4653347581710355e-05,
+      "loss": 0.7298,
+      "step": 8432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.464404917218088e-05,
+      "loss": 18.0569,
+      "step": 8433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.463475060064037e-05,
+      "loss": 18.5385,
+      "step": 8434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.462545186741324e-05,
+      "loss": 0.7217,
+      "step": 8435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4616152972823855e-05,
+      "loss": 17.8186,
+      "step": 8436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.460685391719662e-05,
+      "loss": 18.8513,
+      "step": 8437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4597554700855946e-05,
+      "loss": 17.895,
+      "step": 8438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4588255324126235e-05,
+      "loss": 20.1712,
+      "step": 8439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.457895578733192e-05,
+      "loss": 20.0856,
+      "step": 8440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4569656090797405e-05,
+      "loss": 0.6859,
+      "step": 8441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.456035623484712e-05,
+      "loss": 18.6146,
+      "step": 8442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45510562198055e-05,
+      "loss": 17.8565,
+      "step": 8443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454175604599697e-05,
+      "loss": 19.0419,
+      "step": 8444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4532455713746e-05,
+      "loss": 19.2494,
+      "step": 8445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.452315522337702e-05,
+      "loss": 19.8739,
+      "step": 8446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.451385457521447e-05,
+      "loss": 20.401,
+      "step": 8447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4504553769582855e-05,
+      "loss": 16.2217,
+      "step": 8448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4495252806806596e-05,
+      "loss": 19.3831,
+      "step": 8449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.448595168721019e-05,
+      "loss": 20.6111,
+      "step": 8450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44766504111181e-05,
+      "loss": 0.744,
+      "step": 8451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4467348978854805e-05,
+      "loss": 18.1217,
+      "step": 8452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4458047390744804e-05,
+      "loss": 17.2148,
+      "step": 8453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4448745647112595e-05,
+      "loss": 17.9055,
+      "step": 8454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443944374828266e-05,
+      "loss": 0.7057,
+      "step": 8455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4430141694579497e-05,
+      "loss": 0.6556,
+      "step": 8456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.442083948632763e-05,
+      "loss": 0.8011,
+      "step": 8457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4411537123851586e-05,
+      "loss": 18.748,
+      "step": 8458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4402234607475853e-05,
+      "loss": 0.7495,
+      "step": 8459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.439293193752497e-05,
+      "loss": 18.4964,
+      "step": 8460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.438362911432347e-05,
+      "loss": 18.4364,
+      "step": 8461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4374326138195895e-05,
+      "loss": 18.0779,
+      "step": 8462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4365023009466774e-05,
+      "loss": 19.9267,
+      "step": 8463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.435571972846065e-05,
+      "loss": 0.6205,
+      "step": 8464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.43464162955021e-05,
+      "loss": 0.6816,
+      "step": 8465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.433711271091565e-05,
+      "loss": 18.0823,
+      "step": 8466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.432780897502589e-05,
+      "loss": 19.9168,
+      "step": 8467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.431850508815736e-05,
+      "loss": 18.4389,
+      "step": 8468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430920105063464e-05,
+      "loss": 19.707,
+      "step": 8469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4299896862782326e-05,
+      "loss": 0.7102,
+      "step": 8470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4290592524924976e-05,
+      "loss": 0.7112,
+      "step": 8471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.428128803738721e-05,
+      "loss": 19.761,
+      "step": 8472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4271983400493595e-05,
+      "loss": 20.1683,
+      "step": 8473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426267861456874e-05,
+      "loss": 17.9738,
+      "step": 8474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.425337367993726e-05,
+      "loss": 18.1583,
+      "step": 8475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4244068596923725e-05,
+      "loss": 17.8052,
+      "step": 8476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4234763365852805e-05,
+      "loss": 0.6751,
+      "step": 8477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4225457987049086e-05,
+      "loss": 18.5965,
+      "step": 8478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4216152460837186e-05,
+      "loss": 18.0397,
+      "step": 8479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4206846787541754e-05,
+      "loss": 20.7256,
+      "step": 8480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.419754096748742e-05,
+      "loss": 20.5553,
+      "step": 8481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.418823500099884e-05,
+      "loss": 0.7019,
+      "step": 8482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4178928888400614e-05,
+      "loss": 0.7328,
+      "step": 8483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4169622630017445e-05,
+      "loss": 20.8162,
+      "step": 8484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.416031622617396e-05,
+      "loss": 18.2898,
+      "step": 8485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4151009677194806e-05,
+      "loss": 18.9508,
+      "step": 8486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4141702983404685e-05,
+      "loss": 18.7585,
+      "step": 8487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.413239614512824e-05,
+      "loss": 0.6934,
+      "step": 8488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.412308916269017e-05,
+      "loss": 19.8725,
+      "step": 8489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.411378203641514e-05,
+      "loss": 18.6231,
+      "step": 8490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.410447476662782e-05,
+      "loss": 18.2246,
+      "step": 8491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4095167353652945e-05,
+      "loss": 16.7699,
+      "step": 8492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4085859797815154e-05,
+      "loss": 19.312,
+      "step": 8493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.407655209943921e-05,
+      "loss": 18.359,
+      "step": 8494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.406724425884977e-05,
+      "loss": 19.2502,
+      "step": 8495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4057936276371565e-05,
+      "loss": 0.7012,
+      "step": 8496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4048628152329314e-05,
+      "loss": 0.7256,
+      "step": 8497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403931988704772e-05,
+      "loss": 19.2954,
+      "step": 8498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403001148085154e-05,
+      "loss": 19.4281,
+      "step": 8499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.402070293406547e-05,
+      "loss": 20.8779,
+      "step": 8500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.401139424701427e-05,
+      "loss": 16.9005,
+      "step": 8501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.400208542002266e-05,
+      "loss": 18.3727,
+      "step": 8502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3992776453415405e-05,
+      "loss": 0.6867,
+      "step": 8503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3983467347517234e-05,
+      "loss": 0.7101,
+      "step": 8504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.397415810265294e-05,
+      "loss": 18.2129,
+      "step": 8505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.396484871914722e-05,
+      "loss": 0.7173,
+      "step": 8506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39555391973249e-05,
+      "loss": 17.953,
+      "step": 8507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.394622953751073e-05,
+      "loss": 19.7287,
+      "step": 8508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.393691974002947e-05,
+      "loss": 19.0634,
+      "step": 8509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.392760980520591e-05,
+      "loss": 0.6289,
+      "step": 8510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.391829973336482e-05,
+      "loss": 19.178,
+      "step": 8511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390898952483101e-05,
+      "loss": 19.7552,
+      "step": 8512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3899679179929264e-05,
+      "loss": 0.7129,
+      "step": 8513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.389036869898438e-05,
+      "loss": 0.6716,
+      "step": 8514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.388105808232117e-05,
+      "loss": 0.7706,
+      "step": 8515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.387174733026441e-05,
+      "loss": 18.4317,
+      "step": 8516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3862436443138954e-05,
+      "loss": 17.0674,
+      "step": 8517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3853125421269576e-05,
+      "loss": 18.6798,
+      "step": 8518
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.384381426498115e-05,
+      "loss": 0.7206,
+      "step": 8519
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.383450297459847e-05,
+      "loss": 18.0318,
+      "step": 8520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.382519155044635e-05,
+      "loss": 0.6991,
+      "step": 8521
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.381587999284966e-05,
+      "loss": 19.0209,
+      "step": 8522
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3806568302133223e-05,
+      "loss": 0.6801,
+      "step": 8523
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37972564786219e-05,
+      "loss": 20.5012,
+      "step": 8524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.378794452264053e-05,
+      "loss": 0.6449,
+      "step": 8525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.377863243451395e-05,
+      "loss": 0.6824,
+      "step": 8526
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376932021456704e-05,
+      "loss": 19.034,
+      "step": 8527
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376000786312466e-05,
+      "loss": 0.6716,
+      "step": 8528
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.375069538051168e-05,
+      "loss": 17.9111,
+      "step": 8529
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3741382767052984e-05,
+      "loss": 20.5102,
+      "step": 8530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.373207002307341e-05,
+      "loss": 18.9766,
+      "step": 8531
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3722757148897885e-05,
+      "loss": 0.725,
+      "step": 8532
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.371344414485127e-05,
+      "loss": 18.305,
+      "step": 8533
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3704131011258455e-05,
+      "loss": 19.8209,
+      "step": 8534
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.369481774844435e-05,
+      "loss": 0.6938,
+      "step": 8535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.368550435673385e-05,
+      "loss": 0.6878,
+      "step": 8536
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3676190836451844e-05,
+      "loss": 0.7274,
+      "step": 8537
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.366687718792326e-05,
+      "loss": 0.6966,
+      "step": 8538
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3657563411473e-05,
+      "loss": 19.0654,
+      "step": 8539
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.364824950742598e-05,
+      "loss": 19.5405,
+      "step": 8540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3638935476107145e-05,
+      "loss": 19.7814,
+      "step": 8541
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.362962131784139e-05,
+      "loss": 0.6991,
+      "step": 8542
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3620307032953666e-05,
+      "loss": 20.2353,
+      "step": 8543
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.361099262176891e-05,
+      "loss": 17.2897,
+      "step": 8544
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.360167808461205e-05,
+      "loss": 16.8647,
+      "step": 8545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3592363421808046e-05,
+      "loss": 18.1294,
+      "step": 8546
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.358304863368183e-05,
+      "loss": 17.502,
+      "step": 8547
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.357373372055835e-05,
+      "loss": 0.6819,
+      "step": 8548
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35644186827626e-05,
+      "loss": 0.6904,
+      "step": 8549
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35551035206195e-05,
+      "loss": 0.722,
+      "step": 8550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.354578823445404e-05,
+      "loss": 0.669,
+      "step": 8551
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.353647282459119e-05,
+      "loss": 19.5335,
+      "step": 8552
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35271572913559e-05,
+      "loss": 18.1157,
+      "step": 8553
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.351784163507319e-05,
+      "loss": 21.3038,
+      "step": 8554
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3508525856068e-05,
+      "loss": 0.6895,
+      "step": 8555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3499209954665365e-05,
+      "loss": 17.9904,
+      "step": 8556
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3489893931190236e-05,
+      "loss": 19.606,
+      "step": 8557
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.348057778596762e-05,
+      "loss": 19.0849,
+      "step": 8558
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347126151932254e-05,
+      "loss": 17.6624,
+      "step": 8559
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.346194513157996e-05,
+      "loss": 19.6016,
+      "step": 8560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3452628623064934e-05,
+      "loss": 0.7205,
+      "step": 8561
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.344331199410245e-05,
+      "loss": 17.3236,
+      "step": 8562
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343399524501753e-05,
+      "loss": 0.7349,
+      "step": 8563
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.342467837613518e-05,
+      "loss": 19.4093,
+      "step": 8564
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341536138778045e-05,
+      "loss": 19.4412,
+      "step": 8565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3406044280278354e-05,
+      "loss": 0.6501,
+      "step": 8566
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.339672705395393e-05,
+      "loss": 0.7601,
+      "step": 8567
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.338740970913223e-05,
+      "loss": 20.8999,
+      "step": 8568
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3378092246138275e-05,
+      "loss": 19.3337,
+      "step": 8569
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3368774665297114e-05,
+      "loss": 18.6183,
+      "step": 8570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.335945696693382e-05,
+      "loss": 19.1198,
+      "step": 8571
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.335013915137341e-05,
+      "loss": 0.7519,
+      "step": 8572
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334082121894098e-05,
+      "loss": 18.7315,
+      "step": 8573
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.333150316996158e-05,
+      "loss": 17.5919,
+      "step": 8574
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332218500476026e-05,
+      "loss": 17.2236,
+      "step": 8575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.331286672366213e-05,
+      "loss": 0.6941,
+      "step": 8576
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.330354832699222e-05,
+      "loss": 0.6253,
+      "step": 8577
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3294229815075636e-05,
+      "loss": 19.8903,
+      "step": 8578
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328491118823745e-05,
+      "loss": 0.7939,
+      "step": 8579
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.327559244680276e-05,
+      "loss": 0.6876,
+      "step": 8580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.326627359109665e-05,
+      "loss": 17.2985,
+      "step": 8581
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3256954621444197e-05,
+      "loss": 17.363,
+      "step": 8582
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.324763553817054e-05,
+      "loss": 17.671,
+      "step": 8583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3238316341600756e-05,
+      "loss": 18.5129,
+      "step": 8584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.322899703205996e-05,
+      "loss": 0.7449,
+      "step": 8585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3219677609873255e-05,
+      "loss": 17.9058,
+      "step": 8586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.321035807536575e-05,
+      "loss": 19.3245,
+      "step": 8587
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.320103842886259e-05,
+      "loss": 0.7167,
+      "step": 8588
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3191718670688884e-05,
+      "loss": 17.4912,
+      "step": 8589
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.318239880116974e-05,
+      "loss": 0.6637,
+      "step": 8590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3173078820630314e-05,
+      "loss": 17.3356,
+      "step": 8591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.316375872939573e-05,
+      "loss": 18.8489,
+      "step": 8592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3154438527791126e-05,
+      "loss": 0.7288,
+      "step": 8593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314511821614164e-05,
+      "loss": 0.6784,
+      "step": 8594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3135797794772436e-05,
+      "loss": 17.7539,
+      "step": 8595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312647726400865e-05,
+      "loss": 0.7642,
+      "step": 8596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3117156624175414e-05,
+      "loss": 17.8293,
+      "step": 8597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.310783587559792e-05,
+      "loss": 0.6708,
+      "step": 8598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3098515018601326e-05,
+      "loss": 17.8646,
+      "step": 8599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3089194053510775e-05,
+      "loss": 19.0752,
+      "step": 8600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.307987298065145e-05,
+      "loss": 0.6532,
+      "step": 8601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.307055180034851e-05,
+      "loss": 0.6707,
+      "step": 8602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3061230512927166e-05,
+      "loss": 17.7361,
+      "step": 8603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.305190911871256e-05,
+      "loss": 0.7793,
+      "step": 8604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3042587618029896e-05,
+      "loss": 17.6777,
+      "step": 8605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3033266011204355e-05,
+      "loss": 0.6472,
+      "step": 8606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3023944298561115e-05,
+      "loss": 18.594,
+      "step": 8607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.301462248042539e-05,
+      "loss": 0.675,
+      "step": 8608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.300530055712238e-05,
+      "loss": 17.904,
+      "step": 8609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.299597852897727e-05,
+      "loss": 0.7618,
+      "step": 8610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.298665639631528e-05,
+      "loss": 17.8097,
+      "step": 8611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2977334159461614e-05,
+      "loss": 0.7502,
+      "step": 8612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.296801181874148e-05,
+      "loss": 16.1987,
+      "step": 8613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.295868937448012e-05,
+      "loss": 0.6819,
+      "step": 8614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294936682700271e-05,
+      "loss": 20.3756,
+      "step": 8615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294004417663451e-05,
+      "loss": 20.209,
+      "step": 8616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2930721423700736e-05,
+      "loss": 19.4723,
+      "step": 8617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2921398568526616e-05,
+      "loss": 19.1392,
+      "step": 8618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.291207561143738e-05,
+      "loss": 18.3884,
+      "step": 8619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2902752552758285e-05,
+      "loss": 18.7723,
+      "step": 8620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.289342939281456e-05,
+      "loss": 20.2106,
+      "step": 8621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.288410613193143e-05,
+      "loss": 18.8744,
+      "step": 8622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2874782770434194e-05,
+      "loss": 17.8969,
+      "step": 8623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.286545930864806e-05,
+      "loss": 20.0007,
+      "step": 8624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.28561357468983e-05,
+      "loss": 18.2414,
+      "step": 8625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2846812085510175e-05,
+      "loss": 0.6556,
+      "step": 8626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2837488324808937e-05,
+      "loss": 0.7024,
+      "step": 8627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2828164465119864e-05,
+      "loss": 17.6936,
+      "step": 8628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.281884050676822e-05,
+      "loss": 16.8889,
+      "step": 8629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.280951645007928e-05,
+      "loss": 18.8131,
+      "step": 8630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.280019229537833e-05,
+      "loss": 0.6723,
+      "step": 8631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.279086804299062e-05,
+      "loss": 0.6501,
+      "step": 8632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2781543693241465e-05,
+      "loss": 0.6622,
+      "step": 8633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2772219246456136e-05,
+      "loss": 0.711,
+      "step": 8634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2762894702959934e-05,
+      "loss": 17.8234,
+      "step": 8635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.275357006307814e-05,
+      "loss": 19.4191,
+      "step": 8636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2744245327136056e-05,
+      "loss": 16.3211,
+      "step": 8637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2734920495458983e-05,
+      "loss": 20.0304,
+      "step": 8638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.272559556837222e-05,
+      "loss": 17.7052,
+      "step": 8639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2716270546201085e-05,
+      "loss": 0.6926,
+      "step": 8640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.270694542927088e-05,
+      "loss": 0.7082,
+      "step": 8641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.269762021790692e-05,
+      "loss": 18.4235,
+      "step": 8642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.268829491243452e-05,
+      "loss": 18.6554,
+      "step": 8643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.267896951317899e-05,
+      "loss": 0.6657,
+      "step": 8644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.266964402046569e-05,
+      "loss": 19.1689,
+      "step": 8645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2660318434619894e-05,
+      "loss": 0.7049,
+      "step": 8646
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.265099275596698e-05,
+      "loss": 19.5101,
+      "step": 8647
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.264166698483225e-05,
+      "loss": 0.6676,
+      "step": 8648
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.263234112154104e-05,
+      "loss": 19.9858,
+      "step": 8649
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.262301516641872e-05,
+      "loss": 16.2156,
+      "step": 8650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.26136891197906e-05,
+      "loss": 18.8192,
+      "step": 8651
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2604362981982037e-05,
+      "loss": 0.6732,
+      "step": 8652
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.259503675331839e-05,
+      "loss": 18.9216,
+      "step": 8653
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258571043412499e-05,
+      "loss": 18.5678,
+      "step": 8654
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.257638402472722e-05,
+      "loss": 17.5587,
+      "step": 8655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256705752545041e-05,
+      "loss": 0.6745,
+      "step": 8656
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.255773093661995e-05,
+      "loss": 0.7237,
+      "step": 8657
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.254840425856119e-05,
+      "loss": 0.7326,
+      "step": 8658
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.253907749159949e-05,
+      "loss": 18.6734,
+      "step": 8659
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2529750636060225e-05,
+      "loss": 18.7637,
+      "step": 8660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2520423692268775e-05,
+      "loss": 18.63,
+      "step": 8661
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251109666055052e-05,
+      "loss": 0.6629,
+      "step": 8662
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.250176954123084e-05,
+      "loss": 17.4626,
+      "step": 8663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2492442334635105e-05,
+      "loss": 0.7058,
+      "step": 8664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2483115041088715e-05,
+      "loss": 18.7904,
+      "step": 8665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.247378766091705e-05,
+      "loss": 20.1955,
+      "step": 8666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2464460194445516e-05,
+      "loss": 18.0675,
+      "step": 8667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2455132641999494e-05,
+      "loss": 0.7064,
+      "step": 8668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2445805003904394e-05,
+      "loss": 19.0007,
+      "step": 8669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2436477280485605e-05,
+      "loss": 17.7929,
+      "step": 8670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2427149472068537e-05,
+      "loss": 18.8088,
+      "step": 8671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.241782157897861e-05,
+      "loss": 18.209,
+      "step": 8672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240849360154121e-05,
+      "loss": 18.4761,
+      "step": 8673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2399165540081776e-05,
+      "loss": 17.9791,
+      "step": 8674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.23898373949257e-05,
+      "loss": 17.8698,
+      "step": 8675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2380509166398405e-05,
+      "loss": 19.4014,
+      "step": 8676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.237118085482534e-05,
+      "loss": 0.6628,
+      "step": 8677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236185246053189e-05,
+      "loss": 17.5804,
+      "step": 8678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.235252398384353e-05,
+      "loss": 17.1753,
+      "step": 8679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234319542508564e-05,
+      "loss": 0.7273,
+      "step": 8680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2333866784583686e-05,
+      "loss": 0.7109,
+      "step": 8681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2324538062663085e-05,
+      "loss": 0.6505,
+      "step": 8682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.23152092596493e-05,
+      "loss": 18.5315,
+      "step": 8683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2305880375867755e-05,
+      "loss": 19.3349,
+      "step": 8684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.22965514116439e-05,
+      "loss": 0.7094,
+      "step": 8685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2287222367303176e-05,
+      "loss": 18.3415,
+      "step": 8686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.227789324317106e-05,
+      "loss": 21.4477,
+      "step": 8687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2268564039572964e-05,
+      "loss": 0.73,
+      "step": 8688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2259234756834376e-05,
+      "loss": 0.6918,
+      "step": 8689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.224990539528074e-05,
+      "loss": 19.1565,
+      "step": 8690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2240575955237524e-05,
+      "loss": 0.6834,
+      "step": 8691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.223124643703019e-05,
+      "loss": 20.4279,
+      "step": 8692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.222191684098421e-05,
+      "loss": 0.6622,
+      "step": 8693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2212587167425034e-05,
+      "loss": 0.6721,
+      "step": 8694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2203257416678164e-05,
+      "loss": 0.6536,
+      "step": 8695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2193927589069045e-05,
+      "loss": 0.687,
+      "step": 8696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218459768492319e-05,
+      "loss": 18.8616,
+      "step": 8697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.217526770456604e-05,
+      "loss": 18.2896,
+      "step": 8698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.216593764832311e-05,
+      "loss": 20.6075,
+      "step": 8699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2156607516519885e-05,
+      "loss": 19.19,
+      "step": 8700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2147277309481815e-05,
+      "loss": 19.0417,
+      "step": 8701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.213794702753443e-05,
+      "loss": 0.688,
+      "step": 8702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.212861667100321e-05,
+      "loss": 0.7046,
+      "step": 8703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.211928624021365e-05,
+      "loss": 18.3256,
+      "step": 8704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210995573549127e-05,
+      "loss": 17.9125,
+      "step": 8705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210062515716152e-05,
+      "loss": 16.3474,
+      "step": 8706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.209129450554996e-05,
+      "loss": 0.6664,
+      "step": 8707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.208196378098207e-05,
+      "loss": 0.7103,
+      "step": 8708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2072632983783355e-05,
+      "loss": 20.6448,
+      "step": 8709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2063302114279346e-05,
+      "loss": 19.2492,
+      "step": 8710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2053971172795535e-05,
+      "loss": 0.7045,
+      "step": 8711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2044640159657466e-05,
+      "loss": 17.9767,
+      "step": 8712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.203530907519062e-05,
+      "loss": 20.5852,
+      "step": 8713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.202597791972055e-05,
+      "loss": 18.6767,
+      "step": 8714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.201664669357278e-05,
+      "loss": 0.6503,
+      "step": 8715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.200731539707281e-05,
+      "loss": 18.5913,
+      "step": 8716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1997984030546185e-05,
+      "loss": 18.396,
+      "step": 8717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1988652594318446e-05,
+      "loss": 0.6852,
+      "step": 8718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.197932108871512e-05,
+      "loss": 18.0875,
+      "step": 8719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1969989514061736e-05,
+      "loss": 0.7237,
+      "step": 8720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1960657870683836e-05,
+      "loss": 0.6903,
+      "step": 8721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1951326158906975e-05,
+      "loss": 18.508,
+      "step": 8722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194199437905667e-05,
+      "loss": 18.2579,
+      "step": 8723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19326625314585e-05,
+      "loss": 0.6359,
+      "step": 8724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192333061643798e-05,
+      "loss": 19.1852,
+      "step": 8725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.191399863432067e-05,
+      "loss": 17.4182,
+      "step": 8726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190466658543216e-05,
+      "loss": 0.6208,
+      "step": 8727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189533447009794e-05,
+      "loss": 17.6882,
+      "step": 8728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.188600228864362e-05,
+      "loss": 19.3667,
+      "step": 8729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.187667004139474e-05,
+      "loss": 0.6553,
+      "step": 8730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1867337728676856e-05,
+      "loss": 19.4566,
+      "step": 8731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185800535081555e-05,
+      "loss": 18.5681,
+      "step": 8732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1848672908136366e-05,
+      "loss": 0.679,
+      "step": 8733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.18393404009649e-05,
+      "loss": 17.6049,
+      "step": 8734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1830007829626694e-05,
+      "loss": 19.4184,
+      "step": 8735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182067519444735e-05,
+      "loss": 17.9975,
+      "step": 8736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.181134249575244e-05,
+      "loss": 0.7184,
+      "step": 8737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.180200973386751e-05,
+      "loss": 0.6907,
+      "step": 8738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1792676909118176e-05,
+      "loss": 0.6448,
+      "step": 8739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1783344021829994e-05,
+      "loss": 0.6938,
+      "step": 8740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1774011072328576e-05,
+      "loss": 20.145,
+      "step": 8741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176467806093949e-05,
+      "loss": 0.7268,
+      "step": 8742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175534498798832e-05,
+      "loss": 0.7398,
+      "step": 8743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.174601185380068e-05,
+      "loss": 17.7417,
+      "step": 8744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.173667865870214e-05,
+      "loss": 17.8298,
+      "step": 8745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1727345403018315e-05,
+      "loss": 0.6939,
+      "step": 8746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1718012087074785e-05,
+      "loss": 17.9342,
+      "step": 8747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1708678711197155e-05,
+      "loss": 0.6953,
+      "step": 8748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169934527571103e-05,
+      "loss": 17.7524,
+      "step": 8749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169001178094201e-05,
+      "loss": 0.6486,
+      "step": 8750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.168067822721571e-05,
+      "loss": 16.2875,
+      "step": 8751
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.167134461485773e-05,
+      "loss": 16.4394,
+      "step": 8752
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.166201094419367e-05,
+      "loss": 0.6912,
+      "step": 8753
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1652677215549175e-05,
+      "loss": 20.8785,
+      "step": 8754
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1643343429249814e-05,
+      "loss": 0.6658,
+      "step": 8755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.163400958562124e-05,
+      "loss": 17.4519,
+      "step": 8756
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.162467568498903e-05,
+      "loss": 0.7176,
+      "step": 8757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.161534172767885e-05,
+      "loss": 18.2488,
+      "step": 8758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.16060077140163e-05,
+      "loss": 19.2311,
+      "step": 8759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1596673644327e-05,
+      "loss": 18.7782,
+      "step": 8760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1587339518936585e-05,
+      "loss": 0.6738,
+      "step": 8761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1578005338170674e-05,
+      "loss": 0.7186,
+      "step": 8762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1568671102354906e-05,
+      "loss": 0.6931,
+      "step": 8763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1559336811814914e-05,
+      "loss": 17.7295,
+      "step": 8764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.155000246687631e-05,
+      "loss": 17.8699,
+      "step": 8765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.154066806786476e-05,
+      "loss": 17.177,
+      "step": 8766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153133361510588e-05,
+      "loss": 0.7294,
+      "step": 8767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.152199910892532e-05,
+      "loss": 16.5518,
+      "step": 8768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.151266454964871e-05,
+      "loss": 0.6961,
+      "step": 8769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1503329937601706e-05,
+      "loss": 0.7747,
+      "step": 8770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.149399527310995e-05,
+      "loss": 18.8088,
+      "step": 8771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.148466055649906e-05,
+      "loss": 19.4251,
+      "step": 8772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.147532578809474e-05,
+      "loss": 17.964,
+      "step": 8773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.14659909682226e-05,
+      "loss": 18.205,
+      "step": 8774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.145665609720829e-05,
+      "loss": 0.6836,
+      "step": 8775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1447321175377496e-05,
+      "loss": 16.3869,
+      "step": 8776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.143798620305583e-05,
+      "loss": 19.5342,
+      "step": 8777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.142865118056899e-05,
+      "loss": 18.094,
+      "step": 8778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.141931610824261e-05,
+      "loss": 0.7153,
+      "step": 8779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.140998098640235e-05,
+      "loss": 18.3414,
+      "step": 8780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.140064581537389e-05,
+      "loss": 17.3006,
+      "step": 8781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1391310595482876e-05,
+      "loss": 19.4101,
+      "step": 8782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1381975327055e-05,
+      "loss": 0.756,
+      "step": 8783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.137264001041588e-05,
+      "loss": 18.8825,
+      "step": 8784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.136330464589125e-05,
+      "loss": 19.2168,
+      "step": 8785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.135396923380673e-05,
+      "loss": 18.5499,
+      "step": 8786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.134463377448802e-05,
+      "loss": 0.6976,
+      "step": 8787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.133529826826078e-05,
+      "loss": 19.5502,
+      "step": 8788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.132596271545068e-05,
+      "loss": 17.8532,
+      "step": 8789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131662711638342e-05,
+      "loss": 17.3366,
+      "step": 8790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1307291471384675e-05,
+      "loss": 18.3038,
+      "step": 8791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12979557807801e-05,
+      "loss": 17.145,
+      "step": 8792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.128862004489542e-05,
+      "loss": 20.2715,
+      "step": 8793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.127928426405628e-05,
+      "loss": 0.7371,
+      "step": 8794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1269948438588374e-05,
+      "loss": 18.5274,
+      "step": 8795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12606125688174e-05,
+      "loss": 17.2543,
+      "step": 8796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1251276655069046e-05,
+      "loss": 0.6926,
+      "step": 8797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1241940697669e-05,
+      "loss": 17.8736,
+      "step": 8798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.123260469694294e-05,
+      "loss": 0.6903,
+      "step": 8799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122326865321658e-05,
+      "loss": 18.5626,
+      "step": 8800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1213932566815605e-05,
+      "loss": 19.2137,
+      "step": 8801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12045964380657e-05,
+      "loss": 0.7407,
+      "step": 8802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.119526026729259e-05,
+      "loss": 16.6356,
+      "step": 8803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118592405482196e-05,
+      "loss": 18.2813,
+      "step": 8804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.117658780097949e-05,
+      "loss": 19.4262,
+      "step": 8805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.116725150609092e-05,
+      "loss": 18.5752,
+      "step": 8806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1157915170481916e-05,
+      "loss": 0.75,
+      "step": 8807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1148578794478205e-05,
+      "loss": 18.8739,
+      "step": 8808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1139242378405506e-05,
+      "loss": 0.6997,
+      "step": 8809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1129905922589506e-05,
+      "loss": 19.657,
+      "step": 8810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.112056942735591e-05,
+      "loss": 18.2337,
+      "step": 8811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1111232893030435e-05,
+      "loss": 17.7259,
+      "step": 8812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.11018963199388e-05,
+      "loss": 19.7183,
+      "step": 8813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1092559708406695e-05,
+      "loss": 18.2578,
+      "step": 8814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.108322305875988e-05,
+      "loss": 0.6914,
+      "step": 8815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.107388637132403e-05,
+      "loss": 17.9976,
+      "step": 8816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.106454964642486e-05,
+      "loss": 17.5123,
+      "step": 8817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.105521288438813e-05,
+      "loss": 0.7139,
+      "step": 8818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.10458760855395e-05,
+      "loss": 0.6522,
+      "step": 8819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.103653925020475e-05,
+      "loss": 17.2431,
+      "step": 8820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102720237870957e-05,
+      "loss": 18.3666,
+      "step": 8821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1017865471379686e-05,
+      "loss": 0.7199,
+      "step": 8822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1008528528540835e-05,
+      "loss": 17.2424,
+      "step": 8823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0999191550518723e-05,
+      "loss": 0.6917,
+      "step": 8824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.098985453763909e-05,
+      "loss": 18.6747,
+      "step": 8825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0980517490227675e-05,
+      "loss": 21.1595,
+      "step": 8826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0971180408610174e-05,
+      "loss": 17.7988,
+      "step": 8827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0961843293112365e-05,
+      "loss": 19.4041,
+      "step": 8828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.095250614405993e-05,
+      "loss": 0.6744,
+      "step": 8829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.094316896177864e-05,
+      "loss": 20.9657,
+      "step": 8830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093383174659423e-05,
+      "loss": 17.3112,
+      "step": 8831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.09244944988324e-05,
+      "loss": 18.1957,
+      "step": 8832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091515721881892e-05,
+      "loss": 19.7834,
+      "step": 8833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0905819906879506e-05,
+      "loss": 17.5269,
+      "step": 8834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089648256333992e-05,
+      "loss": 18.6184,
+      "step": 8835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0887145188525884e-05,
+      "loss": 0.6876,
+      "step": 8836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0877807782763145e-05,
+      "loss": 18.1296,
+      "step": 8837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.086847034637745e-05,
+      "loss": 19.1063,
+      "step": 8838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.085913287969453e-05,
+      "loss": 17.5749,
+      "step": 8839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084979538304013e-05,
+      "loss": 0.6781,
+      "step": 8840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084045785674001e-05,
+      "loss": 18.9054,
+      "step": 8841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0831120301119904e-05,
+      "loss": 18.1093,
+      "step": 8842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0821782716505576e-05,
+      "loss": 18.7475,
+      "step": 8843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.081244510322274e-05,
+      "loss": 0.7312,
+      "step": 8844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080310746159718e-05,
+      "loss": 0.7177,
+      "step": 8845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.079376979195461e-05,
+      "loss": 18.0206,
+      "step": 8846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.078443209462084e-05,
+      "loss": 19.171,
+      "step": 8847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077509436992157e-05,
+      "loss": 0.6835,
+      "step": 8848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.076575661818257e-05,
+      "loss": 18.3799,
+      "step": 8849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07564188397296e-05,
+      "loss": 0.7365,
+      "step": 8850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07470810348884e-05,
+      "loss": 18.3729,
+      "step": 8851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0737743203984756e-05,
+      "loss": 18.0026,
+      "step": 8852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072840534734439e-05,
+      "loss": 18.8941,
+      "step": 8853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.071906746529307e-05,
+      "loss": 19.8279,
+      "step": 8854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070972955815657e-05,
+      "loss": 0.7428,
+      "step": 8855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0700391626260635e-05,
+      "loss": 0.7014,
+      "step": 8856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0691053669931034e-05,
+      "loss": 0.6474,
+      "step": 8857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068171568949353e-05,
+      "loss": 0.7387,
+      "step": 8858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.067237768527386e-05,
+      "loss": 18.6023,
+      "step": 8859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.066303965759782e-05,
+      "loss": 0.6823,
+      "step": 8860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.065370160679115e-05,
+      "loss": 0.7543,
+      "step": 8861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.064436353317964e-05,
+      "loss": 20.0151,
+      "step": 8862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063502543708903e-05,
+      "loss": 0.7153,
+      "step": 8863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0625687318845106e-05,
+      "loss": 16.8434,
+      "step": 8864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.061634917877363e-05,
+      "loss": 0.6461,
+      "step": 8865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.060701101720035e-05,
+      "loss": 0.7176,
+      "step": 8866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059767283445106e-05,
+      "loss": 19.6106,
+      "step": 8867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.058833463085151e-05,
+      "loss": 19.1053,
+      "step": 8868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057899640672749e-05,
+      "loss": 0.6801,
+      "step": 8869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0569658162404754e-05,
+      "loss": 17.497,
+      "step": 8870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0560319898209075e-05,
+      "loss": 0.6588,
+      "step": 8871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.055098161446624e-05,
+      "loss": 20.1333,
+      "step": 8872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0541643311502e-05,
+      "loss": 0.7211,
+      "step": 8873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0532304989642153e-05,
+      "loss": 18.4194,
+      "step": 8874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052296664921245e-05,
+      "loss": 0.6556,
+      "step": 8875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0513628290538684e-05,
+      "loss": 0.6296,
+      "step": 8876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0504289913946626e-05,
+      "loss": 0.6546,
+      "step": 8877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0494951519762025e-05,
+      "loss": 0.7481,
+      "step": 8878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0485613108310705e-05,
+      "loss": 18.7264,
+      "step": 8879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0476274679918414e-05,
+      "loss": 0.6202,
+      "step": 8880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.046693623491092e-05,
+      "loss": 0.6695,
+      "step": 8881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0457597773614044e-05,
+      "loss": 18.4193,
+      "step": 8882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0448259296353515e-05,
+      "loss": 18.0961,
+      "step": 8883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.043892080345514e-05,
+      "loss": 0.7108,
+      "step": 8884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0429582295244704e-05,
+      "loss": 0.7062,
+      "step": 8885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.042024377204796e-05,
+      "loss": 0.6487,
+      "step": 8886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.041090523419072e-05,
+      "loss": 16.9796,
+      "step": 8887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.040156668199876e-05,
+      "loss": 18.2674,
+      "step": 8888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.039222811579783e-05,
+      "loss": 17.9656,
+      "step": 8889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0382889535913766e-05,
+      "loss": 0.6954,
+      "step": 8890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.037355094267231e-05,
+      "loss": 19.3406,
+      "step": 8891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0364212336399265e-05,
+      "loss": 18.3312,
+      "step": 8892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035487371742039e-05,
+      "loss": 17.4566,
+      "step": 8893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.034553508606151e-05,
+      "loss": 18.2972,
+      "step": 8894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0336196442648376e-05,
+      "loss": 18.3535,
+      "step": 8895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032685778750679e-05,
+      "loss": 17.7853,
+      "step": 8896
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.031751912096253e-05,
+      "loss": 0.6926,
+      "step": 8897
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030818044334138e-05,
+      "loss": 17.9636,
+      "step": 8898
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.029884175496914e-05,
+      "loss": 18.8472,
+      "step": 8899
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028950305617158e-05,
+      "loss": 16.9815,
+      "step": 8900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.02801643472745e-05,
+      "loss": 20.2344,
+      "step": 8901
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.027082562860368e-05,
+      "loss": 18.2792,
+      "step": 8902
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026148690048491e-05,
+      "loss": 17.3181,
+      "step": 8903
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0252148163243995e-05,
+      "loss": 18.9455,
+      "step": 8904
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0242809417206706e-05,
+      "loss": 0.6603,
+      "step": 8905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.023347066269881e-05,
+      "loss": 0.6031,
+      "step": 8906
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.022413190004615e-05,
+      "loss": 0.7481,
+      "step": 8907
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.021479312957446e-05,
+      "loss": 0.7291,
+      "step": 8908
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.020545435160957e-05,
+      "loss": 17.9673,
+      "step": 8909
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0196115566477264e-05,
+      "loss": 19.3272,
+      "step": 8910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0186776774503305e-05,
+      "loss": 20.1653,
+      "step": 8911
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.017743797601352e-05,
+      "loss": 18.6308,
+      "step": 8912
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.016809917133366e-05,
+      "loss": 18.878,
+      "step": 8913
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.015876036078956e-05,
+      "loss": 17.0254,
+      "step": 8914
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.014942154470699e-05,
+      "loss": 17.9298,
+      "step": 8915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.014008272341173e-05,
+      "loss": 0.6956,
+      "step": 8916
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.013074389722958e-05,
+      "loss": 0.6821,
+      "step": 8917
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012140506648635e-05,
+      "loss": 17.2451,
+      "step": 8918
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.011206623150782e-05,
+      "loss": 17.8907,
+      "step": 8919
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.010272739261975e-05,
+      "loss": 0.6929,
+      "step": 8920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.009338855014798e-05,
+      "loss": 17.2683,
+      "step": 8921
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008404970441828e-05,
+      "loss": 18.216,
+      "step": 8922
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0074710855756435e-05,
+      "loss": 18.0043,
+      "step": 8923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0065372004488265e-05,
+      "loss": 0.6688,
+      "step": 8924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.005603315093953e-05,
+      "loss": 20.3138,
+      "step": 8925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0046694295436046e-05,
+      "loss": 0.7095,
+      "step": 8926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.00373554383036e-05,
+      "loss": 0.6891,
+      "step": 8927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0028016579867975e-05,
+      "loss": 18.6091,
+      "step": 8928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.001867772045498e-05,
+      "loss": 18.0988,
+      "step": 8929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.000933886039038e-05,
+      "loss": 18.6704,
+      "step": 8930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5e-05,
+      "loss": 19.2486,
+      "step": 8931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9990661139609626e-05,
+      "loss": 0.637,
+      "step": 8932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.998132227954503e-05,
+      "loss": 19.5263,
+      "step": 8933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.997198342013204e-05,
+      "loss": 0.7704,
+      "step": 8934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.996264456169641e-05,
+      "loss": 18.5716,
+      "step": 8935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9953305704563966e-05,
+      "loss": 18.5532,
+      "step": 8936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.994396684906048e-05,
+      "loss": 18.8907,
+      "step": 8937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9934627995511746e-05,
+      "loss": 0.6937,
+      "step": 8938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.992528914424357e-05,
+      "loss": 0.7297,
+      "step": 8939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9915950295581726e-05,
+      "loss": 0.6878,
+      "step": 8940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.990661144985203e-05,
+      "loss": 0.6436,
+      "step": 8941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.989727260738026e-05,
+      "loss": 17.6343,
+      "step": 8942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9887933768492194e-05,
+      "loss": 18.5163,
+      "step": 8943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.987859493351365e-05,
+      "loss": 19.3501,
+      "step": 8944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.986925610277041e-05,
+      "loss": 0.6814,
+      "step": 8945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9859917276588275e-05,
+      "loss": 0.6936,
+      "step": 8946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.985057845529302e-05,
+      "loss": 0.7097,
+      "step": 8947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.984123963921043e-05,
+      "loss": 17.3394,
+      "step": 8948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983190082866633e-05,
+      "loss": 0.6918,
+      "step": 8949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9822562023986485e-05,
+      "loss": 0.7476,
+      "step": 8950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98132232254967e-05,
+      "loss": 17.4417,
+      "step": 8951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9803884433522754e-05,
+      "loss": 16.6548,
+      "step": 8952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.979454564839043e-05,
+      "loss": 0.7209,
+      "step": 8953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978520687042554e-05,
+      "loss": 16.5735,
+      "step": 8954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.977586809995386e-05,
+      "loss": 20.574,
+      "step": 8955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.976652933730119e-05,
+      "loss": 19.5135,
+      "step": 8956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975719058279331e-05,
+      "loss": 0.6943,
+      "step": 8957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.974785183675601e-05,
+      "loss": 19.3253,
+      "step": 8958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9738513099515096e-05,
+      "loss": 18.3716,
+      "step": 8959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9729174371396334e-05,
+      "loss": 0.6728,
+      "step": 8960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971983565272553e-05,
+      "loss": 0.7229,
+      "step": 8961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971049694382844e-05,
+      "loss": 20.8363,
+      "step": 8962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9701158245030887e-05,
+      "loss": 18.2485,
+      "step": 8963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969181955665864e-05,
+      "loss": 0.6822,
+      "step": 8964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9682480879037494e-05,
+      "loss": 0.7123,
+      "step": 8965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.967314221249324e-05,
+      "loss": 0.6435,
+      "step": 8966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.966380355735164e-05,
+      "loss": 0.6881,
+      "step": 8967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965446491393851e-05,
+      "loss": 0.6841,
+      "step": 8968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.964512628257961e-05,
+      "loss": 0.7041,
+      "step": 8969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.963578766360076e-05,
+      "loss": 17.4088,
+      "step": 8970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.962644905732771e-05,
+      "loss": 18.5391,
+      "step": 8971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9617110464086245e-05,
+      "loss": 0.7408,
+      "step": 8972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.960777188420217e-05,
+      "loss": 0.6878,
+      "step": 8973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9598433318001254e-05,
+      "loss": 18.8709,
+      "step": 8974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9589094765809284e-05,
+      "loss": 20.511,
+      "step": 8975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.957975622795205e-05,
+      "loss": 17.7107,
+      "step": 8976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.957041770475531e-05,
+      "loss": 19.7396,
+      "step": 8977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9561079196544866e-05,
+      "loss": 16.8515,
+      "step": 8978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9551740703646496e-05,
+      "loss": 0.699,
+      "step": 8979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9542402226385975e-05,
+      "loss": 0.7158,
+      "step": 8980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.953306376508909e-05,
+      "loss": 16.9646,
+      "step": 8981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.95237253200816e-05,
+      "loss": 17.4122,
+      "step": 8982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951438689168931e-05,
+      "loss": 0.6252,
+      "step": 8983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.950504848023799e-05,
+      "loss": 19.4522,
+      "step": 8984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949571008605339e-05,
+      "loss": 17.7802,
+      "step": 8985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9486371709461334e-05,
+      "loss": 18.4841,
+      "step": 8986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.947703335078755e-05,
+      "loss": 0.6184,
+      "step": 8987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9467695010357865e-05,
+      "loss": 20.8444,
+      "step": 8988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.945835668849801e-05,
+      "loss": 19.8622,
+      "step": 8989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.944901838553377e-05,
+      "loss": 0.7747,
+      "step": 8990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9439680101790937e-05,
+      "loss": 0.7154,
+      "step": 8991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.943034183759526e-05,
+      "loss": 18.447,
+      "step": 8992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942100359327252e-05,
+      "loss": 18.4728,
+      "step": 8993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94116653691485e-05,
+      "loss": 0.72,
+      "step": 8994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.940232716554895e-05,
+      "loss": 0.6872,
+      "step": 8995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9392988982799664e-05,
+      "loss": 0.6201,
+      "step": 8996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938365082122638e-05,
+      "loss": 18.4667,
+      "step": 8997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93743126811549e-05,
+      "loss": 17.889,
+      "step": 8998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936497456291097e-05,
+      "loss": 0.637,
+      "step": 8999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.935563646682036e-05,
+      "loss": 18.8667,
+      "step": 9000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.934629839320885e-05,
+      "loss": 17.1176,
+      "step": 9001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9336960342402175e-05,
+      "loss": 18.458,
+      "step": 9002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.932762231472614e-05,
+      "loss": 0.7267,
+      "step": 9003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9318284310506484e-05,
+      "loss": 17.8266,
+      "step": 9004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9308946330068964e-05,
+      "loss": 18.0584,
+      "step": 9005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929960837373936e-05,
+      "loss": 17.9182,
+      "step": 9006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929027044184343e-05,
+      "loss": 0.6665,
+      "step": 9007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.928093253470693e-05,
+      "loss": 0.709,
+      "step": 9008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927159465265562e-05,
+      "loss": 18.2774,
+      "step": 9009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.926225679601525e-05,
+      "loss": 19.5686,
+      "step": 9010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9252918965111597e-05,
+      "loss": 17.8164,
+      "step": 9011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9243581160270396e-05,
+      "loss": 19.0997,
+      "step": 9012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.923424338181745e-05,
+      "loss": 18.672,
+      "step": 9013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9224905630078445e-05,
+      "loss": 18.7309,
+      "step": 9014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9215567905379185e-05,
+      "loss": 18.062,
+      "step": 9015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.920623020804539e-05,
+      "loss": 19.2048,
+      "step": 9016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.919689253840285e-05,
+      "loss": 0.7505,
+      "step": 9017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9187554896777285e-05,
+      "loss": 18.9412,
+      "step": 9018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9178217283494456e-05,
+      "loss": 18.637,
+      "step": 9019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9168879698880114e-05,
+      "loss": 17.7802,
+      "step": 9020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.915954214326001e-05,
+      "loss": 0.7427,
+      "step": 9021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9150204616959886e-05,
+      "loss": 0.7112,
+      "step": 9022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91408671203055e-05,
+      "loss": 0.6667,
+      "step": 9023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.913152965362257e-05,
+      "loss": 18.3107,
+      "step": 9024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.912219221723687e-05,
+      "loss": 0.6374,
+      "step": 9025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911285481147413e-05,
+      "loss": 18.7484,
+      "step": 9026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9103517436660096e-05,
+      "loss": 17.9389,
+      "step": 9027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9094180093120505e-05,
+      "loss": 16.5541,
+      "step": 9028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908484278118109e-05,
+      "loss": 18.8979,
+      "step": 9029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.907550550116761e-05,
+      "loss": 18.1858,
+      "step": 9030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906616825340578e-05,
+      "loss": 18.1243,
+      "step": 9031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9056831038221365e-05,
+      "loss": 19.14,
+      "step": 9032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9047493855940076e-05,
+      "loss": 0.6764,
+      "step": 9033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9038156706887654e-05,
+      "loss": 0.7431,
+      "step": 9034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.902881959138984e-05,
+      "loss": 0.6983,
+      "step": 9035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9019482509772344e-05,
+      "loss": 18.7091,
+      "step": 9036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.901014546236092e-05,
+      "loss": 18.1125,
+      "step": 9037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9000808449481295e-05,
+      "loss": 0.7049,
+      "step": 9038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.899147147145918e-05,
+      "loss": 0.7223,
+      "step": 9039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8982134528620326e-05,
+      "loss": 19.4708,
+      "step": 9040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897279762129044e-05,
+      "loss": 0.657,
+      "step": 9041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8963460749795256e-05,
+      "loss": 18.3949,
+      "step": 9042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8954123914460506e-05,
+      "loss": 17.3336,
+      "step": 9043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.894478711561189e-05,
+      "loss": 20.2948,
+      "step": 9044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.893545035357514e-05,
+      "loss": 18.7357,
+      "step": 9045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8926113628675986e-05,
+      "loss": 18.6131,
+      "step": 9046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8916776941240135e-05,
+      "loss": 0.7297,
+      "step": 9047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.890744029159331e-05,
+      "loss": 18.8427,
+      "step": 9048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.889810368006121e-05,
+      "loss": 19.0087,
+      "step": 9049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.888876710696958e-05,
+      "loss": 0.7032,
+      "step": 9050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88794305726441e-05,
+      "loss": 0.674,
+      "step": 9051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8870094077410506e-05,
+      "loss": 0.6935,
+      "step": 9052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8860757621594505e-05,
+      "loss": 19.323,
+      "step": 9053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8851421205521786e-05,
+      "loss": 18.5227,
+      "step": 9054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.884208482951809e-05,
+      "loss": 17.769,
+      "step": 9055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.883274849390909e-05,
+      "loss": 18.2737,
+      "step": 9056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8823412199020505e-05,
+      "loss": 0.6627,
+      "step": 9057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8814075945178054e-05,
+      "loss": 17.7139,
+      "step": 9058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880473973270741e-05,
+      "loss": 0.6727,
+      "step": 9059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.87954035619343e-05,
+      "loss": 0.7152,
+      "step": 9060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.878606743318439e-05,
+      "loss": 18.5378,
+      "step": 9061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.877673134678342e-05,
+      "loss": 18.204,
+      "step": 9062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.876739530305706e-05,
+      "loss": 0.6771,
+      "step": 9063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.875805930233101e-05,
+      "loss": 17.6399,
+      "step": 9064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.874872334493096e-05,
+      "loss": 17.2953,
+      "step": 9065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873938743118259e-05,
+      "loss": 17.359,
+      "step": 9066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873005156141163e-05,
+      "loss": 20.4382,
+      "step": 9067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8720715735943746e-05,
+      "loss": 20.3242,
+      "step": 9068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8711379955104605e-05,
+      "loss": 18.5824,
+      "step": 9069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.870204421921991e-05,
+      "loss": 0.6879,
+      "step": 9070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.869270852861534e-05,
+      "loss": 19.5042,
+      "step": 9071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.86833728836166e-05,
+      "loss": 0.6896,
+      "step": 9072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.867403728454933e-05,
+      "loss": 17.5545,
+      "step": 9073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8664701731739245e-05,
+      "loss": 0.7307,
+      "step": 9074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8655366225511994e-05,
+      "loss": 17.9322,
+      "step": 9075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8646030766193285e-05,
+      "loss": 18.4669,
+      "step": 9076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.863669535410877e-05,
+      "loss": 0.7045,
+      "step": 9077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862735998958412e-05,
+      "loss": 0.7649,
+      "step": 9078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.861802467294503e-05,
+      "loss": 18.5398,
+      "step": 9079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8608689404517136e-05,
+      "loss": 19.3604,
+      "step": 9080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859935418462612e-05,
+      "loss": 17.7387,
+      "step": 9081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859001901359766e-05,
+      "loss": 0.6629,
+      "step": 9082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8580683891757406e-05,
+      "loss": 0.6988,
+      "step": 9083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.857134881943103e-05,
+      "loss": 18.6903,
+      "step": 9084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8562013796944185e-05,
+      "loss": 18.389,
+      "step": 9085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855267882462252e-05,
+      "loss": 16.7409,
+      "step": 9086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8543343902791714e-05,
+      "loss": 0.7046,
+      "step": 9087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.853400903177741e-05,
+      "loss": 0.7366,
+      "step": 9088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.852467421190527e-05,
+      "loss": 19.4098,
+      "step": 9089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851533944350094e-05,
+      "loss": 19.9723,
+      "step": 9090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.850600472689006e-05,
+      "loss": 18.2999,
+      "step": 9091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8496670062398306e-05,
+      "loss": 19.8343,
+      "step": 9092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.848733545035129e-05,
+      "loss": 18.9895,
+      "step": 9093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.847800089107469e-05,
+      "loss": 16.1624,
+      "step": 9094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846866638489413e-05,
+      "loss": 19.5716,
+      "step": 9095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.845933193213525e-05,
+      "loss": 0.6732,
+      "step": 9096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84499975331237e-05,
+      "loss": 0.653,
+      "step": 9097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84406631881851e-05,
+      "loss": 19.3521,
+      "step": 9098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84313288976451e-05,
+      "loss": 17.6575,
+      "step": 9099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.842199466182934e-05,
+      "loss": 0.698,
+      "step": 9100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8412660481063427e-05,
+      "loss": 0.7357,
+      "step": 9101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.840332635567301e-05,
+      "loss": 18.3409,
+      "step": 9102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.839399228598371e-05,
+      "loss": 0.7259,
+      "step": 9103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8384658272321156e-05,
+      "loss": 18.9759,
+      "step": 9104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.837532431501098e-05,
+      "loss": 0.6735,
+      "step": 9105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.836599041437878e-05,
+      "loss": 0.7063,
+      "step": 9106
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8356656570750204e-05,
+      "loss": 18.22,
+      "step": 9107
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.834732278445084e-05,
+      "loss": 16.8725,
+      "step": 9108
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.833798905580633e-05,
+      "loss": 0.6604,
+      "step": 9109
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8328655385142283e-05,
+      "loss": 17.4733,
+      "step": 9110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.831932177278429e-05,
+      "loss": 18.7739,
+      "step": 9111
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8309988219058e-05,
+      "loss": 19.3288,
+      "step": 9112
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8300654724288967e-05,
+      "loss": 0.69,
+      "step": 9113
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.829132128880285e-05,
+      "loss": 18.2356,
+      "step": 9114
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8281987912925227e-05,
+      "loss": 17.7522,
+      "step": 9115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.827265459698169e-05,
+      "loss": 17.077,
+      "step": 9116
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8263321341297865e-05,
+      "loss": 18.7012,
+      "step": 9117
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.825398814619932e-05,
+      "loss": 0.6428,
+      "step": 9118
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824465501201168e-05,
+      "loss": 19.2611,
+      "step": 9119
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.823532193906052e-05,
+      "loss": 19.2645,
+      "step": 9120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.822598892767143e-05,
+      "loss": 18.6076,
+      "step": 9121
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.821665597817001e-05,
+      "loss": 18.5048,
+      "step": 9122
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.820732309088185e-05,
+      "loss": 17.8405,
+      "step": 9123
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.819799026613252e-05,
+      "loss": 19.0273,
+      "step": 9124
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8188657504247594e-05,
+      "loss": 18.6811,
+      "step": 9125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8179324805552673e-05,
+      "loss": 17.954,
+      "step": 9126
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.816999217037331e-05,
+      "loss": 0.6668,
+      "step": 9127
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8160659599035126e-05,
+      "loss": 19.278,
+      "step": 9128
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.815132709186365e-05,
+      "loss": 0.6871,
+      "step": 9129
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8141994649184466e-05,
+      "loss": 0.6834,
+      "step": 9130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.813266227132316e-05,
+      "loss": 19.1165,
+      "step": 9131
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.812332995860528e-05,
+      "loss": 16.5799,
+      "step": 9132
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8113997711356395e-05,
+      "loss": 0.7141,
+      "step": 9133
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8104665529902075e-05,
+      "loss": 18.3664,
+      "step": 9134
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.809533341456786e-05,
+      "loss": 0.6993,
+      "step": 9135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808600136567933e-05,
+      "loss": 0.6857,
+      "step": 9136
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8076669383562026e-05,
+      "loss": 18.4938,
+      "step": 9137
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.806733746854152e-05,
+      "loss": 18.0214,
+      "step": 9138
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.805800562094334e-05,
+      "loss": 0.7604,
+      "step": 9139
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8048673841093036e-05,
+      "loss": 0.6934,
+      "step": 9140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8039342129316175e-05,
+      "loss": 17.7284,
+      "step": 9141
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.803001048593827e-05,
+      "loss": 0.7742,
+      "step": 9142
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8020678911284895e-05,
+      "loss": 20.2154,
+      "step": 9143
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8011347405681566e-05,
+      "loss": 18.3281,
+      "step": 9144
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8002015969453826e-05,
+      "loss": 0.6799,
+      "step": 9145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.799268460292721e-05,
+      "loss": 0.7196,
+      "step": 9146
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798335330642724e-05,
+      "loss": 0.6666,
+      "step": 9147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7974022080279464e-05,
+      "loss": 0.6819,
+      "step": 9148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7964690924809394e-05,
+      "loss": 0.7306,
+      "step": 9149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.795535984034255e-05,
+      "loss": 20.0653,
+      "step": 9150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7946028827204476e-05,
+      "loss": 0.7178,
+      "step": 9151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.793669788572066e-05,
+      "loss": 0.6833,
+      "step": 9152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.792736701621665e-05,
+      "loss": 16.9977,
+      "step": 9153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.791803621901795e-05,
+      "loss": 18.4264,
+      "step": 9154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.790870549445005e-05,
+      "loss": 18.5571,
+      "step": 9155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789937484283849e-05,
+      "loss": 0.6787,
+      "step": 9156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789004426450875e-05,
+      "loss": 18.0739,
+      "step": 9157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.788071375978636e-05,
+      "loss": 17.8396,
+      "step": 9158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.78713833289968e-05,
+      "loss": 0.6401,
+      "step": 9159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786205297246558e-05,
+      "loss": 0.6207,
+      "step": 9160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7852722690518196e-05,
+      "loss": 18.5285,
+      "step": 9161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.784339248348013e-05,
+      "loss": 17.7987,
+      "step": 9162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.78340623516769e-05,
+      "loss": 16.924,
+      "step": 9163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.782473229543397e-05,
+      "loss": 19.2384,
+      "step": 9164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.781540231507682e-05,
+      "loss": 0.6685,
+      "step": 9165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.780607241093096e-05,
+      "loss": 19.2694,
+      "step": 9166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7796742583321854e-05,
+      "loss": 17.8052,
+      "step": 9167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.778741283257497e-05,
+      "loss": 15.2302,
+      "step": 9168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77780831590158e-05,
+      "loss": 17.8746,
+      "step": 9169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.776875356296981e-05,
+      "loss": 0.7112,
+      "step": 9170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.775942404476248e-05,
+      "loss": 0.7494,
+      "step": 9171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.775009460471926e-05,
+      "loss": 19.632,
+      "step": 9172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.774076524316562e-05,
+      "loss": 18.3607,
+      "step": 9173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.773143596042704e-05,
+      "loss": 20.7406,
+      "step": 9174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7722106756828944e-05,
+      "loss": 17.7767,
+      "step": 9175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.771277763269682e-05,
+      "loss": 19.4378,
+      "step": 9176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.770344858835611e-05,
+      "loss": 17.7349,
+      "step": 9177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.769411962413226e-05,
+      "loss": 18.2756,
+      "step": 9178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7684790740350713e-05,
+      "loss": 18.0626,
+      "step": 9179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.767546193733693e-05,
+      "loss": 18.5121,
+      "step": 9180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.766613321541634e-05,
+      "loss": 18.5328,
+      "step": 9181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.765680457491438e-05,
+      "loss": 19.636,
+      "step": 9182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.76474760161565e-05,
+      "loss": 0.664,
+      "step": 9183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.763814753946812e-05,
+      "loss": 18.2531,
+      "step": 9184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7628819145174684e-05,
+      "loss": 18.3419,
+      "step": 9185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761949083360161e-05,
+      "loss": 19.347,
+      "step": 9186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761016260507432e-05,
+      "loss": 17.8701,
+      "step": 9187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.760083445991825e-05,
+      "loss": 17.9931,
+      "step": 9188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75915063984588e-05,
+      "loss": 0.7089,
+      "step": 9189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7582178421021406e-05,
+      "loss": 19.6598,
+      "step": 9190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.757285052793148e-05,
+      "loss": 18.2668,
+      "step": 9191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.756352271951441e-05,
+      "loss": 18.7859,
+      "step": 9192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7554194996095624e-05,
+      "loss": 19.2141,
+      "step": 9193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.754486735800051e-05,
+      "loss": 0.6977,
+      "step": 9194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7535539805554495e-05,
+      "loss": 0.6728,
+      "step": 9195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752621233908296e-05,
+      "loss": 20.9277,
+      "step": 9196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75168849589113e-05,
+      "loss": 19.4935,
+      "step": 9197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7507557665364914e-05,
+      "loss": 17.271,
+      "step": 9198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.749823045876917e-05,
+      "loss": 18.6515,
+      "step": 9199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7488903339449484e-05,
+      "loss": 19.613,
+      "step": 9200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.747957630773124e-05,
+      "loss": 19.2391,
+      "step": 9201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.747024936393979e-05,
+      "loss": 17.2258,
+      "step": 9202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746092250840053e-05,
+      "loss": 16.4255,
+      "step": 9203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.745159574143882e-05,
+      "loss": 17.6362,
+      "step": 9204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.744226906338006e-05,
+      "loss": 0.6274,
+      "step": 9205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.74329424745496e-05,
+      "loss": 16.8661,
+      "step": 9206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.742361597527279e-05,
+      "loss": 17.2069,
+      "step": 9207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741428956587501e-05,
+      "loss": 16.9676,
+      "step": 9208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.740496324668162e-05,
+      "loss": 18.49,
+      "step": 9209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.739563701801797e-05,
+      "loss": 0.6704,
+      "step": 9210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7386310880209416e-05,
+      "loss": 20.9883,
+      "step": 9211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.737698483358129e-05,
+      "loss": 0.7078,
+      "step": 9212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.736765887845896e-05,
+      "loss": 0.695,
+      "step": 9213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.735833301516776e-05,
+      "loss": 18.0313,
+      "step": 9214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7349007244033034e-05,
+      "loss": 0.6767,
+      "step": 9215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.733968156538011e-05,
+      "loss": 17.5221,
+      "step": 9216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7330355979534325e-05,
+      "loss": 18.8755,
+      "step": 9217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.732103048682101e-05,
+      "loss": 17.6921,
+      "step": 9218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7311705087565483e-05,
+      "loss": 0.6465,
+      "step": 9219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7302379782093085e-05,
+      "loss": 0.7156,
+      "step": 9220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.729305457072913e-05,
+      "loss": 17.4077,
+      "step": 9221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.728372945379891e-05,
+      "loss": 0.6933,
+      "step": 9222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7274404431627777e-05,
+      "loss": 0.7226,
+      "step": 9223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7265079504541015e-05,
+      "loss": 18.9998,
+      "step": 9224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725575467286394e-05,
+      "loss": 17.2529,
+      "step": 9225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.724642993692187e-05,
+      "loss": 17.701,
+      "step": 9226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.723710529704007e-05,
+      "loss": 17.8086,
+      "step": 9227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.722778075354387e-05,
+      "loss": 17.2112,
+      "step": 9228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7218456306758526e-05,
+      "loss": 0.7254,
+      "step": 9229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.720913195700938e-05,
+      "loss": 0.6686,
+      "step": 9230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7199807704621685e-05,
+      "loss": 0.6644,
+      "step": 9231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.719048354992074e-05,
+      "loss": 0.7194,
+      "step": 9232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7181159493231794e-05,
+      "loss": 17.5616,
+      "step": 9233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.717183553488016e-05,
+      "loss": 18.8091,
+      "step": 9234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.716251167519109e-05,
+      "loss": 0.7411,
+      "step": 9235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.715318791448985e-05,
+      "loss": 18.3202,
+      "step": 9236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.714386425310173e-05,
+      "loss": 18.3705,
+      "step": 9237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.713454069135196e-05,
+      "loss": 0.5841,
+      "step": 9238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712521722956583e-05,
+      "loss": 18.7551,
+      "step": 9239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7115893868068585e-05,
+      "loss": 18.3226,
+      "step": 9240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.710657060718546e-05,
+      "loss": 18.9963,
+      "step": 9241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.709724744724174e-05,
+      "loss": 19.3312,
+      "step": 9242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7087924388562635e-05,
+      "loss": 17.4088,
+      "step": 9243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.70786014314734e-05,
+      "loss": 18.2597,
+      "step": 9244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.706927857629929e-05,
+      "loss": 18.4994,
+      "step": 9245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.70599558233655e-05,
+      "loss": 17.953,
+      "step": 9246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.705063317299731e-05,
+      "loss": 18.9977,
+      "step": 9247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.704131062551989e-05,
+      "loss": 16.6333,
+      "step": 9248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7031988181258525e-05,
+      "loss": 19.4573,
+      "step": 9249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.70226658405384e-05,
+      "loss": 0.6915,
+      "step": 9250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.701334360368473e-05,
+      "loss": 0.6932,
+      "step": 9251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700402147102274e-05,
+      "loss": 0.6824,
+      "step": 9252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6994699442877624e-05,
+      "loss": 0.6635,
+      "step": 9253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.698537751957462e-05,
+      "loss": 19.0513,
+      "step": 9254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6976055701438896e-05,
+      "loss": 0.7143,
+      "step": 9255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6966733988795663e-05,
+      "loss": 18.9039,
+      "step": 9256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.695741238197012e-05,
+      "loss": 0.6481,
+      "step": 9257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6948090881287445e-05,
+      "loss": 0.6576,
+      "step": 9258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6938769487072846e-05,
+      "loss": 0.6914,
+      "step": 9259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6929448199651496e-05,
+      "loss": 18.8406,
+      "step": 9260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6920127019348556e-05,
+      "loss": 19.2881,
+      "step": 9261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.691080594648924e-05,
+      "loss": 19.4655,
+      "step": 9262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6901484981398686e-05,
+      "loss": 0.7067,
+      "step": 9263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689216412440209e-05,
+      "loss": 16.4908,
+      "step": 9264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.68828433758246e-05,
+      "loss": 0.7144,
+      "step": 9265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.687352273599137e-05,
+      "loss": 18.8747,
+      "step": 9266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.686420220522758e-05,
+      "loss": 0.7577,
+      "step": 9267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685488178385837e-05,
+      "loss": 18.7755,
+      "step": 9268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6845561472208885e-05,
+      "loss": 17.2978,
+      "step": 9269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.683624127060428e-05,
+      "loss": 17.8989,
+      "step": 9270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682692117936969e-05,
+      "loss": 18.482,
+      "step": 9271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6817601198830264e-05,
+      "loss": 16.7296,
+      "step": 9272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6808281329311135e-05,
+      "loss": 19.1837,
+      "step": 9273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.679896157113741e-05,
+      "loss": 19.3233,
+      "step": 9274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6789641924634246e-05,
+      "loss": 18.6387,
+      "step": 9275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.678032239012674e-05,
+      "loss": 16.9163,
+      "step": 9276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6771002967940045e-05,
+      "loss": 0.6904,
+      "step": 9277
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.676168365839925e-05,
+      "loss": 17.5201,
+      "step": 9278
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.675236446182946e-05,
+      "loss": 0.6818,
+      "step": 9279
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.67430453785558e-05,
+      "loss": 0.7246,
+      "step": 9280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6733726408903353e-05,
+      "loss": 17.8256,
+      "step": 9281
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.672440755319725e-05,
+      "loss": 0.6988,
+      "step": 9282
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.671508881176256e-05,
+      "loss": 0.7084,
+      "step": 9283
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.670577018492437e-05,
+      "loss": 0.7053,
+      "step": 9284
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.669645167300779e-05,
+      "loss": 19.5942,
+      "step": 9285
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.66871332763379e-05,
+      "loss": 20.5074,
+      "step": 9286
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.667781499523975e-05,
+      "loss": 18.299,
+      "step": 9287
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6668496830038435e-05,
+      "loss": 18.2742,
+      "step": 9288
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.665917878105904e-05,
+      "loss": 19.7504,
+      "step": 9289
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6649860848626595e-05,
+      "loss": 0.6908,
+      "step": 9290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.664054303306621e-05,
+      "loss": 17.4325,
+      "step": 9291
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663122533470291e-05,
+      "loss": 18.4316,
+      "step": 9292
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.662190775386175e-05,
+      "loss": 16.8215,
+      "step": 9293
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6612590290867795e-05,
+      "loss": 19.8694,
+      "step": 9294
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.660327294604608e-05,
+      "loss": 0.71,
+      "step": 9295
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6593955719721664e-05,
+      "loss": 17.7521,
+      "step": 9296
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6584638612219577e-05,
+      "loss": 18.0201,
+      "step": 9297
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6575321623864836e-05,
+      "loss": 18.1166,
+      "step": 9298
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6566004754982494e-05,
+      "loss": 16.814,
+      "step": 9299
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.655668800589757e-05,
+      "loss": 17.9773,
+      "step": 9300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.654737137693508e-05,
+      "loss": 16.5858,
+      "step": 9301
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.653805486842004e-05,
+      "loss": 17.7974,
+      "step": 9302
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652873848067747e-05,
+      "loss": 18.2696,
+      "step": 9303
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6519422214032385e-05,
+      "loss": 19.1835,
+      "step": 9304
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6510106068809775e-05,
+      "loss": 16.7224,
+      "step": 9305
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.650079004533465e-05,
+      "loss": 18.8884,
+      "step": 9306
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.649147414393201e-05,
+      "loss": 17.3004,
+      "step": 9307
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.648215836492682e-05,
+      "loss": 18.3868,
+      "step": 9308
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.647284270864411e-05,
+      "loss": 0.6323,
+      "step": 9309
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.646352717540883e-05,
+      "loss": 18.3467,
+      "step": 9310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6454211765545975e-05,
+      "loss": 17.0087,
+      "step": 9311
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6444896479380515e-05,
+      "loss": 0.6916,
+      "step": 9312
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6435581317237414e-05,
+      "loss": 18.2985,
+      "step": 9313
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.642626627944166e-05,
+      "loss": 19.2191,
+      "step": 9314
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641695136631819e-05,
+      "loss": 0.6768,
+      "step": 9315
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.640763657819197e-05,
+      "loss": 0.6786,
+      "step": 9316
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.639832191538797e-05,
+      "loss": 0.6401,
+      "step": 9317
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6389007378231095e-05,
+      "loss": 18.8661,
+      "step": 9318
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637969296704634e-05,
+      "loss": 0.7511,
+      "step": 9319
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637037868215861e-05,
+      "loss": 17.3532,
+      "step": 9320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6361064523892874e-05,
+      "loss": 16.9349,
+      "step": 9321
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.635175049257403e-05,
+      "loss": 17.0892,
+      "step": 9322
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6342436588527013e-05,
+      "loss": 18.3669,
+      "step": 9323
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.633312281207676e-05,
+      "loss": 17.7898,
+      "step": 9324
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.632380916354817e-05,
+      "loss": 0.7277,
+      "step": 9325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6314495643266166e-05,
+      "loss": 18.9153,
+      "step": 9326
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630518225155566e-05,
+      "loss": 18.9007,
+      "step": 9327
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.629586898874154e-05,
+      "loss": 0.6863,
+      "step": 9328
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.628655585514874e-05,
+      "loss": 17.271,
+      "step": 9329
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6277242851102113e-05,
+      "loss": 17.6823,
+      "step": 9330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6267929976926585e-05,
+      "loss": 18.784,
+      "step": 9331
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625861723294703e-05,
+      "loss": 18.3939,
+      "step": 9332
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.624930461948831e-05,
+      "loss": 17.6645,
+      "step": 9333
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6239992136875336e-05,
+      "loss": 19.3035,
+      "step": 9334
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.623067978543295e-05,
+      "loss": 19.0657,
+      "step": 9335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6221367565486054e-05,
+      "loss": 0.6719,
+      "step": 9336
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6212055477359486e-05,
+      "loss": 0.702,
+      "step": 9337
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.62027435213781e-05,
+      "loss": 18.9738,
+      "step": 9338
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6193431697866775e-05,
+      "loss": 18.5375,
+      "step": 9339
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.618412000715033e-05,
+      "loss": 0.6965,
+      "step": 9340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.617480844955367e-05,
+      "loss": 16.0434,
+      "step": 9341
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.616549702540156e-05,
+      "loss": 17.856,
+      "step": 9342
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6156185735018874e-05,
+      "loss": 0.6204,
+      "step": 9343
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.614687457873043e-05,
+      "loss": 0.7065,
+      "step": 9344
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.613756355686107e-05,
+      "loss": 18.9215,
+      "step": 9345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.612825266973561e-05,
+      "loss": 0.6596,
+      "step": 9346
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.611894191767886e-05,
+      "loss": 17.893,
+      "step": 9347
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.610963130101564e-05,
+      "loss": 18.5647,
+      "step": 9348
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.610032082007075e-05,
+      "loss": 0.7074,
+      "step": 9349
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609101047516901e-05,
+      "loss": 17.1787,
+      "step": 9350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6081700266635194e-05,
+      "loss": 18.148,
+      "step": 9351
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6072390194794114e-05,
+      "loss": 18.3874,
+      "step": 9352
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.606308025997055e-05,
+      "loss": 0.6857,
+      "step": 9353
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.605377046248928e-05,
+      "loss": 17.6481,
+      "step": 9354
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6044460802675106e-05,
+      "loss": 0.6471,
+      "step": 9355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.603515128085278e-05,
+      "loss": 0.7108,
+      "step": 9356
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.602584189734708e-05,
+      "loss": 18.2092,
+      "step": 9357
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.601653265248277e-05,
+      "loss": 0.6987,
+      "step": 9358
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60072235465846e-05,
+      "loss": 18.47,
+      "step": 9359
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5997914579977344e-05,
+      "loss": 17.5881,
+      "step": 9360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598860575298575e-05,
+      "loss": 18.2222,
+      "step": 9361
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5979297065934536e-05,
+      "loss": 17.9958,
+      "step": 9362
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5969988519148475e-05,
+      "loss": 19.3789,
+      "step": 9363
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5960680112952285e-05,
+      "loss": 17.3135,
+      "step": 9364
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.595137184767069e-05,
+      "loss": 18.3629,
+      "step": 9365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.594206372362845e-05,
+      "loss": 17.7338,
+      "step": 9366
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5932755741150236e-05,
+      "loss": 0.6939,
+      "step": 9367
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5923447900560805e-05,
+      "loss": 19.553,
+      "step": 9368
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.591414020218485e-05,
+      "loss": 19.3488,
+      "step": 9369
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.590483264634707e-05,
+      "loss": 0.7099,
+      "step": 9370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.589552523337219e-05,
+      "loss": 18.9197,
+      "step": 9371
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.588621796358488e-05,
+      "loss": 19.8296,
+      "step": 9372
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5876910837309844e-05,
+      "loss": 17.8793,
+      "step": 9373
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.586760385487177e-05,
+      "loss": 0.6985,
+      "step": 9374
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5858297016595326e-05,
+      "loss": 0.6273,
+      "step": 9375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5848990322805205e-05,
+      "loss": 0.6773,
+      "step": 9376
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.583968377382605e-05,
+      "loss": 0.7367,
+      "step": 9377
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5830377369982566e-05,
+      "loss": 16.8693,
+      "step": 9378
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582107111159939e-05,
+      "loss": 18.269,
+      "step": 9379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.581176499900117e-05,
+      "loss": 0.682,
+      "step": 9380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.580245903251258e-05,
+      "loss": 17.4099,
+      "step": 9381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.579315321245824e-05,
+      "loss": 20.9054,
+      "step": 9382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578384753916282e-05,
+      "loss": 17.5574,
+      "step": 9383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5774542012950926e-05,
+      "loss": 0.6307,
+      "step": 9384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.576523663414719e-05,
+      "loss": 18.1469,
+      "step": 9385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5755931403076266e-05,
+      "loss": 0.6553,
+      "step": 9386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.574662632006275e-05,
+      "loss": 20.5229,
+      "step": 9387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.573732138543126e-05,
+      "loss": 17.8816,
+      "step": 9388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.572801659950641e-05,
+      "loss": 17.6364,
+      "step": 9389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5718711962612785e-05,
+      "loss": 0.704,
+      "step": 9390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.570940747507502e-05,
+      "loss": 18.9264,
+      "step": 9391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.570010313721768e-05,
+      "loss": 0.7238,
+      "step": 9392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.569079894936536e-05,
+      "loss": 19.1131,
+      "step": 9393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.568149491184265e-05,
+      "loss": 18.8606,
+      "step": 9394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.567219102497412e-05,
+      "loss": 0.6959,
+      "step": 9395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5662887289084366e-05,
+      "loss": 19.7177,
+      "step": 9396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565358370449792e-05,
+      "loss": 0.7201,
+      "step": 9397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.564428027153936e-05,
+      "loss": 0.6987,
+      "step": 9398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.563497699053324e-05,
+      "loss": 0.634,
+      "step": 9399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562567386180412e-05,
+      "loss": 17.1132,
+      "step": 9400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.561637088567654e-05,
+      "loss": 19.5925,
+      "step": 9401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.560706806247505e-05,
+      "loss": 18.0615,
+      "step": 9402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.559776539252417e-05,
+      "loss": 17.7559,
+      "step": 9403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.558846287614843e-05,
+      "loss": 0.6672,
+      "step": 9404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.557916051367238e-05,
+      "loss": 18.9454,
+      "step": 9405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556985830542051e-05,
+      "loss": 0.6578,
+      "step": 9406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556055625171736e-05,
+      "loss": 16.9032,
+      "step": 9407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.555125435288743e-05,
+      "loss": 0.7312,
+      "step": 9408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.55419526092552e-05,
+      "loss": 18.9457,
+      "step": 9409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5532651021145206e-05,
+      "loss": 0.6958,
+      "step": 9410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5523349588881915e-05,
+      "loss": 0.6759,
+      "step": 9411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551404831278982e-05,
+      "loss": 18.838,
+      "step": 9412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.550474719319342e-05,
+      "loss": 0.7126,
+      "step": 9413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.549544623041716e-05,
+      "loss": 20.5725,
+      "step": 9414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5486145424785533e-05,
+      "loss": 0.6546,
+      "step": 9415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.547684477662299e-05,
+      "loss": 0.6663,
+      "step": 9416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.546754428625402e-05,
+      "loss": 18.7934,
+      "step": 9417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5458243954003046e-05,
+      "loss": 0.7234,
+      "step": 9418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5448943780194515e-05,
+      "loss": 20.3239,
+      "step": 9419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54396437651529e-05,
+      "loss": 0.7226,
+      "step": 9420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5430343909202606e-05,
+      "loss": 19.3433,
+      "step": 9421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.542104421266809e-05,
+      "loss": 20.4998,
+      "step": 9422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.541174467587378e-05,
+      "loss": 18.3352,
+      "step": 9423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.540244529914406e-05,
+      "loss": 18.2605,
+      "step": 9424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5393146082803396e-05,
+      "loss": 19.4346,
+      "step": 9425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.538384702717615e-05,
+      "loss": 0.7262,
+      "step": 9426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5374548132586774e-05,
+      "loss": 16.8492,
+      "step": 9427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536524939935964e-05,
+      "loss": 19.3428,
+      "step": 9428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5355950827819124e-05,
+      "loss": 0.7039,
+      "step": 9429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5346652418289656e-05,
+      "loss": 17.139,
+      "step": 9430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.533735417109558e-05,
+      "loss": 18.845,
+      "step": 9431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.53280560865613e-05,
+      "loss": 0.7287,
+      "step": 9432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.531875816501117e-05,
+      "loss": 16.4425,
+      "step": 9433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530946040676955e-05,
+      "loss": 0.6618,
+      "step": 9434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530016281216082e-05,
+      "loss": 18.6481,
+      "step": 9435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5290865381509306e-05,
+      "loss": 17.7839,
+      "step": 9436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528156811513938e-05,
+      "loss": 0.7209,
+      "step": 9437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.527227101337538e-05,
+      "loss": 19.8486,
+      "step": 9438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.526297407654162e-05,
+      "loss": 0.6979,
+      "step": 9439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525367730496246e-05,
+      "loss": 18.8786,
+      "step": 9440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5244380698962205e-05,
+      "loss": 0.6728,
+      "step": 9441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523508425886519e-05,
+      "loss": 0.6924,
+      "step": 9442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.522578798499571e-05,
+      "loss": 0.697,
+      "step": 9443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5216491877678076e-05,
+      "loss": 0.7612,
+      "step": 9444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.52071959372366e-05,
+      "loss": 18.0791,
+      "step": 9445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519790016399556e-05,
+      "loss": 17.0928,
+      "step": 9446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.518860455827928e-05,
+      "loss": 0.6708,
+      "step": 9447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.517930912041201e-05,
+      "loss": 17.455,
+      "step": 9448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5170013850718034e-05,
+      "loss": 0.7225,
+      "step": 9449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.516071874952165e-05,
+      "loss": 17.5217,
+      "step": 9450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5151423817147096e-05,
+      "loss": 0.733,
+      "step": 9451
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5142129053918636e-05,
+      "loss": 0.6661,
+      "step": 9452
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5132834460160524e-05,
+      "loss": 17.9919,
+      "step": 9453
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512354003619702e-05,
+      "loss": 18.1548,
+      "step": 9454
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.511424578235235e-05,
+      "loss": 0.6509,
+      "step": 9455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.510495169895077e-05,
+      "loss": 18.5333,
+      "step": 9456
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5095657786316506e-05,
+      "loss": 18.7483,
+      "step": 9457
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.508636404477377e-05,
+      "loss": 0.6653,
+      "step": 9458
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50770704746468e-05,
+      "loss": 0.6929,
+      "step": 9459
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5067777076259785e-05,
+      "loss": 18.9874,
+      "step": 9460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.505848384993696e-05,
+      "loss": 17.5987,
+      "step": 9461
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.504919079600251e-05,
+      "loss": 0.6714,
+      "step": 9462
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5039897914780615e-05,
+      "loss": 19.6719,
+      "step": 9463
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50306052065955e-05,
+      "loss": 18.7167,
+      "step": 9464
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.502131267177133e-05,
+      "loss": 0.6455,
+      "step": 9465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.501202031063226e-05,
+      "loss": 16.7981,
+      "step": 9466
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50027281235025e-05,
+      "loss": 0.703,
+      "step": 9467
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4993436110706177e-05,
+      "loss": 0.6642,
+      "step": 9468
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4984144272567486e-05,
+      "loss": 16.9275,
+      "step": 9469
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.497485260941056e-05,
+      "loss": 0.6467,
+      "step": 9470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4965561121559537e-05,
+      "loss": 0.6931,
+      "step": 9471
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4956269809338575e-05,
+      "loss": 0.7011,
+      "step": 9472
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4946978673071796e-05,
+      "loss": 0.626,
+      "step": 9473
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.493768771308334e-05,
+      "loss": 0.7222,
+      "step": 9474
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4928396929697316e-05,
+      "loss": 0.6836,
+      "step": 9475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.491910632323784e-05,
+      "loss": 18.7508,
+      "step": 9476
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.490981589402904e-05,
+      "loss": 0.6657,
+      "step": 9477
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4900525642394995e-05,
+      "loss": 18.2287,
+      "step": 9478
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4891235568659815e-05,
+      "loss": 0.6693,
+      "step": 9479
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4881945673147604e-05,
+      "loss": 0.6319,
+      "step": 9480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.487265595618241e-05,
+      "loss": 17.1154,
+      "step": 9481
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.486336641808835e-05,
+      "loss": 0.6849,
+      "step": 9482
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.485407705918947e-05,
+      "loss": 17.3893,
+      "step": 9483
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4844787879809855e-05,
+      "loss": 17.8284,
+      "step": 9484
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.483549888027355e-05,
+      "loss": 18.0572,
+      "step": 9485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.482621006090461e-05,
+      "loss": 16.9424,
+      "step": 9486
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.48169214220271e-05,
+      "loss": 0.6999,
+      "step": 9487
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.480763296396503e-05,
+      "loss": 0.7281,
+      "step": 9488
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4798344687042465e-05,
+      "loss": 0.6704,
+      "step": 9489
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4789056591583424e-05,
+      "loss": 15.7075,
+      "step": 9490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477976867791191e-05,
+      "loss": 17.7927,
+      "step": 9491
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4770480946351965e-05,
+      "loss": 18.9874,
+      "step": 9492
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4761193397227575e-05,
+      "loss": 0.7052,
+      "step": 9493
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.475190603086277e-05,
+      "loss": 18.5536,
+      "step": 9494
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.474261884758153e-05,
+      "loss": 17.9487,
+      "step": 9495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4733331847707835e-05,
+      "loss": 0.6832,
+      "step": 9496
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.472404503156569e-05,
+      "loss": 19.2777,
+      "step": 9497
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4714758399479053e-05,
+      "loss": 17.7561,
+      "step": 9498
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.470547195177191e-05,
+      "loss": 20.0138,
+      "step": 9499
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.469618568876823e-05,
+      "loss": 19.3519,
+      "step": 9500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4686899610791946e-05,
+      "loss": 18.3371,
+      "step": 9501
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.467761371816703e-05,
+      "loss": 0.6792,
+      "step": 9502
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.466832801121742e-05,
+      "loss": 17.6456,
+      "step": 9503
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.465904249026706e-05,
+      "loss": 0.6892,
+      "step": 9504
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4649757155639895e-05,
+      "loss": 0.6846,
+      "step": 9505
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.464047200765982e-05,
+      "loss": 18.2013,
+      "step": 9506
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.463118704665076e-05,
+      "loss": 0.6682,
+      "step": 9507
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.462190227293664e-05,
+      "loss": 18.1983,
+      "step": 9508
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4612617686841365e-05,
+      "loss": 18.7822,
+      "step": 9509
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.460333328868882e-05,
+      "loss": 18.2348,
+      "step": 9510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4594049078802925e-05,
+      "loss": 17.6118,
+      "step": 9511
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.458476505750754e-05,
+      "loss": 0.6576,
+      "step": 9512
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.457548122512656e-05,
+      "loss": 18.4925,
+      "step": 9513
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.456619758198386e-05,
+      "loss": 0.7051,
+      "step": 9514
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.455691412840328e-05,
+      "loss": 0.6723,
+      "step": 9515
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4547630864708725e-05,
+      "loss": 0.695,
+      "step": 9516
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.453834779122401e-05,
+      "loss": 18.8426,
+      "step": 9517
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4529064908273e-05,
+      "loss": 0.641,
+      "step": 9518
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451978221617953e-05,
+      "loss": 0.6995,
+      "step": 9519
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451049971526743e-05,
+      "loss": 18.0161,
+      "step": 9520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.450121740586053e-05,
+      "loss": 0.6522,
+      "step": 9521
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.449193528828265e-05,
+      "loss": 18.1203,
+      "step": 9522
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.448265336285762e-05,
+      "loss": 0.6944,
+      "step": 9523
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4473371629909225e-05,
+      "loss": 17.7415,
+      "step": 9524
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.446409008976126e-05,
+      "loss": 0.6998,
+      "step": 9525
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.445480874273754e-05,
+      "loss": 16.9499,
+      "step": 9526
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.444552758916183e-05,
+      "loss": 18.3479,
+      "step": 9527
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4436246629357934e-05,
+      "loss": 0.6839,
+      "step": 9528
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4426965863649604e-05,
+      "loss": 19.0262,
+      "step": 9529
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.441768529236061e-05,
+      "loss": 0.7018,
+      "step": 9530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.440840491581473e-05,
+      "loss": 0.6873,
+      "step": 9531
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.439912473433569e-05,
+      "loss": 0.6562,
+      "step": 9532
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.438984474824725e-05,
+      "loss": 0.7402,
+      "step": 9533
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.438056495787315e-05,
+      "loss": 18.8088,
+      "step": 9534
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4371285363537114e-05,
+      "loss": 0.7298,
+      "step": 9535
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.436200596556288e-05,
+      "loss": 0.6929,
+      "step": 9536
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4352726764274146e-05,
+      "loss": 19.1152,
+      "step": 9537
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434344775999465e-05,
+      "loss": 0.624,
+      "step": 9538
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4334168953048085e-05,
+      "loss": 17.1869,
+      "step": 9539
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4324890343758136e-05,
+      "loss": 18.2936,
+      "step": 9540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.431561193244852e-05,
+      "loss": 0.6516,
+      "step": 9541
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.430633371944289e-05,
+      "loss": 18.5978,
+      "step": 9542
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.429705570506495e-05,
+      "loss": 18.0,
+      "step": 9543
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.428777788963836e-05,
+      "loss": 0.6899,
+      "step": 9544
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.427850027348678e-05,
+      "loss": 17.7421,
+      "step": 9545
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.426922285693388e-05,
+      "loss": 19.2584,
+      "step": 9546
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.425994564030328e-05,
+      "loss": 17.5195,
+      "step": 9547
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4250668623918654e-05,
+      "loss": 0.7062,
+      "step": 9548
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.424139180810363e-05,
+      "loss": 17.09,
+      "step": 9549
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.423211519318182e-05,
+      "loss": 18.2626,
+      "step": 9550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.422283877947687e-05,
+      "loss": 0.7246,
+      "step": 9551
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4213562567312374e-05,
+      "loss": 18.0758,
+      "step": 9552
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4204286557011944e-05,
+      "loss": 0.6546,
+      "step": 9553
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4195010748899186e-05,
+      "loss": 19.5536,
+      "step": 9554
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.418573514329769e-05,
+      "loss": 0.6997,
+      "step": 9555
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.417645974053104e-05,
+      "loss": 0.6353,
+      "step": 9556
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.416718454092283e-05,
+      "loss": 0.6897,
+      "step": 9557
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.41579095447966e-05,
+      "loss": 16.571,
+      "step": 9558
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.414863475247595e-05,
+      "loss": 0.6842,
+      "step": 9559
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4139360164284436e-05,
+      "loss": 16.8034,
+      "step": 9560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.413008578054558e-05,
+      "loss": 17.7052,
+      "step": 9561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4120811601582934e-05,
+      "loss": 16.6848,
+      "step": 9562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.411153762772005e-05,
+      "loss": 19.038,
+      "step": 9563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.410226385928044e-05,
+      "loss": 18.6334,
+      "step": 9564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.409299029658764e-05,
+      "loss": 17.5985,
+      "step": 9565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.408371693996517e-05,
+      "loss": 0.6803,
+      "step": 9566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4074443789736505e-05,
+      "loss": 16.9083,
+      "step": 9567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.406517084622519e-05,
+      "loss": 20.0065,
+      "step": 9568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.405589810975468e-05,
+      "loss": 18.2592,
+      "step": 9569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404662558064849e-05,
+      "loss": 17.4954,
+      "step": 9570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.403735325923009e-05,
+      "loss": 0.635,
+      "step": 9571
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4028081145822937e-05,
+      "loss": 18.473,
+      "step": 9572
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.401880924075051e-05,
+      "loss": 0.6784,
+      "step": 9573
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.400953754433626e-05,
+      "loss": 0.6786,
+      "step": 9574
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.400026605690365e-05,
+      "loss": 19.5611,
+      "step": 9575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3990994778776115e-05,
+      "loss": 17.9266,
+      "step": 9576
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398172371027707e-05,
+      "loss": 17.9622,
+      "step": 9577
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.397245285172997e-05,
+      "loss": 16.0609,
+      "step": 9578
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.396318220345822e-05,
+      "loss": 17.387,
+      "step": 9579
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395391176578525e-05,
+      "loss": 0.7289,
+      "step": 9580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.394464153903445e-05,
+      "loss": 0.6796,
+      "step": 9581
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3935371523529225e-05,
+      "loss": 17.167,
+      "step": 9582
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.392610171959297e-05,
+      "loss": 0.7008,
+      "step": 9583
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3916832127549054e-05,
+      "loss": 16.974,
+      "step": 9584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.390756274772087e-05,
+      "loss": 19.9991,
+      "step": 9585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.389829358043178e-05,
+      "loss": 0.6635,
+      "step": 9586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3889024626005146e-05,
+      "loss": 15.8162,
+      "step": 9587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3879755884764324e-05,
+      "loss": 18.6076,
+      "step": 9588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3870487357032645e-05,
+      "loss": 0.7411,
+      "step": 9589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3861219043133484e-05,
+      "loss": 19.0457,
+      "step": 9590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3851950943390144e-05,
+      "loss": 18.0725,
+      "step": 9591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.384268305812595e-05,
+      "loss": 19.2991,
+      "step": 9592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.383341538766423e-05,
+      "loss": 0.6999,
+      "step": 9593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3824147932328285e-05,
+      "loss": 0.7032,
+      "step": 9594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.381488069244143e-05,
+      "loss": 18.7734,
+      "step": 9595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.380561366832695e-05,
+      "loss": 0.6883,
+      "step": 9596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.379634686030813e-05,
+      "loss": 17.4583,
+      "step": 9597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3787080268708244e-05,
+      "loss": 0.6915,
+      "step": 9598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3777813893850574e-05,
+      "loss": 0.726,
+      "step": 9599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3768547736058385e-05,
+      "loss": 19.1826,
+      "step": 9600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.375928179565494e-05,
+      "loss": 18.2184,
+      "step": 9601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3750016072963455e-05,
+      "loss": 0.6547,
+      "step": 9602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.374075056830721e-05,
+      "loss": 0.694,
+      "step": 9603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.373148528200941e-05,
+      "loss": 19.073,
+      "step": 9604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.372222021439332e-05,
+      "loss": 0.7213,
+      "step": 9605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371295536578212e-05,
+      "loss": 19.146,
+      "step": 9606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.370369073649902e-05,
+      "loss": 0.7056,
+      "step": 9607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369442632686726e-05,
+      "loss": 0.6383,
+      "step": 9608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.368516213721e-05,
+      "loss": 17.9159,
+      "step": 9609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.367589816785045e-05,
+      "loss": 18.1152,
+      "step": 9610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.366663441911178e-05,
+      "loss": 0.6644,
+      "step": 9611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.365737089131715e-05,
+      "loss": 0.6692,
+      "step": 9612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3648107584789763e-05,
+      "loss": 0.6968,
+      "step": 9613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.363884449985275e-05,
+      "loss": 18.6622,
+      "step": 9614
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.362958163682926e-05,
+      "loss": 0.7302,
+      "step": 9615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3620318996042417e-05,
+      "loss": 0.7048,
+      "step": 9616
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3611056577815397e-05,
+      "loss": 20.1055,
+      "step": 9617
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3601794382471287e-05,
+      "loss": 18.5104,
+      "step": 9618
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.359253241033323e-05,
+      "loss": 0.6945,
+      "step": 9619
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358327066172434e-05,
+      "loss": 17.5856,
+      "step": 9620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.357400913696769e-05,
+      "loss": 18.1556,
+      "step": 9621
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356474783638641e-05,
+      "loss": 0.6816,
+      "step": 9622
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.355548676030356e-05,
+      "loss": 18.3784,
+      "step": 9623
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3546225909042246e-05,
+      "loss": 19.8512,
+      "step": 9624
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3536965282925516e-05,
+      "loss": 16.5247,
+      "step": 9625
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3527704882276434e-05,
+      "loss": 0.6548,
+      "step": 9626
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.351844470741808e-05,
+      "loss": 18.7309,
+      "step": 9627
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.350918475867347e-05,
+      "loss": 19.6395,
+      "step": 9628
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.349992503636566e-05,
+      "loss": 0.6293,
+      "step": 9629
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.34906655408177e-05,
+      "loss": 17.5765,
+      "step": 9630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.348140627235258e-05,
+      "loss": 0.6623,
+      "step": 9631
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347214723129334e-05,
+      "loss": 0.6385,
+      "step": 9632
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3462888417962974e-05,
+      "loss": 18.8287,
+      "step": 9633
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.345362983268449e-05,
+      "loss": 18.2832,
+      "step": 9634
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3444371475780894e-05,
+      "loss": 0.6815,
+      "step": 9635
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3435113347575134e-05,
+      "loss": 19.3766,
+      "step": 9636
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.342585544839023e-05,
+      "loss": 19.6756,
+      "step": 9637
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3416597778549106e-05,
+      "loss": 17.7568,
+      "step": 9638
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.340734033837477e-05,
+      "loss": 19.6789,
+      "step": 9639
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.339808312819014e-05,
+      "loss": 18.1556,
+      "step": 9640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3388826148318164e-05,
+      "loss": 0.6609,
+      "step": 9641
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.33795693990818e-05,
+      "loss": 16.4995,
+      "step": 9642
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.337031288080394e-05,
+      "loss": 0.7415,
+      "step": 9643
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.336105659380755e-05,
+      "loss": 0.7388,
+      "step": 9644
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.335180053841551e-05,
+      "loss": 0.7428,
+      "step": 9645
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334254471495072e-05,
+      "loss": 0.7274,
+      "step": 9646
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.33332891237361e-05,
+      "loss": 17.1922,
+      "step": 9647
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3324033765094535e-05,
+      "loss": 18.8981,
+      "step": 9648
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.331477863934888e-05,
+      "loss": 18.7984,
+      "step": 9649
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3305523746822035e-05,
+      "loss": 0.7262,
+      "step": 9650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3296269087836845e-05,
+      "loss": 16.9408,
+      "step": 9651
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.328701466271617e-05,
+      "loss": 17.1107,
+      "step": 9652
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3277760471782876e-05,
+      "loss": 18.1783,
+      "step": 9653
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.326850651535977e-05,
+      "loss": 19.7155,
+      "step": 9654
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.32592527937697e-05,
+      "loss": 0.6666,
+      "step": 9655
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3249999307335495e-05,
+      "loss": 0.7319,
+      "step": 9656
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.324074605637996e-05,
+      "loss": 0.7143,
+      "step": 9657
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.323149304122591e-05,
+      "loss": 17.2782,
+      "step": 9658
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3222240262196125e-05,
+      "loss": 18.7326,
+      "step": 9659
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3212987719613414e-05,
+      "loss": 16.9377,
+      "step": 9660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.320373541380054e-05,
+      "loss": 18.6959,
+      "step": 9661
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.31944833450803e-05,
+      "loss": 0.6695,
+      "step": 9662
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.318523151377545e-05,
+      "loss": 19.14,
+      "step": 9663
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317597992020873e-05,
+      "loss": 19.129,
+      "step": 9664
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.316672856470291e-05,
+      "loss": 0.7314,
+      "step": 9665
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315747744758072e-05,
+      "loss": 18.0449,
+      "step": 9666
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.31482265691649e-05,
+      "loss": 0.6553,
+      "step": 9667
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.313897592977817e-05,
+      "loss": 17.015,
+      "step": 9668
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.312972552974326e-05,
+      "loss": 0.6921,
+      "step": 9669
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.312047536938283e-05,
+      "loss": 18.0627,
+      "step": 9670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.311122544901962e-05,
+      "loss": 19.427,
+      "step": 9671
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.310197576897631e-05,
+      "loss": 16.6045,
+      "step": 9672
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.309272632957558e-05,
+      "loss": 17.0814,
+      "step": 9673
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3083477131140106e-05,
+      "loss": 0.6763,
+      "step": 9674
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.307422817399254e-05,
+      "loss": 0.7383,
+      "step": 9675
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3064979458455564e-05,
+      "loss": 19.0491,
+      "step": 9676
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3055730984851805e-05,
+      "loss": 0.7257,
+      "step": 9677
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.304648275350391e-05,
+      "loss": 19.0674,
+      "step": 9678
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3037234764734516e-05,
+      "loss": 19.9447,
+      "step": 9679
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302798701886622e-05,
+      "loss": 16.9979,
+      "step": 9680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.301873951622168e-05,
+      "loss": 0.6918,
+      "step": 9681
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300949225712347e-05,
+      "loss": 17.816,
+      "step": 9682
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300024524189419e-05,
+      "loss": 17.3303,
+      "step": 9683
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2990998470856444e-05,
+      "loss": 16.6336,
+      "step": 9684
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.298175194433279e-05,
+      "loss": 17.7368,
+      "step": 9685
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2972505662645824e-05,
+      "loss": 21.0898,
+      "step": 9686
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.29632596261181e-05,
+      "loss": 16.677,
+      "step": 9687
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2954013835072156e-05,
+      "loss": 0.6609,
+      "step": 9688
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.294476828983056e-05,
+      "loss": 16.1133,
+      "step": 9689
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293552299071584e-05,
+      "loss": 18.468,
+      "step": 9690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.292627793805053e-05,
+      "loss": 0.6726,
+      "step": 9691
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291703313215715e-05,
+      "loss": 0.7196,
+      "step": 9692
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.290778857335819e-05,
+      "loss": 0.6956,
+      "step": 9693
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2898544261976194e-05,
+      "loss": 0.7081,
+      "step": 9694
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.288930019833362e-05,
+      "loss": 18.1767,
+      "step": 9695
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.288005638275298e-05,
+      "loss": 18.2999,
+      "step": 9696
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2870812815556745e-05,
+      "loss": 18.1755,
+      "step": 9697
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286156949706736e-05,
+      "loss": 0.6706,
+      "step": 9698
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.285232642760731e-05,
+      "loss": 18.8411,
+      "step": 9699
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284308360749904e-05,
+      "loss": 17.7142,
+      "step": 9700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2833841037065e-05,
+      "loss": 0.6821,
+      "step": 9701
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2824598716627615e-05,
+      "loss": 19.7905,
+      "step": 9702
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.28153566465093e-05,
+      "loss": 17.7347,
+      "step": 9703
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280611482703248e-05,
+      "loss": 19.5903,
+      "step": 9704
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.279687325851957e-05,
+      "loss": 0.6857,
+      "step": 9705
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2787631941292964e-05,
+      "loss": 17.8684,
+      "step": 9706
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2778390875675064e-05,
+      "loss": 0.6614,
+      "step": 9707
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2769150061988216e-05,
+      "loss": 18.8405,
+      "step": 9708
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2759909500554827e-05,
+      "loss": 17.6315,
+      "step": 9709
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.275066919169724e-05,
+      "loss": 17.9919,
+      "step": 9710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.274142913573784e-05,
+      "loss": 0.6534,
+      "step": 9711
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.273218933299894e-05,
+      "loss": 17.7915,
+      "step": 9712
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.272294978380288e-05,
+      "loss": 19.1652,
+      "step": 9713
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2713710488472006e-05,
+      "loss": 0.6406,
+      "step": 9714
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.270447144732863e-05,
+      "loss": 15.6791,
+      "step": 9715
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2695232660695055e-05,
+      "loss": 0.663,
+      "step": 9716
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.26859941288936e-05,
+      "loss": 17.5772,
+      "step": 9717
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267675585224654e-05,
+      "loss": 18.7649,
+      "step": 9718
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2667517831076165e-05,
+      "loss": 16.9498,
+      "step": 9719
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.265828006570475e-05,
+      "loss": 0.6675,
+      "step": 9720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2649042556454574e-05,
+      "loss": 17.7649,
+      "step": 9721
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.263980530364788e-05,
+      "loss": 18.4129,
+      "step": 9722
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.263056830760693e-05,
+      "loss": 0.6945,
+      "step": 9723
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2621331568653925e-05,
+      "loss": 19.1147,
+      "step": 9724
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2612095087111136e-05,
+      "loss": 17.971,
+      "step": 9725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2602858863300774e-05,
+      "loss": 16.1891,
+      "step": 9726
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.259362289754504e-05,
+      "loss": 0.6894,
+      "step": 9727
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258438719016615e-05,
+      "loss": 15.791,
+      "step": 9728
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.257515174148629e-05,
+      "loss": 0.6155,
+      "step": 9729
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2565916551827656e-05,
+      "loss": 17.5953,
+      "step": 9730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.255668162151242e-05,
+      "loss": 17.8231,
+      "step": 9731
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.254744695086273e-05,
+      "loss": 18.1726,
+      "step": 9732
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.253821254020078e-05,
+      "loss": 0.6895,
+      "step": 9733
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252897838984869e-05,
+      "loss": 17.0167,
+      "step": 9734
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251974450012862e-05,
+      "loss": 18.0725,
+      "step": 9735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251051087136269e-05,
+      "loss": 0.7509,
+      "step": 9736
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.250127750387301e-05,
+      "loss": 17.9301,
+      "step": 9737
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.249204439798171e-05,
+      "loss": 18.152,
+      "step": 9738
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.248281155401089e-05,
+      "loss": 0.678,
+      "step": 9739
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2473578972282644e-05,
+      "loss": 16.7871,
+      "step": 9740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.246434665311907e-05,
+      "loss": 17.5789,
+      "step": 9741
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2455114596842214e-05,
+      "loss": 19.1012,
+      "step": 9742
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.244588280377417e-05,
+      "loss": 0.7364,
+      "step": 9743
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.243665127423698e-05,
+      "loss": 0.6916,
+      "step": 9744
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.24274200085527e-05,
+      "loss": 17.2293,
+      "step": 9745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.241818900704338e-05,
+      "loss": 17.2672,
+      "step": 9746
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.240895827003102e-05,
+      "loss": 17.7375,
+      "step": 9747
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2399727797837674e-05,
+      "loss": 0.6679,
+      "step": 9748
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239049759078534e-05,
+      "loss": 17.4399,
+      "step": 9749
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2381267649196e-05,
+      "loss": 21.0305,
+      "step": 9750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2372037973391684e-05,
+      "loss": 19.208,
+      "step": 9751
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.236280856369435e-05,
+      "loss": 18.0346,
+      "step": 9752
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2353579420425986e-05,
+      "loss": 18.3682,
+      "step": 9753
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.234435054390856e-05,
+      "loss": 0.6525,
+      "step": 9754
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2335121934464005e-05,
+      "loss": 0.7174,
+      "step": 9755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2325893592414296e-05,
+      "loss": 18.4404,
+      "step": 9756
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2316665518081345e-05,
+      "loss": 0.6537,
+      "step": 9757
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2307437711787104e-05,
+      "loss": 19.633,
+      "step": 9758
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.229821017385348e-05,
+      "loss": 17.5628,
+      "step": 9759
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.228898290460237e-05,
+      "loss": 19.0097,
+      "step": 9760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2279755904355704e-05,
+      "loss": 19.0695,
+      "step": 9761
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227052917343534e-05,
+      "loss": 0.6439,
+      "step": 9762
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2261302712163185e-05,
+      "loss": 0.6551,
+      "step": 9763
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.225207652086111e-05,
+      "loss": 19.756,
+      "step": 9764
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2242850599850944e-05,
+      "loss": 17.7679,
+      "step": 9765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2233624949454586e-05,
+      "loss": 18.1203,
+      "step": 9766
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.222439956999384e-05,
+      "loss": 18.2572,
+      "step": 9767
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.221517446179057e-05,
+      "loss": 19.35,
+      "step": 9768
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.220594962516659e-05,
+      "loss": 18.2728,
+      "step": 9769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21967250604437e-05,
+      "loss": 0.6865,
+      "step": 9770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.218750076794373e-05,
+      "loss": 19.157,
+      "step": 9771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2178276747988446e-05,
+      "loss": 0.695,
+      "step": 9772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.216905300089968e-05,
+      "loss": 14.8148,
+      "step": 9773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2159829526999175e-05,
+      "loss": 18.0552,
+      "step": 9774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2150606326608694e-05,
+      "loss": 0.7283,
+      "step": 9775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.214138340005002e-05,
+      "loss": 0.6258,
+      "step": 9776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.213216074764488e-05,
+      "loss": 0.6798,
+      "step": 9777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.212293836971505e-05,
+      "loss": 18.9581,
+      "step": 9778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21137162665822e-05,
+      "loss": 18.2906,
+      "step": 9779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2104494438568087e-05,
+      "loss": 17.613,
+      "step": 9780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.209527288599441e-05,
+      "loss": 17.6924,
+      "step": 9781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2086051609182887e-05,
+      "loss": 0.598,
+      "step": 9782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207683060845519e-05,
+      "loss": 0.6706,
+      "step": 9783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2067609884133005e-05,
+      "loss": 18.7422,
+      "step": 9784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2058389436538006e-05,
+      "loss": 0.7111,
+      "step": 9785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204916926599185e-05,
+      "loss": 0.7433,
+      "step": 9786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.20399493728162e-05,
+      "loss": 0.7467,
+      "step": 9787
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.20307297573327e-05,
+      "loss": 16.7237,
+      "step": 9788
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202151041986296e-05,
+      "loss": 15.1519,
+      "step": 9789
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.201229136072863e-05,
+      "loss": 16.5868,
+      "step": 9790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2003072580251295e-05,
+      "loss": 16.5792,
+      "step": 9791
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1993854078752596e-05,
+      "loss": 0.6484,
+      "step": 9792
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1984635856554106e-05,
+      "loss": 0.6469,
+      "step": 9793
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1975417913977394e-05,
+      "loss": 0.6853,
+      "step": 9794
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196620025134407e-05,
+      "loss": 18.1342,
+      "step": 9795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1956982868975663e-05,
+      "loss": 17.429,
+      "step": 9796
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.194776576719376e-05,
+      "loss": 17.4524,
+      "step": 9797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.193854894631989e-05,
+      "loss": 18.2528,
+      "step": 9798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1929332406675574e-05,
+      "loss": 18.0551,
+      "step": 9799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.192011614858237e-05,
+      "loss": 17.9609,
+      "step": 9800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1910900172361764e-05,
+      "loss": 0.6726,
+      "step": 9801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.190168447833529e-05,
+      "loss": 18.4743,
+      "step": 9802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.189246906682442e-05,
+      "loss": 16.3715,
+      "step": 9803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.188325393815065e-05,
+      "loss": 0.7263,
+      "step": 9804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1874039092635466e-05,
+      "loss": 18.237,
+      "step": 9805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.186482453060031e-05,
+      "loss": 0.7406,
+      "step": 9806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.185561025236667e-05,
+      "loss": 17.8621,
+      "step": 9807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.184639625825597e-05,
+      "loss": 0.6663,
+      "step": 9808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.183718254858965e-05,
+      "loss": 0.7215,
+      "step": 9809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.182796912368914e-05,
+      "loss": 16.7031,
+      "step": 9810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181875598387586e-05,
+      "loss": 18.2927,
+      "step": 9811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.180954312947121e-05,
+      "loss": 17.4508,
+      "step": 9812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18003305607966e-05,
+      "loss": 0.6901,
+      "step": 9813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1791118278173395e-05,
+      "loss": 17.8248,
+      "step": 9814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1781906281923e-05,
+      "loss": 0.669,
+      "step": 9815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.177269457236676e-05,
+      "loss": 0.6527,
+      "step": 9816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176348314982604e-05,
+      "loss": 0.5838,
+      "step": 9817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17542720146222e-05,
+      "loss": 18.9507,
+      "step": 9818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1745061167076545e-05,
+      "loss": 0.6489,
+      "step": 9819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.173585060751043e-05,
+      "loss": 18.9086,
+      "step": 9820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.172664033624516e-05,
+      "loss": 17.6501,
+      "step": 9821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.171743035360205e-05,
+      "loss": 0.6676,
+      "step": 9822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17082206599024e-05,
+      "loss": 19.7788,
+      "step": 9823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.169901125546747e-05,
+      "loss": 16.5999,
+      "step": 9824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.168980214061857e-05,
+      "loss": 0.6922,
+      "step": 9825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1680593315676944e-05,
+      "loss": 18.7792,
+      "step": 9826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1671384780963874e-05,
+      "loss": 18.9408,
+      "step": 9827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.166217653680058e-05,
+      "loss": 18.1112,
+      "step": 9828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.16529685835083e-05,
+      "loss": 18.4802,
+      "step": 9829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.164376092140828e-05,
+      "loss": 18.0381,
+      "step": 9830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.163455355082171e-05,
+      "loss": 20.3324,
+      "step": 9831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.162534647206983e-05,
+      "loss": 0.674,
+      "step": 9832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1616139685473816e-05,
+      "loss": 0.6639,
+      "step": 9833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.160693319135484e-05,
+      "loss": 18.2035,
+      "step": 9834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1597726990034085e-05,
+      "loss": 16.852,
+      "step": 9835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1588521081832735e-05,
+      "loss": 18.0999,
+      "step": 9836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.157931546707193e-05,
+      "loss": 0.64,
+      "step": 9837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.15701101460728e-05,
+      "loss": 0.7268,
+      "step": 9838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1560905119156506e-05,
+      "loss": 0.7208,
+      "step": 9839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.155170038664415e-05,
+      "loss": 0.6899,
+      "step": 9840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154249594885687e-05,
+      "loss": 0.6891,
+      "step": 9841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.153329180611575e-05,
+      "loss": 18.8189,
+      "step": 9842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1524087958741884e-05,
+      "loss": 0.686,
+      "step": 9843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1514884407056375e-05,
+      "loss": 18.116,
+      "step": 9844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.150568115138028e-05,
+      "loss": 18.2373,
+      "step": 9845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1496478192034646e-05,
+      "loss": 0.6657,
+      "step": 9846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.148727552934055e-05,
+      "loss": 0.6455,
+      "step": 9847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.147807316361902e-05,
+      "loss": 17.6012,
+      "step": 9848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14688710951911e-05,
+      "loss": 17.6339,
+      "step": 9849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14596693243778e-05,
+      "loss": 19.4458,
+      "step": 9850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1450467851500127e-05,
+      "loss": 18.9084,
+      "step": 9851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.144126667687909e-05,
+      "loss": 0.6826,
+      "step": 9852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.143206580083567e-05,
+      "loss": 18.955,
+      "step": 9853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.142286522369087e-05,
+      "loss": 0.6974,
+      "step": 9854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.141366494576563e-05,
+      "loss": 19.2383,
+      "step": 9855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1404464967380914e-05,
+      "loss": 18.1376,
+      "step": 9856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139526528885768e-05,
+      "loss": 0.666,
+      "step": 9857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.138606591051686e-05,
+      "loss": 0.71,
+      "step": 9858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1376866832679385e-05,
+      "loss": 18.8112,
+      "step": 9859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1367668055666165e-05,
+      "loss": 0.7209,
+      "step": 9860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.135846957979811e-05,
+      "loss": 0.6776,
+      "step": 9861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.134927140539611e-05,
+      "loss": 19.7342,
+      "step": 9862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1340073532781054e-05,
+      "loss": 18.3552,
+      "step": 9863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1330875962273825e-05,
+      "loss": 0.646,
+      "step": 9864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.132167869419528e-05,
+      "loss": 18.8653,
+      "step": 9865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1312481728866257e-05,
+      "loss": 17.1574,
+      "step": 9866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.130328506660763e-05,
+      "loss": 0.7209,
+      "step": 9867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.12940887077402e-05,
+      "loss": 16.6493,
+      "step": 9868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1284892652584804e-05,
+      "loss": 18.3233,
+      "step": 9869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1275696901462266e-05,
+      "loss": 18.7378,
+      "step": 9870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1266501454693356e-05,
+      "loss": 17.9083,
+      "step": 9871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.125730631259889e-05,
+      "loss": 0.6863,
+      "step": 9872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.124811147549964e-05,
+      "loss": 0.6057,
+      "step": 9873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1238916943716364e-05,
+      "loss": 0.6254,
+      "step": 9874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122972271756983e-05,
+      "loss": 17.4523,
+      "step": 9875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122052879738078e-05,
+      "loss": 19.6074,
+      "step": 9876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.121133518346997e-05,
+      "loss": 18.5591,
+      "step": 9877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120214187615809e-05,
+      "loss": 18.7743,
+      "step": 9878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1192948875765885e-05,
+      "loss": 19.3574,
+      "step": 9879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.118375618261405e-05,
+      "loss": 0.6642,
+      "step": 9880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117456379702328e-05,
+      "loss": 0.7018,
+      "step": 9881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.116537171931425e-05,
+      "loss": 18.0713,
+      "step": 9882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1156179949807636e-05,
+      "loss": 17.1349,
+      "step": 9883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1146988488824115e-05,
+      "loss": 0.7453,
+      "step": 9884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.113779733668432e-05,
+      "loss": 0.6951,
+      "step": 9885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.112860649370888e-05,
+      "loss": 19.3465,
+      "step": 9886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.111941596021846e-05,
+      "loss": 0.6671,
+      "step": 9887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1110225736533664e-05,
+      "loss": 19.5155,
+      "step": 9888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1101035822975085e-05,
+      "loss": 20.2346,
+      "step": 9889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1091846219863317e-05,
+      "loss": 17.1334,
+      "step": 9890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.108265692751897e-05,
+      "loss": 17.4843,
+      "step": 9891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.10734679462626e-05,
+      "loss": 0.7227,
+      "step": 9892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1064279276414784e-05,
+      "loss": 18.9359,
+      "step": 9893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.105509091829607e-05,
+      "loss": 0.6992,
+      "step": 9894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.104590287222699e-05,
+      "loss": 18.3283,
+      "step": 9895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1036715138528106e-05,
+      "loss": 0.7588,
+      "step": 9896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1027527717519896e-05,
+      "loss": 0.7151,
+      "step": 9897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1018340609522906e-05,
+      "loss": 19.8408,
+      "step": 9898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.100915381485762e-05,
+      "loss": 17.0291,
+      "step": 9899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.099996733384452e-05,
+      "loss": 18.1861,
+      "step": 9900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.09907811668041e-05,
+      "loss": 0.6465,
+      "step": 9901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098159531405681e-05,
+      "loss": 17.49,
+      "step": 9902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0972409775923114e-05,
+      "loss": 0.7419,
+      "step": 9903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.096322455272346e-05,
+      "loss": 0.7464,
+      "step": 9904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0954039644778266e-05,
+      "loss": 0.7249,
+      "step": 9905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0944855052407967e-05,
+      "loss": 0.7035,
+      "step": 9906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.093567077593297e-05,
+      "loss": 0.6771,
+      "step": 9907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.092648681567368e-05,
+      "loss": 19.181,
+      "step": 9908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.091730317195048e-05,
+      "loss": 17.7644,
+      "step": 9909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0908119845083736e-05,
+      "loss": 16.8665,
+      "step": 9910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.089893683539385e-05,
+      "loss": 17.8349,
+      "step": 9911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0889754143201135e-05,
+      "loss": 0.6749,
+      "step": 9912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.088057176882598e-05,
+      "loss": 18.8342,
+      "step": 9913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0871389712588686e-05,
+      "loss": 18.257,
+      "step": 9914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.086220797480958e-05,
+      "loss": 17.5988,
+      "step": 9915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.085302655580899e-05,
+      "loss": 0.7018,
+      "step": 9916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.084384545590719e-05,
+      "loss": 17.8483,
+      "step": 9917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0834664675424505e-05,
+      "loss": 17.2741,
+      "step": 9918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.082548421468119e-05,
+      "loss": 0.6507,
+      "step": 9919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0816304073997505e-05,
+      "loss": 18.3957,
+      "step": 9920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.080712425369373e-05,
+      "loss": 17.3778,
+      "step": 9921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.079794475409009e-05,
+      "loss": 0.6435,
+      "step": 9922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078876557550683e-05,
+      "loss": 19.0711,
+      "step": 9923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.077958671826416e-05,
+      "loss": 18.298,
+      "step": 9924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.077040818268229e-05,
+      "loss": 19.0762,
+      "step": 9925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.076122996908145e-05,
+      "loss": 17.5191,
+      "step": 9926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.075205207778179e-05,
+      "loss": 15.766,
+      "step": 9927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.074287450910351e-05,
+      "loss": 0.7163,
+      "step": 9928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.073369726336677e-05,
+      "loss": 0.6573,
+      "step": 9929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072452034089172e-05,
+      "loss": 19.0916,
+      "step": 9930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.071534374199852e-05,
+      "loss": 0.6656,
+      "step": 9931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.070616746700729e-05,
+      "loss": 17.84,
+      "step": 9932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.069699151623814e-05,
+      "loss": 18.0119,
+      "step": 9933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06878158900112e-05,
+      "loss": 18.4211,
+      "step": 9934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.067864058864655e-05,
+      "loss": 17.8203,
+      "step": 9935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06694656124643e-05,
+      "loss": 18.8762,
+      "step": 9936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066029096178451e-05,
+      "loss": 19.2756,
+      "step": 9937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.065111663692724e-05,
+      "loss": 18.2075,
+      "step": 9938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.064194263821256e-05,
+      "loss": 17.8911,
+      "step": 9939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06327689659605e-05,
+      "loss": 19.3877,
+      "step": 9940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.062359562049109e-05,
+      "loss": 20.9169,
+      "step": 9941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.061442260212437e-05,
+      "loss": 17.1404,
+      "step": 9942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.060524991118031e-05,
+      "loss": 16.8071,
+      "step": 9943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.059607754797892e-05,
+      "loss": 17.97,
+      "step": 9944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0586905512840194e-05,
+      "loss": 0.6375,
+      "step": 9945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.057773380608411e-05,
+      "loss": 0.6367,
+      "step": 9946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0568562428030596e-05,
+      "loss": 17.0886,
+      "step": 9947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055939137899964e-05,
+      "loss": 0.6911,
+      "step": 9948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055022065931115e-05,
+      "loss": 0.695,
+      "step": 9949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0541050269285096e-05,
+      "loss": 0.7369,
+      "step": 9950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053188020924136e-05,
+      "loss": 20.2886,
+      "step": 9951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0522710479499834e-05,
+      "loss": 0.6715,
+      "step": 9952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0513541080380455e-05,
+      "loss": 0.7341,
+      "step": 9953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.050437201220306e-05,
+      "loss": 19.6985,
+      "step": 9954
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.049520327528755e-05,
+      "loss": 19.4677,
+      "step": 9955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.048603486995378e-05,
+      "loss": 16.5154,
+      "step": 9956
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.047686679652156e-05,
+      "loss": 18.3938,
+      "step": 9957
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.046769905531077e-05,
+      "loss": 18.6312,
+      "step": 9958
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.045853164664121e-05,
+      "loss": 18.1955,
+      "step": 9959
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044936457083271e-05,
+      "loss": 16.2001,
+      "step": 9960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044019782820505e-05,
+      "loss": 17.3839,
+      "step": 9961
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.043103141907802e-05,
+      "loss": 18.4704,
+      "step": 9962
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.042186534377141e-05,
+      "loss": 17.1273,
+      "step": 9963
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.041269960260497e-05,
+      "loss": 18.8424,
+      "step": 9964
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0403534195898466e-05,
+      "loss": 17.7909,
+      "step": 9965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.039436912397164e-05,
+      "loss": 18.4084,
+      "step": 9966
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0385204387144206e-05,
+      "loss": 18.1522,
+      "step": 9967
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.037603998573591e-05,
+      "loss": 17.3016,
+      "step": 9968
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0366875920066425e-05,
+      "loss": 19.0655,
+      "step": 9969
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0357712190455474e-05,
+      "loss": 17.2498,
+      "step": 9970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.034854879722274e-05,
+      "loss": 16.9923,
+      "step": 9971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.033938574068787e-05,
+      "loss": 18.0827,
+      "step": 9972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0330223021170543e-05,
+      "loss": 18.8744,
+      "step": 9973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.03210606389904e-05,
+      "loss": 0.6913,
+      "step": 9974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0311898594467086e-05,
+      "loss": 17.627,
+      "step": 9975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0302736887920225e-05,
+      "loss": 17.0735,
+      "step": 9976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.02935755196694e-05,
+      "loss": 19.6875,
+      "step": 9977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0284414490034264e-05,
+      "loss": 18.0136,
+      "step": 9978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0275253799334355e-05,
+      "loss": 19.9286,
+      "step": 9979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0266093447889296e-05,
+      "loss": 0.6717,
+      "step": 9980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.025693343601863e-05,
+      "loss": 19.6117,
+      "step": 9981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.024777376404189e-05,
+      "loss": 21.9948,
+      "step": 9982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.023861443227865e-05,
+      "loss": 17.6438,
+      "step": 9983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.022945544104843e-05,
+      "loss": 21.0132,
+      "step": 9984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.022029679067075e-05,
+      "loss": 17.5881,
+      "step": 9985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.021113848146512e-05,
+      "loss": 0.6458,
+      "step": 9986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.020198051375101e-05,
+      "loss": 17.716,
+      "step": 9987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0192822887847934e-05,
+      "loss": 0.7257,
+      "step": 9988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.018366560407534e-05,
+      "loss": 20.8845,
+      "step": 9989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.01745086627527e-05,
+      "loss": 16.179,
+      "step": 9990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0165352064199455e-05,
+      "loss": 0.6666,
+      "step": 9991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015619580873503e-05,
+      "loss": 18.1731,
+      "step": 9992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.014703989667887e-05,
+      "loss": 18.4052,
+      "step": 9993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.013788432835036e-05,
+      "loss": 0.7032,
+      "step": 9994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.012872910406892e-05,
+      "loss": 0.6931,
+      "step": 9995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0119574224153944e-05,
+      "loss": 18.2058,
+      "step": 9996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.011041968892478e-05,
+      "loss": 19.8687,
+      "step": 9997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.010126549870079e-05,
+      "loss": 0.714,
+      "step": 9998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.009211165380135e-05,
+      "loss": 17.8387,
+      "step": 9999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0082958154545777e-05,
+      "loss": 0.6557,
+      "step": 10000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.007380500125341e-05,
+      "loss": 0.7364,
+      "step": 10001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.006465219424355e-05,
+      "loss": 16.9675,
+      "step": 10002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.005549973383551e-05,
+      "loss": 0.6883,
+      "step": 10003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0046347620348586e-05,
+      "loss": 0.6952,
+      "step": 10004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0037195854102043e-05,
+      "loss": 17.9461,
+      "step": 10005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002804443541515e-05,
+      "loss": 0.6675,
+      "step": 10006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.001889336460717e-05,
+      "loss": 18.8607,
+      "step": 10007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0009742641997325e-05,
+      "loss": 15.1942,
+      "step": 10008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.000059226790487e-05,
+      "loss": 18.2992,
+      "step": 10009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9991442242649006e-05,
+      "loss": 0.6552,
+      "step": 10010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9982292566548934e-05,
+      "loss": 19.055,
+      "step": 10011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9973143239923874e-05,
+      "loss": 19.9944,
+      "step": 10012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.996399426309296e-05,
+      "loss": 18.2593,
+      "step": 10013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9954845636375415e-05,
+      "loss": 0.6835,
+      "step": 10014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994569736009036e-05,
+      "loss": 20.3339,
+      "step": 10015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9936549434556945e-05,
+      "loss": 0.6909,
+      "step": 10016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.992740186009431e-05,
+      "loss": 17.0538,
+      "step": 10017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.991825463702157e-05,
+      "loss": 19.0156,
+      "step": 10018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.990910776565784e-05,
+      "loss": 0.717,
+      "step": 10019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98999612463222e-05,
+      "loss": 16.8414,
+      "step": 10020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.989081507933374e-05,
+      "loss": 19.3275,
+      "step": 10021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9881669265011534e-05,
+      "loss": 19.9809,
+      "step": 10022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9872523803674636e-05,
+      "loss": 0.7022,
+      "step": 10023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.986337869564209e-05,
+      "loss": 0.6391,
+      "step": 10024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985423394123295e-05,
+      "loss": 18.45,
+      "step": 10025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.984508954076621e-05,
+      "loss": 18.0149,
+      "step": 10026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.983594549456089e-05,
+      "loss": 18.6129,
+      "step": 10027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9826801802935984e-05,
+      "loss": 0.6543,
+      "step": 10028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.981765846621049e-05,
+      "loss": 0.7324,
+      "step": 10029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.980851548470337e-05,
+      "loss": 17.3025,
+      "step": 10030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.979937285873356e-05,
+      "loss": 18.1985,
+      "step": 10031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.979023058862006e-05,
+      "loss": 18.5347,
+      "step": 10032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9781088674681764e-05,
+      "loss": 0.7027,
+      "step": 10033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.977194711723759e-05,
+      "loss": 18.3044,
+      "step": 10034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9762805916606484e-05,
+      "loss": 20.2389,
+      "step": 10035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.97536650731073e-05,
+      "loss": 21.1257,
+      "step": 10036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.974452458705897e-05,
+      "loss": 18.1634,
+      "step": 10037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9735384458780334e-05,
+      "loss": 19.4419,
+      "step": 10038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.972624468859025e-05,
+      "loss": 0.6643,
+      "step": 10039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.971710527680759e-05,
+      "loss": 19.0649,
+      "step": 10040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.970796622375116e-05,
+      "loss": 15.9813,
+      "step": 10041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.96988275297398e-05,
+      "loss": 16.4026,
+      "step": 10042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9689689195092325e-05,
+      "loss": 17.834,
+      "step": 10043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.968055122012751e-05,
+      "loss": 17.3038,
+      "step": 10044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.967141360516417e-05,
+      "loss": 18.6416,
+      "step": 10045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.966227635052105e-05,
+      "loss": 18.1613,
+      "step": 10046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.965313945651693e-05,
+      "loss": 0.6424,
+      "step": 10047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.964400292347055e-05,
+      "loss": 0.7942,
+      "step": 10048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9634866751700625e-05,
+      "loss": 18.5579,
+      "step": 10049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9625730941525915e-05,
+      "loss": 18.2477,
+      "step": 10050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.961659549326512e-05,
+      "loss": 17.5833,
+      "step": 10051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960746040723691e-05,
+      "loss": 18.1156,
+      "step": 10052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.959832568375997e-05,
+      "loss": 18.4536,
+      "step": 10053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.958919132315301e-05,
+      "loss": 0.6681,
+      "step": 10054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9580057325734644e-05,
+      "loss": 19.0961,
+      "step": 10055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.957092369182355e-05,
+      "loss": 0.7159,
+      "step": 10056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.956179042173836e-05,
+      "loss": 0.7006,
+      "step": 10057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.955265751579767e-05,
+      "loss": 18.0676,
+      "step": 10058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.954352497432011e-05,
+      "loss": 16.597,
+      "step": 10059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9534392797624265e-05,
+      "loss": 18.9296,
+      "step": 10060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952526098602873e-05,
+      "loss": 16.4147,
+      "step": 10061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.951612953985207e-05,
+      "loss": 17.887,
+      "step": 10062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.950699845941282e-05,
+      "loss": 16.4305,
+      "step": 10063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9497867745029563e-05,
+      "loss": 0.7667,
+      "step": 10064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9488737397020797e-05,
+      "loss": 18.7742,
+      "step": 10065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.947960741570507e-05,
+      "loss": 16.6994,
+      "step": 10066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.947047780140086e-05,
+      "loss": 18.2132,
+      "step": 10067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.946134855442667e-05,
+      "loss": 18.7397,
+      "step": 10068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9452219675100994e-05,
+      "loss": 17.9759,
+      "step": 10069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9443091163742274e-05,
+      "loss": 19.1577,
+      "step": 10070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9433963020668994e-05,
+      "loss": 0.7599,
+      "step": 10071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942483524619958e-05,
+      "loss": 17.135,
+      "step": 10072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9415707840652446e-05,
+      "loss": 17.9795,
+      "step": 10073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9406580804346046e-05,
+      "loss": 0.661,
+      "step": 10074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9397454137598745e-05,
+      "loss": 19.0614,
+      "step": 10075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9388327840728966e-05,
+      "loss": 0.7323,
+      "step": 10076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9379201914055066e-05,
+      "loss": 0.6575,
+      "step": 10077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.93700763578954e-05,
+      "loss": 0.6646,
+      "step": 10078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.936095117256835e-05,
+      "loss": 17.357,
+      "step": 10079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.935182635839223e-05,
+      "loss": 16.8683,
+      "step": 10080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.934270191568538e-05,
+      "loss": 20.1783,
+      "step": 10081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9333577844766114e-05,
+      "loss": 18.8651,
+      "step": 10082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.932445414595271e-05,
+      "loss": 0.7036,
+      "step": 10083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.931533081956348e-05,
+      "loss": 0.6862,
+      "step": 10084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.930620786591668e-05,
+      "loss": 18.3094,
+      "step": 10085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.929708528533058e-05,
+      "loss": 17.5033,
+      "step": 10086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9287963078123434e-05,
+      "loss": 16.0016,
+      "step": 10087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9278841244613465e-05,
+      "loss": 0.762,
+      "step": 10088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9269719785118904e-05,
+      "loss": 18.3016,
+      "step": 10089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.926059869995794e-05,
+      "loss": 0.7114,
+      "step": 10090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.92514779894488e-05,
+      "loss": 18.7195,
+      "step": 10091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.924235765390964e-05,
+      "loss": 17.4188,
+      "step": 10092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.923323769365864e-05,
+      "loss": 19.4334,
+      "step": 10093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9224118109013966e-05,
+      "loss": 18.6342,
+      "step": 10094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.921499890029374e-05,
+      "loss": 17.316,
+      "step": 10095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.920588006781611e-05,
+      "loss": 20.351,
+      "step": 10096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9196761611899194e-05,
+      "loss": 17.9428,
+      "step": 10097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918764353286107e-05,
+      "loss": 0.7334,
+      "step": 10098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9178525831019857e-05,
+      "loss": 18.2613,
+      "step": 10099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916940850669362e-05,
+      "loss": 0.6748,
+      "step": 10100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916029156020043e-05,
+      "loss": 16.6485,
+      "step": 10101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9151174991858343e-05,
+      "loss": 19.282,
+      "step": 10102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.914205880198537e-05,
+      "loss": 18.6531,
+      "step": 10103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.913294299089957e-05,
+      "loss": 18.8324,
+      "step": 10104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.912382755891893e-05,
+      "loss": 17.7234,
+      "step": 10105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9114712506361486e-05,
+      "loss": 17.9877,
+      "step": 10106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910559783354517e-05,
+      "loss": 18.2161,
+      "step": 10107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.909648354078799e-05,
+      "loss": 0.669,
+      "step": 10108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.908736962840787e-05,
+      "loss": 18.2123,
+      "step": 10109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9078256096722795e-05,
+      "loss": 18.565,
+      "step": 10110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.906914294605069e-05,
+      "loss": 19.106,
+      "step": 10111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.906003017670945e-05,
+      "loss": 18.4929,
+      "step": 10112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9050917789017013e-05,
+      "loss": 0.7246,
+      "step": 10113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9041805783291234e-05,
+      "loss": 0.7156,
+      "step": 10114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9032694159850025e-05,
+      "loss": 17.532,
+      "step": 10115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.902358291901124e-05,
+      "loss": 19.891,
+      "step": 10116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.901447206109272e-05,
+      "loss": 0.7037,
+      "step": 10117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9005361586412325e-05,
+      "loss": 19.1339,
+      "step": 10118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8996251495287856e-05,
+      "loss": 18.1081,
+      "step": 10119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8987141788037154e-05,
+      "loss": 0.6677,
+      "step": 10120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8978032464978e-05,
+      "loss": 18.8241,
+      "step": 10121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8968923526428166e-05,
+      "loss": 0.7016,
+      "step": 10122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8959814972705455e-05,
+      "loss": 18.226,
+      "step": 10123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.89507068041276e-05,
+      "loss": 0.6864,
+      "step": 10124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8941599021012364e-05,
+      "loss": 17.0057,
+      "step": 10125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8932491623677477e-05,
+      "loss": 0.6762,
+      "step": 10126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.892338461244063e-05,
+      "loss": 19.5544,
+      "step": 10127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.891427798761957e-05,
+      "loss": 0.6462,
+      "step": 10128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.890517174953196e-05,
+      "loss": 19.195,
+      "step": 10129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.889606589849548e-05,
+      "loss": 16.9263,
+      "step": 10130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.88869604348278e-05,
+      "loss": 18.4475,
+      "step": 10131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.887785535884657e-05,
+      "loss": 19.1241,
+      "step": 10132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.886875067086943e-05,
+      "loss": 18.5516,
+      "step": 10133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8859646371214e-05,
+      "loss": 0.6829,
+      "step": 10134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.885054246019788e-05,
+      "loss": 18.8412,
+      "step": 10135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8841438938138694e-05,
+      "loss": 0.6768,
+      "step": 10136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.883233580535399e-05,
+      "loss": 18.8857,
+      "step": 10137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.882323306216137e-05,
+      "loss": 0.6217,
+      "step": 10138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.881413070887838e-05,
+      "loss": 17.4536,
+      "step": 10139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8805028745822534e-05,
+      "loss": 0.6983,
+      "step": 10140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8795927173311404e-05,
+      "loss": 18.3806,
+      "step": 10141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8786825991662476e-05,
+      "loss": 0.7354,
+      "step": 10142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.877772520119326e-05,
+      "loss": 19.6914,
+      "step": 10143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.876862480222125e-05,
+      "loss": 18.4641,
+      "step": 10144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8759524795063904e-05,
+      "loss": 17.6438,
+      "step": 10145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8750425180038705e-05,
+      "loss": 0.7107,
+      "step": 10146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8741325957463075e-05,
+      "loss": 18.5339,
+      "step": 10147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873222712765447e-05,
+      "loss": 0.7142,
+      "step": 10148
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8723128690930296e-05,
+      "loss": 18.4272,
+      "step": 10149
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8714030647607955e-05,
+      "loss": 18.1867,
+      "step": 10150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8704932998004844e-05,
+      "loss": 0.7254,
+      "step": 10151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.869583574243834e-05,
+      "loss": 0.7094,
+      "step": 10152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.868673888122582e-05,
+      "loss": 17.6534,
+      "step": 10153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867764241468462e-05,
+      "loss": 17.8296,
+      "step": 10154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8668546343132075e-05,
+      "loss": 18.0726,
+      "step": 10155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.865945066688552e-05,
+      "loss": 0.7496,
+      "step": 10156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.865035538626225e-05,
+      "loss": 18.7879,
+      "step": 10157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8641260501579575e-05,
+      "loss": 19.927,
+      "step": 10158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863216601315477e-05,
+      "loss": 18.4063,
+      "step": 10159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.862307192130512e-05,
+      "loss": 17.8845,
+      "step": 10160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861397822634784e-05,
+      "loss": 0.7167,
+      "step": 10161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.860488492860019e-05,
+      "loss": 16.251,
+      "step": 10162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8595792028379404e-05,
+      "loss": 0.7301,
+      "step": 10163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.858669952600269e-05,
+      "loss": 20.1421,
+      "step": 10164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8577607421787246e-05,
+      "loss": 19.01,
+      "step": 10165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.856851571605025e-05,
+      "loss": 15.9122,
+      "step": 10166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.855942440910888e-05,
+      "loss": 0.6086,
+      "step": 10167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.85503335012803e-05,
+      "loss": 22.0325,
+      "step": 10168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8541242992881624e-05,
+      "loss": 0.6474,
+      "step": 10169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.853215288423002e-05,
+      "loss": 18.1943,
+      "step": 10170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.852306317564258e-05,
+      "loss": 18.5534,
+      "step": 10171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.851397386743641e-05,
+      "loss": 0.7661,
+      "step": 10172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.85048849599286e-05,
+      "loss": 18.2145,
+      "step": 10173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8495796453436206e-05,
+      "loss": 0.6488,
+      "step": 10174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.848670834827631e-05,
+      "loss": 0.6398,
+      "step": 10175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8477620644765945e-05,
+      "loss": 18.9378,
+      "step": 10176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.846853334322215e-05,
+      "loss": 18.1299,
+      "step": 10177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8459446443961944e-05,
+      "loss": 16.8515,
+      "step": 10178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8450359947302304e-05,
+      "loss": 16.5851,
+      "step": 10179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844127385356026e-05,
+      "loss": 0.7078,
+      "step": 10180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.843218816305274e-05,
+      "loss": 0.7255,
+      "step": 10181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842310287609675e-05,
+      "loss": 18.1549,
+      "step": 10182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.841401799300922e-05,
+      "loss": 18.5534,
+      "step": 10183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.840493351410706e-05,
+      "loss": 18.9736,
+      "step": 10184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.839584943970722e-05,
+      "loss": 19.2501,
+      "step": 10185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8386765770126585e-05,
+      "loss": 17.1567,
+      "step": 10186
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.837768250568207e-05,
+      "loss": 17.43,
+      "step": 10187
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.836859964669053e-05,
+      "loss": 0.7251,
+      "step": 10188
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835951719346882e-05,
+      "loss": 0.6749,
+      "step": 10189
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835043514633381e-05,
+      "loss": 17.5748,
+      "step": 10190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.834135350560232e-05,
+      "loss": 16.4188,
+      "step": 10191
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8332272271591176e-05,
+      "loss": 17.6912,
+      "step": 10192
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.832319144461718e-05,
+      "loss": 19.193,
+      "step": 10193
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.831411102499712e-05,
+      "loss": 0.6257,
+      "step": 10194
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.830503101304778e-05,
+      "loss": 17.237,
+      "step": 10195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.829595140908591e-05,
+      "loss": 18.5436,
+      "step": 10196
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8286872213428285e-05,
+      "loss": 0.6527,
+      "step": 10197
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.827779342639162e-05,
+      "loss": 0.7013,
+      "step": 10198
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8268715048292624e-05,
+      "loss": 18.7276,
+      "step": 10199
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.825963707944803e-05,
+      "loss": 18.0727,
+      "step": 10200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.82505595201745e-05,
+      "loss": 17.5111,
+      "step": 10201
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.824148237078875e-05,
+      "loss": 0.717,
+      "step": 10202
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.823240563160741e-05,
+      "loss": 0.6911,
+      "step": 10203
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.822332930294713e-05,
+      "loss": 18.7252,
+      "step": 10204
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8214253385124565e-05,
+      "loss": 0.6587,
+      "step": 10205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.820517787845631e-05,
+      "loss": 19.5719,
+      "step": 10206
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8196102783258994e-05,
+      "loss": 16.5742,
+      "step": 10207
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.81870280998492e-05,
+      "loss": 17.4161,
+      "step": 10208
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8177953828543496e-05,
+      "loss": 17.3204,
+      "step": 10209
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.816887996965846e-05,
+      "loss": 19.8283,
+      "step": 10210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815980652351062e-05,
+      "loss": 17.9047,
+      "step": 10211
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815073349041653e-05,
+      "loss": 17.0511,
+      "step": 10212
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8141660870692704e-05,
+      "loss": 0.6927,
+      "step": 10213
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8132588664655636e-05,
+      "loss": 18.6414,
+      "step": 10214
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.812351687262183e-05,
+      "loss": 0.6541,
+      "step": 10215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.811444549490776e-05,
+      "loss": 17.4169,
+      "step": 10216
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.810537453182988e-05,
+      "loss": 16.6369,
+      "step": 10217
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.809630398370463e-05,
+      "loss": 0.6432,
+      "step": 10218
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.808723385084847e-05,
+      "loss": 18.5982,
+      "step": 10219
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8078164133577786e-05,
+      "loss": 18.388,
+      "step": 10220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8069094832209005e-05,
+      "loss": 18.5784,
+      "step": 10221
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.806002594705851e-05,
+      "loss": 0.688,
+      "step": 10222
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8050957478442665e-05,
+      "loss": 0.6921,
+      "step": 10223
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8041889426677854e-05,
+      "loss": 18.2478,
+      "step": 10224
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8032821792080394e-05,
+      "loss": 17.8231,
+      "step": 10225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.802375457496664e-05,
+      "loss": 19.6717,
+      "step": 10226
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8014687775652896e-05,
+      "loss": 0.6827,
+      "step": 10227
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.800562139445546e-05,
+      "loss": 0.6958,
+      "step": 10228
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.799655543169063e-05,
+      "loss": 17.8606,
+      "step": 10229
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7987489887674684e-05,
+      "loss": 0.6149,
+      "step": 10230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.797842476272385e-05,
+      "loss": 19.6942,
+      "step": 10231
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.796936005715441e-05,
+      "loss": 0.739,
+      "step": 10232
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7960295771282556e-05,
+      "loss": 18.7509,
+      "step": 10233
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7951231905424534e-05,
+      "loss": 18.1659,
+      "step": 10234
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7942168459896535e-05,
+      "loss": 0.7582,
+      "step": 10235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.793310543501473e-05,
+      "loss": 18.6554,
+      "step": 10236
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.79240428310953e-05,
+      "loss": 19.4363,
+      "step": 10237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.791498064845439e-05,
+      "loss": 19.4665,
+      "step": 10238
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7905918887408163e-05,
+      "loss": 18.0284,
+      "step": 10239
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.789685754827273e-05,
+      "loss": 0.6109,
+      "step": 10240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.78877966313642e-05,
+      "loss": 0.6702,
+      "step": 10241
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7878736136998674e-05,
+      "loss": 0.6874,
+      "step": 10242
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.786967606549223e-05,
+      "loss": 0.7466,
+      "step": 10243
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.786061641716095e-05,
+      "loss": 0.7135,
+      "step": 10244
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7851557192320875e-05,
+      "loss": 17.1055,
+      "step": 10245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7842498391288025e-05,
+      "loss": 19.1472,
+      "step": 10246
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7833440014378464e-05,
+      "loss": 0.6324,
+      "step": 10247
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.782438206190816e-05,
+      "loss": 17.955,
+      "step": 10248
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.781532453419313e-05,
+      "loss": 0.6849,
+      "step": 10249
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7806267431549355e-05,
+      "loss": 18.3061,
+      "step": 10250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7797210754292765e-05,
+      "loss": 18.1336,
+      "step": 10251
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7788154502739356e-05,
+      "loss": 16.7003,
+      "step": 10252
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.777909867720502e-05,
+      "loss": 19.1484,
+      "step": 10253
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.777004327800571e-05,
+      "loss": 0.6193,
+      "step": 10254
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.776098830545732e-05,
+      "loss": 17.4797,
+      "step": 10255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7751933759875714e-05,
+      "loss": 19.2002,
+      "step": 10256
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.77428796415768e-05,
+      "loss": 0.6517,
+      "step": 10257
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.773382595087642e-05,
+      "loss": 17.2348,
+      "step": 10258
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7724772688090425e-05,
+      "loss": 0.7635,
+      "step": 10259
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.771571985353465e-05,
+      "loss": 17.0469,
+      "step": 10260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7706667447524876e-05,
+      "loss": 17.6279,
+      "step": 10261
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7697615470376956e-05,
+      "loss": 17.3278,
+      "step": 10262
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.768856392240663e-05,
+      "loss": 17.9383,
+      "step": 10263
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.76795128039297e-05,
+      "loss": 0.6603,
+      "step": 10264
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7670462115261906e-05,
+      "loss": 0.6154,
+      "step": 10265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7661411856718974e-05,
+      "loss": 0.6795,
+      "step": 10266
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.765236202861666e-05,
+      "loss": 0.6794,
+      "step": 10267
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7643312631270635e-05,
+      "loss": 16.6726,
+      "step": 10268
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7634263664996655e-05,
+      "loss": 17.267,
+      "step": 10269
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.762521513011033e-05,
+      "loss": 19.0734,
+      "step": 10270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7616167026927364e-05,
+      "loss": 19.5848,
+      "step": 10271
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.760711935576338e-05,
+      "loss": 0.6858,
+      "step": 10272
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7598072116934056e-05,
+      "loss": 16.0679,
+      "step": 10273
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.758902531075497e-05,
+      "loss": 18.9451,
+      "step": 10274
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.757997893754174e-05,
+      "loss": 0.6594,
+      "step": 10275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7570932997609964e-05,
+      "loss": 0.6167,
+      "step": 10276
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.756188749127521e-05,
+      "loss": 18.0641,
+      "step": 10277
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7552842418853034e-05,
+      "loss": 0.6758,
+      "step": 10278
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.754379778065899e-05,
+      "loss": 18.1718,
+      "step": 10279
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7534753577008584e-05,
+      "loss": 0.6805,
+      "step": 10280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7525709808217354e-05,
+      "loss": 19.4024,
+      "step": 10281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7516666474600787e-05,
+      "loss": 18.9875,
+      "step": 10282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750762357647437e-05,
+      "loss": 17.7536,
+      "step": 10283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.749858111415358e-05,
+      "loss": 16.9743,
+      "step": 10284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.748953908795384e-05,
+      "loss": 18.0792,
+      "step": 10285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.748049749819063e-05,
+      "loss": 19.5831,
+      "step": 10286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7471456345179324e-05,
+      "loss": 15.9652,
+      "step": 10287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.746241562923537e-05,
+      "loss": 18.3515,
+      "step": 10288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7453375350674155e-05,
+      "loss": 0.6335,
+      "step": 10289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7444335509811024e-05,
+      "loss": 19.6694,
+      "step": 10290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.743529610696137e-05,
+      "loss": 17.2843,
+      "step": 10291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.742625714244052e-05,
+      "loss": 18.4931,
+      "step": 10292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7417218616563835e-05,
+      "loss": 18.5732,
+      "step": 10293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7408180529646596e-05,
+      "loss": 18.399,
+      "step": 10294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.739914288200411e-05,
+      "loss": 0.652,
+      "step": 10295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.739010567395168e-05,
+      "loss": 18.522,
+      "step": 10296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.738106890580456e-05,
+      "loss": 17.0295,
+      "step": 10297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.737203257787801e-05,
+      "loss": 16.1841,
+      "step": 10298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.736299669048727e-05,
+      "loss": 0.6708,
+      "step": 10299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7353961243947544e-05,
+      "loss": 0.6145,
+      "step": 10300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.734492623857407e-05,
+      "loss": 18.1938,
+      "step": 10301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.733589167468202e-05,
+      "loss": 18.56,
+      "step": 10302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7326857552586594e-05,
+      "loss": 19.5361,
+      "step": 10303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7317823872602934e-05,
+      "loss": 17.4269,
+      "step": 10304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7308790635046185e-05,
+      "loss": 18.3073,
+      "step": 10305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.729975784023148e-05,
+      "loss": 0.7107,
+      "step": 10306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.729072548847394e-05,
+      "loss": 19.2893,
+      "step": 10307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.728169358008867e-05,
+      "loss": 0.7226,
+      "step": 10308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.727266211539075e-05,
+      "loss": 20.0106,
+      "step": 10309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.726363109469525e-05,
+      "loss": 18.1382,
+      "step": 10310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7254600518317215e-05,
+      "loss": 16.4725,
+      "step": 10311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.724557038657169e-05,
+      "loss": 0.7335,
+      "step": 10312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7236540699773706e-05,
+      "loss": 17.8362,
+      "step": 10313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.722751145823825e-05,
+      "loss": 0.7537,
+      "step": 10314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.721848266228033e-05,
+      "loss": 0.6672,
+      "step": 10315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.720945431221492e-05,
+      "loss": 17.6607,
+      "step": 10316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7200426408356984e-05,
+      "loss": 18.9728,
+      "step": 10317
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.719139895102144e-05,
+      "loss": 0.7157,
+      "step": 10318
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.718237194052325e-05,
+      "loss": 0.6838,
+      "step": 10319
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7173345377177315e-05,
+      "loss": 0.701,
+      "step": 10320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.716431926129854e-05,
+      "loss": 19.1884,
+      "step": 10321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.71552935932018e-05,
+      "loss": 17.0059,
+      "step": 10322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.714626837320195e-05,
+      "loss": 0.668,
+      "step": 10323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7137243601613884e-05,
+      "loss": 0.6527,
+      "step": 10324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.71282192787524e-05,
+      "loss": 0.6843,
+      "step": 10325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.711919540493232e-05,
+      "loss": 0.6264,
+      "step": 10326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.711017198046845e-05,
+      "loss": 0.6704,
+      "step": 10327
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.710114900567559e-05,
+      "loss": 17.7581,
+      "step": 10328
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.70921264808685e-05,
+      "loss": 0.6772,
+      "step": 10329
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708310440636196e-05,
+      "loss": 0.6732,
+      "step": 10330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.707408278247069e-05,
+      "loss": 18.9026,
+      "step": 10331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.70650616095094e-05,
+      "loss": 18.4916,
+      "step": 10332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.705604088779285e-05,
+      "loss": 19.0842,
+      "step": 10333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7047020617635684e-05,
+      "loss": 18.34,
+      "step": 10334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.703800079935262e-05,
+      "loss": 18.2232,
+      "step": 10335
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.702898143325831e-05,
+      "loss": 0.6713,
+      "step": 10336
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7019962519667375e-05,
+      "loss": 18.9208,
+      "step": 10337
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.701094405889448e-05,
+      "loss": 19.1428,
+      "step": 10338
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.700192605125422e-05,
+      "loss": 0.6785,
+      "step": 10339
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6992908497061205e-05,
+      "loss": 19.3857,
+      "step": 10340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.698389139663003e-05,
+      "loss": 16.9846,
+      "step": 10341
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.697487475027522e-05,
+      "loss": 16.7,
+      "step": 10342
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.696585855831138e-05,
+      "loss": 17.0111,
+      "step": 10343
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6956842821053e-05,
+      "loss": 0.6755,
+      "step": 10344
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.694782753881465e-05,
+      "loss": 19.4554,
+      "step": 10345
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6938812711910795e-05,
+      "loss": 19.9464,
+      "step": 10346
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.692979834065594e-05,
+      "loss": 18.4707,
+      "step": 10347
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6920784425364554e-05,
+      "loss": 19.8824,
+      "step": 10348
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6911770966351076e-05,
+      "loss": 18.4073,
+      "step": 10349
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.690275796392999e-05,
+      "loss": 0.7115,
+      "step": 10350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6893745418415694e-05,
+      "loss": 18.5945,
+      "step": 10351
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.688473333012259e-05,
+      "loss": 17.5784,
+      "step": 10352
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6875721699365086e-05,
+      "loss": 18.6023,
+      "step": 10353
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6866710526457544e-05,
+      "loss": 17.9547,
+      "step": 10354
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6857699811714345e-05,
+      "loss": 0.7394,
+      "step": 10355
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.684868955544983e-05,
+      "loss": 18.6716,
+      "step": 10356
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.683967975797831e-05,
+      "loss": 0.6918,
+      "step": 10357
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.683067041961412e-05,
+      "loss": 0.6998,
+      "step": 10358
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6821661540671546e-05,
+      "loss": 17.7473,
+      "step": 10359
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.681265312146488e-05,
+      "loss": 20.2938,
+      "step": 10360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6803645162308376e-05,
+      "loss": 18.695,
+      "step": 10361
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6794637663516276e-05,
+      "loss": 15.611,
+      "step": 10362
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.678563062540283e-05,
+      "loss": 0.6742,
+      "step": 10363
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.677662404828225e-05,
+      "loss": 18.2466,
+      "step": 10364
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.676761793246874e-05,
+      "loss": 17.602,
+      "step": 10365
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.675861227827648e-05,
+      "loss": 16.4237,
+      "step": 10366
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6749607086019635e-05,
+      "loss": 17.7863,
+      "step": 10367
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.674060235601236e-05,
+      "loss": 0.7121,
+      "step": 10368
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6731598088568794e-05,
+      "loss": 0.6418,
+      "step": 10369
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.672259428400306e-05,
+      "loss": 0.6969,
+      "step": 10370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.671359094262927e-05,
+      "loss": 18.3236,
+      "step": 10371
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.670458806476148e-05,
+      "loss": 0.6663,
+      "step": 10372
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.66955856507138e-05,
+      "loss": 18.7966,
+      "step": 10373
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.668658370080026e-05,
+      "loss": 0.7196,
+      "step": 10374
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.667758221533492e-05,
+      "loss": 17.238,
+      "step": 10375
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.666858119463178e-05,
+      "loss": 0.6972,
+      "step": 10376
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6659580639004856e-05,
+      "loss": 0.6646,
+      "step": 10377
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6650580548768153e-05,
+      "loss": 18.0121,
+      "step": 10378
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664158092423564e-05,
+      "loss": 18.6463,
+      "step": 10379
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.663258176572127e-05,
+      "loss": 18.3524,
+      "step": 10380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6623583073538966e-05,
+      "loss": 18.4702,
+      "step": 10381
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.661458484800269e-05,
+      "loss": 16.0098,
+      "step": 10382
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.660558708942632e-05,
+      "loss": 17.5384,
+      "step": 10383
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.659658979812378e-05,
+      "loss": 16.5127,
+      "step": 10384
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.658759297440892e-05,
+      "loss": 18.2425,
+      "step": 10385
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.657859661859562e-05,
+      "loss": 0.7039,
+      "step": 10386
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.656960073099771e-05,
+      "loss": 0.7525,
+      "step": 10387
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.656060531192902e-05,
+      "loss": 19.8559,
+      "step": 10388
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6551610361703375e-05,
+      "loss": 0.6296,
+      "step": 10389
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.654261588063457e-05,
+      "loss": 0.7006,
+      "step": 10390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.653362186903635e-05,
+      "loss": 19.2227,
+      "step": 10391
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.652462832722252e-05,
+      "loss": 16.7782,
+      "step": 10392
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.65156352555068e-05,
+      "loss": 17.7249,
+      "step": 10393
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6506642654202944e-05,
+      "loss": 0.68,
+      "step": 10394
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6497650523624646e-05,
+      "loss": 17.3005,
+      "step": 10395
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.64886588640856e-05,
+      "loss": 0.6685,
+      "step": 10396
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.64796676758995e-05,
+      "loss": 17.8187,
+      "step": 10397
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.647067695937999e-05,
+      "loss": 0.6778,
+      "step": 10398
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.646168671484076e-05,
+      "loss": 19.6638,
+      "step": 10399
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.64526969425954e-05,
+      "loss": 16.9871,
+      "step": 10400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.644370764295753e-05,
+      "loss": 18.7644,
+      "step": 10401
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6434718816240767e-05,
+      "loss": 19.7065,
+      "step": 10402
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6425730462758664e-05,
+      "loss": 18.404,
+      "step": 10403
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6416742582824826e-05,
+      "loss": 16.9921,
+      "step": 10404
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.640775517675277e-05,
+      "loss": 19.1542,
+      "step": 10405
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.639876824485603e-05,
+      "loss": 0.6765,
+      "step": 10406
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.638978178744815e-05,
+      "loss": 19.7604,
+      "step": 10407
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6380795804842584e-05,
+      "loss": 0.7347,
+      "step": 10408
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.637181029735286e-05,
+      "loss": 18.7337,
+      "step": 10409
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.636282526529242e-05,
+      "loss": 19.4583,
+      "step": 10410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.635384070897471e-05,
+      "loss": 0.6821,
+      "step": 10411
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.634485662871317e-05,
+      "loss": 0.6865,
+      "step": 10412
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.633587302482122e-05,
+      "loss": 0.6733,
+      "step": 10413
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.632688989761225e-05,
+      "loss": 18.3355,
+      "step": 10414
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6317907247399654e-05,
+      "loss": 16.9073,
+      "step": 10415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.630892507449678e-05,
+      "loss": 17.8504,
+      "step": 10416
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6299943379217e-05,
+      "loss": 0.64,
+      "step": 10417
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.629096216187364e-05,
+      "loss": 0.6795,
+      "step": 10418
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.628198142278e-05,
+      "loss": 18.4416,
+      "step": 10419
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6273001162249404e-05,
+      "loss": 16.6449,
+      "step": 10420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.626402138059511e-05,
+      "loss": 16.2835,
+      "step": 10421
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.625504207813042e-05,
+      "loss": 18.4527,
+      "step": 10422
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.624606325516855e-05,
+      "loss": 0.7092,
+      "step": 10423
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.623708491202273e-05,
+      "loss": 16.6817,
+      "step": 10424
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.622810704900621e-05,
+      "loss": 0.7219,
+      "step": 10425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.621912966643216e-05,
+      "loss": 17.3979,
+      "step": 10426
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.621015276461378e-05,
+      "loss": 18.6498,
+      "step": 10427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.620117634386423e-05,
+      "loss": 0.7419,
+      "step": 10428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6192200404496636e-05,
+      "loss": 18.357,
+      "step": 10429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.618322494682417e-05,
+      "loss": 16.781,
+      "step": 10430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.617424997115991e-05,
+      "loss": 18.0266,
+      "step": 10431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6165275477816994e-05,
+      "loss": 0.688,
+      "step": 10432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.615630146710848e-05,
+      "loss": 0.6901,
+      "step": 10433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6147327939347435e-05,
+      "loss": 18.6672,
+      "step": 10434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6138354894846894e-05,
+      "loss": 0.6765,
+      "step": 10435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6129382333919914e-05,
+      "loss": 0.6935,
+      "step": 10436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.612041025687949e-05,
+      "loss": 20.9423,
+      "step": 10437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.611143866403862e-05,
+      "loss": 18.2231,
+      "step": 10438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6102467555710295e-05,
+      "loss": 18.7577,
+      "step": 10439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.609349693220747e-05,
+      "loss": 0.7371,
+      "step": 10440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.608452679384311e-05,
+      "loss": 0.7035,
+      "step": 10441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.607555714093013e-05,
+      "loss": 18.2837,
+      "step": 10442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.606658797378142e-05,
+      "loss": 17.1422,
+      "step": 10443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.605761929270992e-05,
+      "loss": 17.8998,
+      "step": 10444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.604865109802848e-05,
+      "loss": 17.4413,
+      "step": 10445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.603968339004998e-05,
+      "loss": 17.3783,
+      "step": 10446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6030716169087256e-05,
+      "loss": 19.0016,
+      "step": 10447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6021749435453114e-05,
+      "loss": 19.1261,
+      "step": 10448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.601278318946041e-05,
+      "loss": 0.6808,
+      "step": 10449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6003817431421895e-05,
+      "loss": 0.6842,
+      "step": 10450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.599485216165038e-05,
+      "loss": 17.8752,
+      "step": 10451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.598588738045861e-05,
+      "loss": 0.639,
+      "step": 10452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.597692308815932e-05,
+      "loss": 0.7012,
+      "step": 10453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.596795928506525e-05,
+      "loss": 0.6716,
+      "step": 10454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5958995971489096e-05,
+      "loss": 16.1238,
+      "step": 10455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.595003314774357e-05,
+      "loss": 0.6973,
+      "step": 10456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.594107081414133e-05,
+      "loss": 17.4078,
+      "step": 10457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.593210897099502e-05,
+      "loss": 0.6342,
+      "step": 10458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5923147618617316e-05,
+      "loss": 15.8647,
+      "step": 10459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5914186757320805e-05,
+      "loss": 0.7001,
+      "step": 10460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5905226387418126e-05,
+      "loss": 17.0142,
+      "step": 10461
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5896266509221846e-05,
+      "loss": 20.0999,
+      "step": 10462
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.588730712304454e-05,
+      "loss": 0.6388,
+      "step": 10463
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.587834822919878e-05,
+      "loss": 16.0523,
+      "step": 10464
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.586938982799707e-05,
+      "loss": 0.723,
+      "step": 10465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5860431919751964e-05,
+      "loss": 18.791,
+      "step": 10466
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.585147450477595e-05,
+      "loss": 18.0825,
+      "step": 10467
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.584251758338151e-05,
+      "loss": 17.6158,
+      "step": 10468
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5833561155881115e-05,
+      "loss": 18.9111,
+      "step": 10469
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5824605222587224e-05,
+      "loss": 17.5267,
+      "step": 10470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.581564978381227e-05,
+      "loss": 0.6673,
+      "step": 10471
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.580669483986866e-05,
+      "loss": 15.4341,
+      "step": 10472
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5797740391068804e-05,
+      "loss": 0.7266,
+      "step": 10473
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.578878643772507e-05,
+      "loss": 0.6608,
+      "step": 10474
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.577983298014984e-05,
+      "loss": 21.3192,
+      "step": 10475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.577088001865546e-05,
+      "loss": 16.5765,
+      "step": 10476
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.576192755355427e-05,
+      "loss": 17.7565,
+      "step": 10477
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5752975585158535e-05,
+      "loss": 19.0575,
+      "step": 10478
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.574402411378061e-05,
+      "loss": 17.2667,
+      "step": 10479
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.573507313973273e-05,
+      "loss": 0.7173,
+      "step": 10480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.57261226633272e-05,
+      "loss": 18.3575,
+      "step": 10481
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.571717268487623e-05,
+      "loss": 0.6295,
+      "step": 10482
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5708223204692045e-05,
+      "loss": 18.5639,
+      "step": 10483
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.569927422308688e-05,
+      "loss": 0.6948,
+      "step": 10484
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5690325740372886e-05,
+      "loss": 0.6841,
+      "step": 10485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.568137775686229e-05,
+      "loss": 0.7051,
+      "step": 10486
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.567243027286721e-05,
+      "loss": 18.6513,
+      "step": 10487
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.566348328869981e-05,
+      "loss": 17.4069,
+      "step": 10488
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.565453680467218e-05,
+      "loss": 19.0357,
+      "step": 10489
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5645590821096455e-05,
+      "loss": 18.5538,
+      "step": 10490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5636645338284705e-05,
+      "loss": 0.6354,
+      "step": 10491
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5627700356548996e-05,
+      "loss": 17.2221,
+      "step": 10492
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.561875587620141e-05,
+      "loss": 0.7133,
+      "step": 10493
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.560981189755394e-05,
+      "loss": 0.679,
+      "step": 10494
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.560086842091864e-05,
+      "loss": 0.6776,
+      "step": 10495
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.559192544660749e-05,
+      "loss": 18.5463,
+      "step": 10496
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.558298297493247e-05,
+      "loss": 16.6888,
+      "step": 10497
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.557404100620556e-05,
+      "loss": 19.0199,
+      "step": 10498
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.556509954073869e-05,
+      "loss": 0.6198,
+      "step": 10499
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.55561585788438e-05,
+      "loss": 17.8706,
+      "step": 10500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.55472181208328e-05,
+      "loss": 17.9653,
+      "step": 10501
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.553827816701758e-05,
+      "loss": 0.6403,
+      "step": 10502
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5529338717710026e-05,
+      "loss": 0.6148,
+      "step": 10503
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5520399773221985e-05,
+      "loss": 18.0788,
+      "step": 10504
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5511461333865314e-05,
+      "loss": 0.6244,
+      "step": 10505
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5502523399951825e-05,
+      "loss": 0.6799,
+      "step": 10506
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.549358597179332e-05,
+      "loss": 0.6791,
+      "step": 10507
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.548464904970161e-05,
+      "loss": 19.8222,
+      "step": 10508
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5475712633988436e-05,
+      "loss": 18.0165,
+      "step": 10509
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.546677672496558e-05,
+      "loss": 0.7025,
+      "step": 10510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.545784132294476e-05,
+      "loss": 18.3927,
+      "step": 10511
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.544890642823768e-05,
+      "loss": 0.6102,
+      "step": 10512
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.543997204115608e-05,
+      "loss": 0.7045,
+      "step": 10513
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.543103816201162e-05,
+      "loss": 0.7108,
+      "step": 10514
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.542210479111596e-05,
+      "loss": 0.6411,
+      "step": 10515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.541317192878075e-05,
+      "loss": 17.2577,
+      "step": 10516
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.540423957531762e-05,
+      "loss": 0.7024,
+      "step": 10517
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.53953077310382e-05,
+      "loss": 17.8407,
+      "step": 10518
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.538637639625406e-05,
+      "loss": 18.2127,
+      "step": 10519
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.537744557127677e-05,
+      "loss": 0.6795,
+      "step": 10520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.536851525641792e-05,
+      "loss": 0.6778,
+      "step": 10521
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.535958545198902e-05,
+      "loss": 0.6786,
+      "step": 10522
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.535065615830162e-05,
+      "loss": 18.8866,
+      "step": 10523
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.53417273756672e-05,
+      "loss": 0.6603,
+      "step": 10524
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.533279910439726e-05,
+      "loss": 19.7861,
+      "step": 10525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5323871344803263e-05,
+      "loss": 17.3227,
+      "step": 10526
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.531494409719667e-05,
+      "loss": 18.8401,
+      "step": 10527
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.53060173618889e-05,
+      "loss": 0.7247,
+      "step": 10528
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.529709113919139e-05,
+      "loss": 17.8726,
+      "step": 10529
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.528816542941551e-05,
+      "loss": 0.6962,
+      "step": 10530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.527924023287267e-05,
+      "loss": 0.652,
+      "step": 10531
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.52703155498742e-05,
+      "loss": 0.6618,
+      "step": 10532
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.526139138073148e-05,
+      "loss": 17.7884,
+      "step": 10533
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.525246772575581e-05,
+      "loss": 18.964,
+      "step": 10534
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5243544585258495e-05,
+      "loss": 18.2964,
+      "step": 10535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.523462195955085e-05,
+      "loss": 15.5663,
+      "step": 10536
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.522569984894412e-05,
+      "loss": 0.7176,
+      "step": 10537
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5216778253749574e-05,
+      "loss": 20.1207,
+      "step": 10538
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.520785717427846e-05,
+      "loss": 0.6883,
+      "step": 10539
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519893661084196e-05,
+      "loss": 0.7365,
+      "step": 10540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519001656375131e-05,
+      "loss": 0.6855,
+      "step": 10541
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.518109703331768e-05,
+      "loss": 19.5742,
+      "step": 10542
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.517217801985225e-05,
+      "loss": 0.7542,
+      "step": 10543
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.516325952366612e-05,
+      "loss": 16.2395,
+      "step": 10544
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5154341545070474e-05,
+      "loss": 18.1271,
+      "step": 10545
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.514542408437638e-05,
+      "loss": 0.674,
+      "step": 10546
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.513650714189495e-05,
+      "loss": 18.5972,
+      "step": 10547
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.512759071793725e-05,
+      "loss": 20.8376,
+      "step": 10548
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5118674812814336e-05,
+      "loss": 18.7013,
+      "step": 10549
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5109759426837265e-05,
+      "loss": 19.0918,
+      "step": 10550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5100844560317026e-05,
+      "loss": 18.8274,
+      "step": 10551
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5091930213564644e-05,
+      "loss": 17.4481,
+      "step": 10552
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.50830163868911e-05,
+      "loss": 18.5983,
+      "step": 10553
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.507410308060733e-05,
+      "loss": 0.7053,
+      "step": 10554
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.506519029502433e-05,
+      "loss": 0.6733,
+      "step": 10555
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5056278030452994e-05,
+      "loss": 0.7074,
+      "step": 10556
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5047366287204254e-05,
+      "loss": 18.5402,
+      "step": 10557
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5038455065588983e-05,
+      "loss": 0.635,
+      "step": 10558
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.502954436591806e-05,
+      "loss": 18.0256,
+      "step": 10559
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.502063418850235e-05,
+      "loss": 19.0009,
+      "step": 10560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.501172453365268e-05,
+      "loss": 18.4842,
+      "step": 10561
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.500281540167989e-05,
+      "loss": 0.731,
+      "step": 10562
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.499390679289476e-05,
+      "loss": 19.0711,
+      "step": 10563
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.498499870760808e-05,
+      "loss": 0.7115,
+      "step": 10564
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.497609114613062e-05,
+      "loss": 18.3554,
+      "step": 10565
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.496718410877311e-05,
+      "loss": 0.6601,
+      "step": 10566
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49582775958463e-05,
+      "loss": 15.9054,
+      "step": 10567
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4949371607660896e-05,
+      "loss": 0.6686,
+      "step": 10568
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.494046614452758e-05,
+      "loss": 19.1879,
+      "step": 10569
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.493156120675704e-05,
+      "loss": 16.5393,
+      "step": 10570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49226567946599e-05,
+      "loss": 19.4562,
+      "step": 10571
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.491375290854684e-05,
+      "loss": 17.2763,
+      "step": 10572
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.490484954872845e-05,
+      "loss": 18.2258,
+      "step": 10573
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4895946715515326e-05,
+      "loss": 0.688,
+      "step": 10574
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4887044409218077e-05,
+      "loss": 17.7326,
+      "step": 10575
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4878142630147236e-05,
+      "loss": 19.2933,
+      "step": 10576
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.486924137861338e-05,
+      "loss": 0.6598,
+      "step": 10577
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.486034065492702e-05,
+      "loss": 0.6195,
+      "step": 10578
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4851440459398645e-05,
+      "loss": 0.6594,
+      "step": 10579
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4842540792338776e-05,
+      "loss": 17.9906,
+      "step": 10580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.483364165405785e-05,
+      "loss": 0.7669,
+      "step": 10581
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4824743044866356e-05,
+      "loss": 18.8403,
+      "step": 10582
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.481584496507472e-05,
+      "loss": 0.6742,
+      "step": 10583
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.480694741499334e-05,
+      "loss": 0.7291,
+      "step": 10584
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4798050394932626e-05,
+      "loss": 19.8356,
+      "step": 10585
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4789153905202956e-05,
+      "loss": 18.0613,
+      "step": 10586
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.478025794611469e-05,
+      "loss": 16.423,
+      "step": 10587
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.477136251797818e-05,
+      "loss": 20.0707,
+      "step": 10588
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.476246762110372e-05,
+      "loss": 18.7747,
+      "step": 10589
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.475357325580165e-05,
+      "loss": 18.5639,
+      "step": 10590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.474467942238223e-05,
+      "loss": 0.6763,
+      "step": 10591
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4735786121155746e-05,
+      "loss": 17.5646,
+      "step": 10592
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.472689335243244e-05,
+      "loss": 17.4944,
+      "step": 10593
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.471800111652254e-05,
+      "loss": 16.9932,
+      "step": 10594
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4709109413736266e-05,
+      "loss": 17.7914,
+      "step": 10595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.470021824438379e-05,
+      "loss": 0.6788,
+      "step": 10596
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.469132760877534e-05,
+      "loss": 20.416,
+      "step": 10597
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4682437507221e-05,
+      "loss": 0.6677,
+      "step": 10598
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.467354794003096e-05,
+      "loss": 17.7853,
+      "step": 10599
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4664658907515306e-05,
+      "loss": 19.901,
+      "step": 10600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4655770409984166e-05,
+      "loss": 0.653,
+      "step": 10601
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4646882447747605e-05,
+      "loss": 17.1352,
+      "step": 10602
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.463799502111569e-05,
+      "loss": 0.6399,
+      "step": 10603
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.462910813039847e-05,
+      "loss": 0.7216,
+      "step": 10604
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4620221775905966e-05,
+      "loss": 17.3868,
+      "step": 10605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4611335957948186e-05,
+      "loss": 0.6578,
+      "step": 10606
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4602450676835117e-05,
+      "loss": 18.2193,
+      "step": 10607
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4593565932876726e-05,
+      "loss": 18.1544,
+      "step": 10608
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.458468172638298e-05,
+      "loss": 0.6894,
+      "step": 10609
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.457579805766379e-05,
+      "loss": 18.1313,
+      "step": 10610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.456691492702907e-05,
+      "loss": 17.0221,
+      "step": 10611
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.455803233478874e-05,
+      "loss": 18.4614,
+      "step": 10612
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4549150281252636e-05,
+      "loss": 0.7031,
+      "step": 10613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.454026876673065e-05,
+      "loss": 19.7349,
+      "step": 10614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.453138779153261e-05,
+      "loss": 0.7144,
+      "step": 10615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.452250735596831e-05,
+      "loss": 19.8124,
+      "step": 10616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4513627460347596e-05,
+      "loss": 16.9928,
+      "step": 10617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.450474810498021e-05,
+      "loss": 17.8629,
+      "step": 10618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.449586929017594e-05,
+      "loss": 16.1913,
+      "step": 10619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.448699101624453e-05,
+      "loss": 0.7226,
+      "step": 10620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.447811328349567e-05,
+      "loss": 18.8829,
+      "step": 10621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4469236092239113e-05,
+      "loss": 18.9569,
+      "step": 10622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4460359442784516e-05,
+      "loss": 17.5114,
+      "step": 10623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.445148333544156e-05,
+      "loss": 0.6793,
+      "step": 10624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.44426077705199e-05,
+      "loss": 20.2596,
+      "step": 10625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4433732748329136e-05,
+      "loss": 0.65,
+      "step": 10626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4424858269178914e-05,
+      "loss": 16.0849,
+      "step": 10627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4415984333378805e-05,
+      "loss": 18.7814,
+      "step": 10628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.44071109412384e-05,
+      "loss": 0.6635,
+      "step": 10629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4398238093067246e-05,
+      "loss": 18.9137,
+      "step": 10630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.438936578917487e-05,
+      "loss": 0.6937,
+      "step": 10631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4380494029870804e-05,
+      "loss": 18.2413,
+      "step": 10632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4371622815464526e-05,
+      "loss": 0.6973,
+      "step": 10633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4362752146265545e-05,
+      "loss": 17.9904,
+      "step": 10634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4353882022583295e-05,
+      "loss": 0.7074,
+      "step": 10635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.434501244472721e-05,
+      "loss": 16.5228,
+      "step": 10636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.433614341300675e-05,
+      "loss": 0.7188,
+      "step": 10637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4327274927731275e-05,
+      "loss": 0.7037,
+      "step": 10638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4318406989210195e-05,
+      "loss": 0.6416,
+      "step": 10639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.430953959775287e-05,
+      "loss": 18.0335,
+      "step": 10640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.430067275366863e-05,
+      "loss": 18.3293,
+      "step": 10641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4291806457266826e-05,
+      "loss": 15.3892,
+      "step": 10642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4282940708856747e-05,
+      "loss": 17.5463,
+      "step": 10643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4274075508747696e-05,
+      "loss": 0.6875,
+      "step": 10644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.426521085724892e-05,
+      "loss": 17.87,
+      "step": 10645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.425634675466968e-05,
+      "loss": 19.7384,
+      "step": 10646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.424748320131922e-05,
+      "loss": 18.1474,
+      "step": 10647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.423862019750673e-05,
+      "loss": 18.4272,
+      "step": 10648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4229757743541416e-05,
+      "loss": 17.4179,
+      "step": 10649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.422089583973245e-05,
+      "loss": 0.6647,
+      "step": 10650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4212034486388975e-05,
+      "loss": 18.0269,
+      "step": 10651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.420317368382015e-05,
+      "loss": 16.3128,
+      "step": 10652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.419431343233507e-05,
+      "loss": 17.3707,
+      "step": 10653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.418545373224283e-05,
+      "loss": 0.6445,
+      "step": 10654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.41765945838525e-05,
+      "loss": 19.4106,
+      "step": 10655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.416773598747316e-05,
+      "loss": 18.7693,
+      "step": 10656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.415887794341384e-05,
+      "loss": 18.9,
+      "step": 10657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.415002045198356e-05,
+      "loss": 18.9994,
+      "step": 10658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.414116351349132e-05,
+      "loss": 16.8231,
+      "step": 10659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.413230712824609e-05,
+      "loss": 0.6892,
+      "step": 10660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4123451296556845e-05,
+      "loss": 0.725,
+      "step": 10661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.411459601873252e-05,
+      "loss": 17.9104,
+      "step": 10662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.410574129508205e-05,
+      "loss": 0.6406,
+      "step": 10663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.409688712591432e-05,
+      "loss": 0.6801,
+      "step": 10664
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.408803351153822e-05,
+      "loss": 0.6634,
+      "step": 10665
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4079180452262626e-05,
+      "loss": 16.6754,
+      "step": 10666
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.407032794839636e-05,
+      "loss": 0.7196,
+      "step": 10667
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.406147600024828e-05,
+      "loss": 0.6996,
+      "step": 10668
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4052624608127175e-05,
+      "loss": 0.6607,
+      "step": 10669
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4043773772341826e-05,
+      "loss": 0.6348,
+      "step": 10670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.403492349320101e-05,
+      "loss": 0.6809,
+      "step": 10671
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4026073771013464e-05,
+      "loss": 18.2448,
+      "step": 10672
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4017224606087936e-05,
+      "loss": 18.0076,
+      "step": 10673
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4008375998733136e-05,
+      "loss": 17.7225,
+      "step": 10674
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.399952794925773e-05,
+      "loss": 17.3699,
+      "step": 10675
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3990680457970404e-05,
+      "loss": 18.4856,
+      "step": 10676
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3981833525179805e-05,
+      "loss": 17.2149,
+      "step": 10677
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.397298715119457e-05,
+      "loss": 18.5788,
+      "step": 10678
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.396414133632331e-05,
+      "loss": 18.1157,
+      "step": 10679
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.395529608087461e-05,
+      "loss": 0.6916,
+      "step": 10680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3946451385157066e-05,
+      "loss": 18.0957,
+      "step": 10681
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.393760724947919e-05,
+      "loss": 19.867,
+      "step": 10682
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3928763674149564e-05,
+      "loss": 17.8857,
+      "step": 10683
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3919920659476666e-05,
+      "loss": 0.6567,
+      "step": 10684
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.391107820576901e-05,
+      "loss": 15.0896,
+      "step": 10685
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.390223631333506e-05,
+      "loss": 0.727,
+      "step": 10686
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.389339498248328e-05,
+      "loss": 19.4109,
+      "step": 10687
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.388455421352211e-05,
+      "loss": 17.4174,
+      "step": 10688
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.387571400675996e-05,
+      "loss": 20.2086,
+      "step": 10689
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.386687436250522e-05,
+      "loss": 17.7623,
+      "step": 10690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.385803528106628e-05,
+      "loss": 18.061,
+      "step": 10691
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3849196762751485e-05,
+      "loss": 19.1937,
+      "step": 10692
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.384035880786919e-05,
+      "loss": 0.6622,
+      "step": 10693
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.383152141672771e-05,
+      "loss": 16.2989,
+      "step": 10694
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.382268458963532e-05,
+      "loss": 17.8833,
+      "step": 10695
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3813848326900334e-05,
+      "loss": 18.7023,
+      "step": 10696
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.380501262883099e-05,
+      "loss": 19.12,
+      "step": 10697
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.379617749573552e-05,
+      "loss": 0.6902,
+      "step": 10698
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.378734292792218e-05,
+      "loss": 0.7042,
+      "step": 10699
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3778508925699124e-05,
+      "loss": 17.3425,
+      "step": 10700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.376967548937457e-05,
+      "loss": 19.8709,
+      "step": 10701
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.376084261925666e-05,
+      "loss": 19.224,
+      "step": 10702
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3752010315653535e-05,
+      "loss": 0.5818,
+      "step": 10703
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.374317857887333e-05,
+      "loss": 0.6802,
+      "step": 10704
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.373434740922413e-05,
+      "loss": 0.7075,
+      "step": 10705
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.372551680701405e-05,
+      "loss": 17.5576,
+      "step": 10706
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3716686772551094e-05,
+      "loss": 17.9296,
+      "step": 10707
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.370785730614335e-05,
+      "loss": 18.534,
+      "step": 10708
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369902840809881e-05,
+      "loss": 19.7994,
+      "step": 10709
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369020007872551e-05,
+      "loss": 17.7579,
+      "step": 10710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3681372318331417e-05,
+      "loss": 16.4356,
+      "step": 10711
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3672545127224474e-05,
+      "loss": 17.7941,
+      "step": 10712
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.366371850571266e-05,
+      "loss": 18.1192,
+      "step": 10713
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3654892454103863e-05,
+      "loss": 20.0415,
+      "step": 10714
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.364606697270602e-05,
+      "loss": 17.1243,
+      "step": 10715
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3637242061826993e-05,
+      "loss": 0.7195,
+      "step": 10716
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.362841772177465e-05,
+      "loss": 18.2133,
+      "step": 10717
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.361959395285684e-05,
+      "loss": 17.4379,
+      "step": 10718
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3610770755381365e-05,
+      "loss": 0.6543,
+      "step": 10719
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.360194812965606e-05,
+      "loss": 0.7218,
+      "step": 10720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3593126075988694e-05,
+      "loss": 0.6839,
+      "step": 10721
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.358430459468702e-05,
+      "loss": 18.1543,
+      "step": 10722
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3575483686058805e-05,
+      "loss": 0.615,
+      "step": 10723
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3566663350411754e-05,
+      "loss": 0.6547,
+      "step": 10724
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3557843588053576e-05,
+      "loss": 0.7385,
+      "step": 10725
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.354902439929196e-05,
+      "loss": 0.669,
+      "step": 10726
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3540205784434564e-05,
+      "loss": 16.626,
+      "step": 10727
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3531387743789036e-05,
+      "loss": 17.5507,
+      "step": 10728
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3522570277662985e-05,
+      "loss": 18.2343,
+      "step": 10729
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3513753386364046e-05,
+      "loss": 0.6593,
+      "step": 10730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.350493707019977e-05,
+      "loss": 16.6554,
+      "step": 10731
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3496121329477734e-05,
+      "loss": 19.2596,
+      "step": 10732
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.348730616450549e-05,
+      "loss": 17.6039,
+      "step": 10733
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.347849157559054e-05,
+      "loss": 18.0775,
+      "step": 10734
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.346967756304041e-05,
+      "loss": 19.4554,
+      "step": 10735
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.346086412716258e-05,
+      "loss": 17.1647,
+      "step": 10736
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.345205126826449e-05,
+      "loss": 0.6956,
+      "step": 10737
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.344323898665361e-05,
+      "loss": 0.7008,
+      "step": 10738
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.343442728263735e-05,
+      "loss": 17.6869,
+      "step": 10739
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.342561615652312e-05,
+      "loss": 18.3656,
+      "step": 10740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.34168056086183e-05,
+      "loss": 18.1897,
+      "step": 10741
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.340799563923024e-05,
+      "loss": 16.7932,
+      "step": 10742
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.33991862486663e-05,
+      "loss": 17.5071,
+      "step": 10743
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.339037743723378e-05,
+      "loss": 16.7933,
+      "step": 10744
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3381569205240014e-05,
+      "loss": 17.0927,
+      "step": 10745
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.337276155299226e-05,
+      "loss": 0.729,
+      "step": 10746
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3363954480797777e-05,
+      "loss": 17.8166,
+      "step": 10747
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.335514798896383e-05,
+      "loss": 0.727,
+      "step": 10748
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.334634207779761e-05,
+      "loss": 16.7726,
+      "step": 10749
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.333753674760633e-05,
+      "loss": 0.657,
+      "step": 10750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.332873199869719e-05,
+      "loss": 0.6513,
+      "step": 10751
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.331992783137731e-05,
+      "loss": 18.016,
+      "step": 10752
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3311124245953864e-05,
+      "loss": 0.6408,
+      "step": 10753
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.330232124273395e-05,
+      "loss": 0.6626,
+      "step": 10754
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.329351882202469e-05,
+      "loss": 0.6897,
+      "step": 10755
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3284716984133145e-05,
+      "loss": 0.7139,
+      "step": 10756
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.327591572936637e-05,
+      "loss": 0.7054,
+      "step": 10757
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.326711505803142e-05,
+      "loss": 17.8288,
+      "step": 10758
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.325831497043529e-05,
+      "loss": 19.2557,
+      "step": 10759
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.324951546688501e-05,
+      "loss": 18.8716,
+      "step": 10760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.324071654768754e-05,
+      "loss": 17.9663,
+      "step": 10761
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.323191821314983e-05,
+      "loss": 0.6523,
+      "step": 10762
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3223120463578806e-05,
+      "loss": 0.657,
+      "step": 10763
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3214323299281414e-05,
+      "loss": 0.6885,
+      "step": 10764
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.320552672056453e-05,
+      "loss": 17.0617,
+      "step": 10765
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.319673072773503e-05,
+      "loss": 16.1984,
+      "step": 10766
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.318793532109977e-05,
+      "loss": 17.9489,
+      "step": 10767
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.317914050096558e-05,
+      "loss": 17.0143,
+      "step": 10768
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.317034626763929e-05,
+      "loss": 19.4566,
+      "step": 10769
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3161552621427693e-05,
+      "loss": 17.9255,
+      "step": 10770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3152759562637536e-05,
+      "loss": 0.6278,
+      "step": 10771
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.314396709157559e-05,
+      "loss": 18.4803,
+      "step": 10772
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3135175208548584e-05,
+      "loss": 18.0096,
+      "step": 10773
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.312638391386323e-05,
+      "loss": 0.6681,
+      "step": 10774
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3117593207826226e-05,
+      "loss": 18.7468,
+      "step": 10775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3108803090744224e-05,
+      "loss": 17.5726,
+      "step": 10776
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.310001356292388e-05,
+      "loss": 17.0158,
+      "step": 10777
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.309122462467183e-05,
+      "loss": 18.1916,
+      "step": 10778
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.308243627629468e-05,
+      "loss": 0.6526,
+      "step": 10779
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.307364851809903e-05,
+      "loss": 0.6648,
+      "step": 10780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.306486135039141e-05,
+      "loss": 18.8014,
+      "step": 10781
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.305607477347841e-05,
+      "loss": 0.6797,
+      "step": 10782
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.304728878766653e-05,
+      "loss": 17.1562,
+      "step": 10783
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.303850339326228e-05,
+      "loss": 0.6543,
+      "step": 10784
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3029718590572156e-05,
+      "loss": 0.6651,
+      "step": 10785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.30209343799026e-05,
+      "loss": 17.8386,
+      "step": 10786
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3012150761560085e-05,
+      "loss": 18.1548,
+      "step": 10787
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3003367735850996e-05,
+      "loss": 0.7007,
+      "step": 10788
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.299458530308178e-05,
+      "loss": 18.0047,
+      "step": 10789
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.29858034635588e-05,
+      "loss": 0.7141,
+      "step": 10790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.297702221758839e-05,
+      "loss": 17.3796,
+      "step": 10791
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.296824156547692e-05,
+      "loss": 17.4475,
+      "step": 10792
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2959461507530705e-05,
+      "loss": 18.5359,
+      "step": 10793
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.295068204405605e-05,
+      "loss": 19.9857,
+      "step": 10794
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.294190317535921e-05,
+      "loss": 17.4983,
+      "step": 10795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.293312490174646e-05,
+      "loss": 0.6827,
+      "step": 10796
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2924347223524034e-05,
+      "loss": 0.6718,
+      "step": 10797
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.291557014099815e-05,
+      "loss": 0.7274,
+      "step": 10798
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2906793654474985e-05,
+      "loss": 0.6568,
+      "step": 10799
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.289801776426075e-05,
+      "loss": 0.6842,
+      "step": 10800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.288924247066155e-05,
+      "loss": 0.7163,
+      "step": 10801
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2880467773983567e-05,
+      "loss": 0.6972,
+      "step": 10802
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.287169367453288e-05,
+      "loss": 0.6912,
+      "step": 10803
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.286292017261557e-05,
+      "loss": 0.6812,
+      "step": 10804
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.285414726853775e-05,
+      "loss": 16.7629,
+      "step": 10805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2845374962605434e-05,
+      "loss": 0.6888,
+      "step": 10806
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.283660325512466e-05,
+      "loss": 18.8917,
+      "step": 10807
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.282783214640145e-05,
+      "loss": 18.0946,
+      "step": 10808
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2819061636741764e-05,
+      "loss": 18.6534,
+      "step": 10809
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.281029172645158e-05,
+      "loss": 17.071,
+      "step": 10810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.280152241583684e-05,
+      "loss": 0.6878,
+      "step": 10811
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.279275370520348e-05,
+      "loss": 19.9457,
+      "step": 10812
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.278398559485739e-05,
+      "loss": 19.4382,
+      "step": 10813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2775218085104456e-05,
+      "loss": 0.6846,
+      "step": 10814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.276645117625054e-05,
+      "loss": 0.6924,
+      "step": 10815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.275768486860149e-05,
+      "loss": 16.8435,
+      "step": 10816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.274891916246311e-05,
+      "loss": 16.8265,
+      "step": 10817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.274015405814119e-05,
+      "loss": 17.7786,
+      "step": 10818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.273138955594153e-05,
+      "loss": 19.0053,
+      "step": 10819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.272262565616987e-05,
+      "loss": 18.7539,
+      "step": 10820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2713862359131956e-05,
+      "loss": 17.4999,
+      "step": 10821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2705099665133506e-05,
+      "loss": 0.5916,
+      "step": 10822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.269633757448018e-05,
+      "loss": 18.752,
+      "step": 10823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.26875760874777e-05,
+      "loss": 0.7544,
+      "step": 10824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.267881520443168e-05,
+      "loss": 15.9966,
+      "step": 10825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.267005492564776e-05,
+      "loss": 17.5963,
+      "step": 10826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.266129525143156e-05,
+      "loss": 0.6716,
+      "step": 10827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2652536182088636e-05,
+      "loss": 19.079,
+      "step": 10828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.264377771792459e-05,
+      "loss": 18.3784,
+      "step": 10829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2635019859244954e-05,
+      "loss": 18.0792,
+      "step": 10830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.262626260635525e-05,
+      "loss": 16.9662,
+      "step": 10831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.261750595956099e-05,
+      "loss": 20.2064,
+      "step": 10832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.260874991916763e-05,
+      "loss": 0.6967,
+      "step": 10833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.259999448548066e-05,
+      "loss": 17.0394,
+      "step": 10834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.25912396588055e-05,
+      "loss": 0.7189,
+      "step": 10835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2582485439447585e-05,
+      "loss": 0.6826,
+      "step": 10836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.257373182771231e-05,
+      "loss": 18.2701,
+      "step": 10837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2564978823905035e-05,
+      "loss": 18.964,
+      "step": 10838
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.255622642833114e-05,
+      "loss": 18.367,
+      "step": 10839
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.254747464129593e-05,
+      "loss": 18.4071,
+      "step": 10840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.253872346310474e-05,
+      "loss": 19.5564,
+      "step": 10841
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.252997289406285e-05,
+      "loss": 16.0213,
+      "step": 10842
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.252122293447552e-05,
+      "loss": 17.1878,
+      "step": 10843
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2512473584648036e-05,
+      "loss": 20.0783,
+      "step": 10844
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.250372484488558e-05,
+      "loss": 19.3673,
+      "step": 10845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2494976715493393e-05,
+      "loss": 0.6515,
+      "step": 10846
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.248622919677664e-05,
+      "loss": 0.6868,
+      "step": 10847
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.247748228904049e-05,
+      "loss": 0.7329,
+      "step": 10848
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2468735992590096e-05,
+      "loss": 0.6563,
+      "step": 10849
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.245999030773055e-05,
+      "loss": 18.1449,
+      "step": 10850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.245124523476699e-05,
+      "loss": 0.6934,
+      "step": 10851
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.244250077400446e-05,
+      "loss": 17.3617,
+      "step": 10852
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.243375692574804e-05,
+      "loss": 18.994,
+      "step": 10853
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.242501369030275e-05,
+      "loss": 18.5617,
+      "step": 10854
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2416271067973605e-05,
+      "loss": 0.6434,
+      "step": 10855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2407529059065624e-05,
+      "loss": 0.6274,
+      "step": 10856
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2398787663883746e-05,
+      "loss": 0.7045,
+      "step": 10857
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2390046882732925e-05,
+      "loss": 20.6911,
+      "step": 10858
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.238130671591811e-05,
+      "loss": 0.6648,
+      "step": 10859
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2372567163744186e-05,
+      "loss": 0.665,
+      "step": 10860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.236382822651606e-05,
+      "loss": 16.9644,
+      "step": 10861
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2355089904538584e-05,
+      "loss": 16.6064,
+      "step": 10862
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.234635219811658e-05,
+      "loss": 0.698,
+      "step": 10863
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.233761510755491e-05,
+      "loss": 18.4525,
+      "step": 10864
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.232887863315834e-05,
+      "loss": 18.2516,
+      "step": 10865
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2320142775231676e-05,
+      "loss": 0.6474,
+      "step": 10866
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.231140753407966e-05,
+      "loss": 0.6667,
+      "step": 10867
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2302672910007016e-05,
+      "loss": 0.7325,
+      "step": 10868
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.229393890331848e-05,
+      "loss": 18.7699,
+      "step": 10869
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.228520551431875e-05,
+      "loss": 0.6985,
+      "step": 10870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.227647274331246e-05,
+      "loss": 16.7239,
+      "step": 10871
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.226774059060428e-05,
+      "loss": 0.7375,
+      "step": 10872
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225900905649884e-05,
+      "loss": 0.6474,
+      "step": 10873
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225027814130074e-05,
+      "loss": 19.4305,
+      "step": 10874
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.224154784531457e-05,
+      "loss": 18.8534,
+      "step": 10875
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2232818168844895e-05,
+      "loss": 0.7158,
+      "step": 10876
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2224089112196234e-05,
+      "loss": 17.3991,
+      "step": 10877
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.221536067567314e-05,
+      "loss": 16.9044,
+      "step": 10878
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2206632859580085e-05,
+      "loss": 0.6959,
+      "step": 10879
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.219790566422156e-05,
+      "loss": 0.6691,
+      "step": 10880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.218917908990201e-05,
+      "loss": 18.1408,
+      "step": 10881
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.218045313692588e-05,
+      "loss": 0.6929,
+      "step": 10882
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.217172780559756e-05,
+      "loss": 0.6924,
+      "step": 10883
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2163003096221464e-05,
+      "loss": 17.5308,
+      "step": 10884
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2154279009101946e-05,
+      "loss": 18.2124,
+      "step": 10885
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.214555554454336e-05,
+      "loss": 19.5045,
+      "step": 10886
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.213683270285001e-05,
+      "loss": 0.6885,
+      "step": 10887
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2128110484326225e-05,
+      "loss": 0.6475,
+      "step": 10888
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.211938888927627e-05,
+      "loss": 18.7268,
+      "step": 10889
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.211066791800441e-05,
+      "loss": 17.7659,
+      "step": 10890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2101947570814894e-05,
+      "loss": 0.6959,
+      "step": 10891
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.20932278480119e-05,
+      "loss": 0.667,
+      "step": 10892
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2084508749899664e-05,
+      "loss": 17.0356,
+      "step": 10893
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.207579027678235e-05,
+      "loss": 0.6645,
+      "step": 10894
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2067072428964075e-05,
+      "loss": 19.5704,
+      "step": 10895
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.205835520674901e-05,
+      "loss": 17.2847,
+      "step": 10896
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2049638610441225e-05,
+      "loss": 17.1369,
+      "step": 10897
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.204092264034484e-05,
+      "loss": 18.5925,
+      "step": 10898
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.20322072967639e-05,
+      "loss": 19.0023,
+      "step": 10899
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.202349258000244e-05,
+      "loss": 0.6841,
+      "step": 10900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.201477849036448e-05,
+      "loss": 17.0757,
+      "step": 10901
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.200606502815403e-05,
+      "loss": 16.8077,
+      "step": 10902
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.199735219367507e-05,
+      "loss": 0.6469,
+      "step": 10903
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.198863998723153e-05,
+      "loss": 19.4491,
+      "step": 10904
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.197992840912736e-05,
+      "loss": 0.6652,
+      "step": 10905
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.197121745966646e-05,
+      "loss": 17.8981,
+      "step": 10906
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.196250713915272e-05,
+      "loss": 18.6603,
+      "step": 10907
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1953797447890013e-05,
+      "loss": 0.736,
+      "step": 10908
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1945088386182176e-05,
+      "loss": 0.7059,
+      "step": 10909
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1936379954333025e-05,
+      "loss": 20.3193,
+      "step": 10910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1927672152646375e-05,
+      "loss": 17.9302,
+      "step": 10911
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.191896498142599e-05,
+      "loss": 0.6963,
+      "step": 10912
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1910258440975636e-05,
+      "loss": 0.7032,
+      "step": 10913
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.190155253159905e-05,
+      "loss": 17.8795,
+      "step": 10914
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1892847253599926e-05,
+      "loss": 17.3024,
+      "step": 10915
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1884142607281976e-05,
+      "loss": 17.4551,
+      "step": 10916
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.187543859294885e-05,
+      "loss": 19.2989,
+      "step": 10917
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.186673521090421e-05,
+      "loss": 19.5719,
+      "step": 10918
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.185803246145166e-05,
+      "loss": 18.1075,
+      "step": 10919
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.184933034489481e-05,
+      "loss": 0.7408,
+      "step": 10920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.184062886153726e-05,
+      "loss": 0.7002,
+      "step": 10921
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.183192801168253e-05,
+      "loss": 0.6466,
+      "step": 10922
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.182322779563418e-05,
+      "loss": 17.9173,
+      "step": 10923
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.181452821369573e-05,
+      "loss": 0.6999,
+      "step": 10924
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.180582926617067e-05,
+      "loss": 0.6671,
+      "step": 10925
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.179713095336243e-05,
+      "loss": 18.429,
+      "step": 10926
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1788433275574504e-05,
+      "loss": 18.5463,
+      "step": 10927
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.177973623311029e-05,
+      "loss": 0.66,
+      "step": 10928
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.17710398262732e-05,
+      "loss": 17.1178,
+      "step": 10929
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.176234405536661e-05,
+      "loss": 16.7259,
+      "step": 10930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.175364892069388e-05,
+      "loss": 16.9232,
+      "step": 10931
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.174495442255836e-05,
+      "loss": 0.6835,
+      "step": 10932
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.173626056126335e-05,
+      "loss": 0.6778,
+      "step": 10933
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.172756733711213e-05,
+      "loss": 17.9218,
+      "step": 10934
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1718874750408004e-05,
+      "loss": 18.8623,
+      "step": 10935
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1710182801454176e-05,
+      "loss": 0.663,
+      "step": 10936
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1701491490553904e-05,
+      "loss": 17.3913,
+      "step": 10937
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.169280081801039e-05,
+      "loss": 18.8733,
+      "step": 10938
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.168411078412678e-05,
+      "loss": 18.1728,
+      "step": 10939
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1675421389206275e-05,
+      "loss": 19.1261,
+      "step": 10940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.166673263355199e-05,
+      "loss": 17.8275,
+      "step": 10941
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.165804451746703e-05,
+      "loss": 19.0661,
+      "step": 10942
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1649357041254514e-05,
+      "loss": 19.4578,
+      "step": 10943
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1640670205217484e-05,
+      "loss": 18.8409,
+      "step": 10944
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1631984009659e-05,
+      "loss": 18.6453,
+      "step": 10945
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.162329845488207e-05,
+      "loss": 19.2368,
+      "step": 10946
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.161461354118973e-05,
+      "loss": 17.0443,
+      "step": 10947
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1605929268884925e-05,
+      "loss": 0.691,
+      "step": 10948
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1597245638270625e-05,
+      "loss": 0.657,
+      "step": 10949
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1588562649649775e-05,
+      "loss": 17.5621,
+      "step": 10950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.157988030332526e-05,
+      "loss": 16.294,
+      "step": 10951
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.15711985996e-05,
+      "loss": 17.965,
+      "step": 10952
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1562517538776856e-05,
+      "loss": 17.8955,
+      "step": 10953
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.155383712115865e-05,
+      "loss": 0.7228,
+      "step": 10954
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1545157347048236e-05,
+      "loss": 17.5929,
+      "step": 10955
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.153647821674839e-05,
+      "loss": 18.1525,
+      "step": 10956
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.15277997305619e-05,
+      "loss": 0.6742,
+      "step": 10957
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1519121888791535e-05,
+      "loss": 17.6727,
+      "step": 10958
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1510444691739996e-05,
+      "loss": 17.3537,
+      "step": 10959
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.150176813971002e-05,
+      "loss": 0.6961,
+      "step": 10960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.149309223300428e-05,
+      "loss": 18.5183,
+      "step": 10961
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.148441697192546e-05,
+      "loss": 0.7195,
+      "step": 10962
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.147574235677618e-05,
+      "loss": 0.7096,
+      "step": 10963
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1467068387859065e-05,
+      "loss": 17.5219,
+      "step": 10964
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1458395065476734e-05,
+      "loss": 17.4799,
+      "step": 10965
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1449722389931724e-05,
+      "loss": 18.5174,
+      "step": 10966
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.144105036152663e-05,
+      "loss": 16.955,
+      "step": 10967
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.143237898056396e-05,
+      "loss": 16.2717,
+      "step": 10968
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.142370824734621e-05,
+      "loss": 0.7188,
+      "step": 10969
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.141503816217588e-05,
+      "loss": 0.7521,
+      "step": 10970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.140636872535544e-05,
+      "loss": 0.6476,
+      "step": 10971
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1397699937187316e-05,
+      "loss": 17.0344,
+      "step": 10972
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1389031797973936e-05,
+      "loss": 0.7162,
+      "step": 10973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.138036430801766e-05,
+      "loss": 17.9173,
+      "step": 10974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.137169746762091e-05,
+      "loss": 18.4429,
+      "step": 10975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.136303127708601e-05,
+      "loss": 17.225,
+      "step": 10976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1354365736715285e-05,
+      "loss": 0.6325,
+      "step": 10977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1345700846811045e-05,
+      "loss": 0.6478,
+      "step": 10978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.133703660767558e-05,
+      "loss": 17.8755,
+      "step": 10979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.132837301961111e-05,
+      "loss": 0.7417,
+      "step": 10980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131971008291991e-05,
+      "loss": 0.6336,
+      "step": 10981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131104779790418e-05,
+      "loss": 17.6965,
+      "step": 10982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.13023861648661e-05,
+      "loss": 16.6945,
+      "step": 10983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.129372518410786e-05,
+      "loss": 18.6425,
+      "step": 10984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1285064855931575e-05,
+      "loss": 16.7988,
+      "step": 10985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1276405180639396e-05,
+      "loss": 0.6328,
+      "step": 10986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.126774615853341e-05,
+      "loss": 17.3774,
+      "step": 10987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.125908778991568e-05,
+      "loss": 0.6967,
+      "step": 10988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1250430075088275e-05,
+      "loss": 20.0435,
+      "step": 10989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.124177301435324e-05,
+      "loss": 17.8846,
+      "step": 10990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.123311660801255e-05,
+      "loss": 0.6552,
+      "step": 10991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.12244608563682e-05,
+      "loss": 18.7131,
+      "step": 10992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.121580575972216e-05,
+      "loss": 0.6728,
+      "step": 10993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.120715131837638e-05,
+      "loss": 18.4936,
+      "step": 10994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1198497532632764e-05,
+      "loss": 17.9696,
+      "step": 10995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.118984440279319e-05,
+      "loss": 18.0121,
+      "step": 10996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1181191929159555e-05,
+      "loss": 17.0839,
+      "step": 10997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.117254011203368e-05,
+      "loss": 0.7021,
+      "step": 10998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1163888951717426e-05,
+      "loss": 0.6637,
+      "step": 10999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.115523844851257e-05,
+      "loss": 0.6891,
+      "step": 11000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.114658860272088e-05,
+      "loss": 17.4136,
+      "step": 11001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.113793941464415e-05,
+      "loss": 18.2395,
+      "step": 11002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.112929088458407e-05,
+      "loss": 17.829,
+      "step": 11003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.112064301284239e-05,
+      "loss": 0.6942,
+      "step": 11004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.111199579972077e-05,
+      "loss": 0.6084,
+      "step": 11005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1103349245520873e-05,
+      "loss": 0.6521,
+      "step": 11006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.109470335054436e-05,
+      "loss": 0.7038,
+      "step": 11007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1086058115092843e-05,
+      "loss": 0.7018,
+      "step": 11008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1077413539467915e-05,
+      "loss": 16.4926,
+      "step": 11009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.106876962397114e-05,
+      "loss": 0.6553,
+      "step": 11010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1060126368904074e-05,
+      "loss": 16.9711,
+      "step": 11011
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.105148377456825e-05,
+      "loss": 0.6515,
+      "step": 11012
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1042841841265153e-05,
+      "loss": 17.1196,
+      "step": 11013
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.103420056929629e-05,
+      "loss": 18.3134,
+      "step": 11014
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.102555995896309e-05,
+      "loss": 17.8286,
+      "step": 11015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1016920010567e-05,
+      "loss": 18.1751,
+      "step": 11016
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.100828072440944e-05,
+      "loss": 18.1388,
+      "step": 11017
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.099964210079178e-05,
+      "loss": 17.9757,
+      "step": 11018
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09910041400154e-05,
+      "loss": 0.6685,
+      "step": 11019
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0982366842381636e-05,
+      "loss": 0.625,
+      "step": 11020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09737302081918e-05,
+      "loss": 19.0269,
+      "step": 11021
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0965094237747196e-05,
+      "loss": 17.7573,
+      "step": 11022
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0956458931349086e-05,
+      "loss": 17.4936,
+      "step": 11023
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.094782428929873e-05,
+      "loss": 0.7078,
+      "step": 11024
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.093919031189735e-05,
+      "loss": 15.3629,
+      "step": 11025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.093055699944614e-05,
+      "loss": 18.4442,
+      "step": 11026
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09219243522463e-05,
+      "loss": 18.3962,
+      "step": 11027
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.091329237059896e-05,
+      "loss": 17.8108,
+      "step": 11028
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.090466105480527e-05,
+      "loss": 18.2124,
+      "step": 11029
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0896030405166344e-05,
+      "loss": 18.5584,
+      "step": 11030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.088740042198325e-05,
+      "loss": 16.1231,
+      "step": 11031
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.087877110555707e-05,
+      "loss": 18.2946,
+      "step": 11032
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.087014245618882e-05,
+      "loss": 0.6786,
+      "step": 11033
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0861514474179575e-05,
+      "loss": 18.942,
+      "step": 11034
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.085288715983025e-05,
+      "loss": 18.1655,
+      "step": 11035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.084426051344186e-05,
+      "loss": 19.0492,
+      "step": 11036
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.083563453531533e-05,
+      "loss": 20.3701,
+      "step": 11037
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.08270092257516e-05,
+      "loss": 0.6732,
+      "step": 11038
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.081838458505157e-05,
+      "loss": 17.2496,
+      "step": 11039
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0809760613516116e-05,
+      "loss": 0.6855,
+      "step": 11040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0801137311446087e-05,
+      "loss": 0.7111,
+      "step": 11041
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.079251467914231e-05,
+      "loss": 0.7006,
+      "step": 11042
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0783892716905607e-05,
+      "loss": 19.5249,
+      "step": 11043
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0775271425036746e-05,
+      "loss": 17.3261,
+      "step": 11044
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0766650803836484e-05,
+      "loss": 0.6936,
+      "step": 11045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0758030853605565e-05,
+      "loss": 0.7489,
+      "step": 11046
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.074941157464471e-05,
+      "loss": 0.6836,
+      "step": 11047
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0740792967254604e-05,
+      "loss": 0.6842,
+      "step": 11048
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.073217503173591e-05,
+      "loss": 0.6338,
+      "step": 11049
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0723557768389266e-05,
+      "loss": 16.9014,
+      "step": 11050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.07149411775153e-05,
+      "loss": 0.6739,
+      "step": 11051
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0706325259414604e-05,
+      "loss": 0.6314,
+      "step": 11052
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.069771001438776e-05,
+      "loss": 17.123,
+      "step": 11053
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06890954427353e-05,
+      "loss": 18.2524,
+      "step": 11054
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.068048154475776e-05,
+      "loss": 0.6161,
+      "step": 11055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.067186832075565e-05,
+      "loss": 0.6775,
+      "step": 11056
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0663255771029425e-05,
+      "loss": 19.123,
+      "step": 11057
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.065464389587956e-05,
+      "loss": 19.4241,
+      "step": 11058
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0646032695606494e-05,
+      "loss": 17.0654,
+      "step": 11059
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06374221705106e-05,
+      "loss": 0.7699,
+      "step": 11060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06288123208923e-05,
+      "loss": 18.5237,
+      "step": 11061
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.062020314705193e-05,
+      "loss": 15.6792,
+      "step": 11062
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.061159464928984e-05,
+      "loss": 0.6118,
+      "step": 11063
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.060298682790635e-05,
+      "loss": 0.6597,
+      "step": 11064
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.059437968320172e-05,
+      "loss": 19.2053,
+      "step": 11065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.058577321547625e-05,
+      "loss": 15.813,
+      "step": 11066
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.057716742503016e-05,
+      "loss": 17.0425,
+      "step": 11067
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.056856231216369e-05,
+      "loss": 17.2056,
+      "step": 11068
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.055995787717703e-05,
+      "loss": 0.6572,
+      "step": 11069
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.055135412037032e-05,
+      "loss": 17.6051,
+      "step": 11070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.054275104204376e-05,
+      "loss": 16.1228,
+      "step": 11071
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.053414864249743e-05,
+      "loss": 0.6856,
+      "step": 11072
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0525546922031464e-05,
+      "loss": 15.4087,
+      "step": 11073
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0516945880945923e-05,
+      "loss": 0.6896,
+      "step": 11074
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.050834551954086e-05,
+      "loss": 0.6719,
+      "step": 11075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.049974583811631e-05,
+      "loss": 0.6441,
+      "step": 11076
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0491146836972272e-05,
+      "loss": 17.7027,
+      "step": 11077
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0482548516408737e-05,
+      "loss": 17.1991,
+      "step": 11078
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0473950876725667e-05,
+      "loss": 18.1099,
+      "step": 11079
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0465353918222976e-05,
+      "loss": 18.9683,
+      "step": 11080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0456757641200606e-05,
+      "loss": 16.7243,
+      "step": 11081
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0448162045958416e-05,
+      "loss": 19.041,
+      "step": 11082
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0439567132796276e-05,
+      "loss": 19.5171,
+      "step": 11083
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0430972902014043e-05,
+      "loss": 18.4872,
+      "step": 11084
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0422379353911512e-05,
+      "loss": 0.7043,
+      "step": 11085
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0413786488788492e-05,
+      "loss": 0.6689,
+      "step": 11086
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.040519430694474e-05,
+      "loss": 17.2701,
+      "step": 11087
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.039660280868e-05,
+      "loss": 0.6738,
+      "step": 11088
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0388011994294014e-05,
+      "loss": 17.4735,
+      "step": 11089
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0379421864086448e-05,
+      "loss": 19.7797,
+      "step": 11090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0370832418356994e-05,
+      "loss": 17.0758,
+      "step": 11091
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0362243657405276e-05,
+      "loss": 20.2418,
+      "step": 11092
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.035365558153096e-05,
+      "loss": 17.8314,
+      "step": 11093
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0345068191033608e-05,
+      "loss": 16.1063,
+      "step": 11094
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0336481486212827e-05,
+      "loss": 18.4846,
+      "step": 11095
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0327895467368156e-05,
+      "loss": 0.6906,
+      "step": 11096
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0319310134799117e-05,
+      "loss": 16.6622,
+      "step": 11097
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.031072548880524e-05,
+      "loss": 0.6791,
+      "step": 11098
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.030214152968598e-05,
+      "loss": 17.5585,
+      "step": 11099
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.029355825774082e-05,
+      "loss": 18.1378,
+      "step": 11100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0284975673269173e-05,
+      "loss": 18.2931,
+      "step": 11101
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.027639377657045e-05,
+      "loss": 18.3645,
+      "step": 11102
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0267812567944053e-05,
+      "loss": 17.8806,
+      "step": 11103
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0259232047689325e-05,
+      "loss": 18.3699,
+      "step": 11104
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0250652216105613e-05,
+      "loss": 0.6617,
+      "step": 11105
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.024207307349224e-05,
+      "loss": 15.9713,
+      "step": 11106
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0233494620148472e-05,
+      "loss": 17.5872,
+      "step": 11107
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0224916856373596e-05,
+      "loss": 16.3814,
+      "step": 11108
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0216339782466842e-05,
+      "loss": 0.6709,
+      "step": 11109
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0207763398727436e-05,
+      "loss": 19.0456,
+      "step": 11110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0199187705454566e-05,
+      "loss": 18.4756,
+      "step": 11111
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.019061270294739e-05,
+      "loss": 16.8371,
+      "step": 11112
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0182038391505073e-05,
+      "loss": 0.689,
+      "step": 11113
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.017346477142673e-05,
+      "loss": 18.2132,
+      "step": 11114
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0164891843011457e-05,
+      "loss": 16.0798,
+      "step": 11115
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0156319606558325e-05,
+      "loss": 18.3231,
+      "step": 11116
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0147748062366373e-05,
+      "loss": 16.6309,
+      "step": 11117
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.013917721073465e-05,
+      "loss": 18.8455,
+      "step": 11118
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0130607051962122e-05,
+      "loss": 17.9488,
+      "step": 11119
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0122037586347806e-05,
+      "loss": 0.6616,
+      "step": 11120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0113468814190632e-05,
+      "loss": 0.6715,
+      "step": 11121
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0104900735789522e-05,
+      "loss": 17.0483,
+      "step": 11122
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0096333351443384e-05,
+      "loss": 15.1321,
+      "step": 11123
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.00877666614511e-05,
+      "loss": 16.5803,
+      "step": 11124
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.007920066611154e-05,
+      "loss": 17.9807,
+      "step": 11125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.007063536572351e-05,
+      "loss": 18.7007,
+      "step": 11126
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0062070760585824e-05,
+      "loss": 0.6657,
+      "step": 11127
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0053506850997278e-05,
+      "loss": 0.7226,
+      "step": 11128
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0044943637256607e-05,
+      "loss": 18.0055,
+      "step": 11129
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.003638111966257e-05,
+      "loss": 18.5892,
+      "step": 11130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.002781929851386e-05,
+      "loss": 19.5854,
+      "step": 11131
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.001925817410916e-05,
+      "loss": 17.7574,
+      "step": 11132
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0010697746747152e-05,
+      "loss": 18.2297,
+      "step": 11133
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0002138016726444e-05,
+      "loss": 17.3675,
+      "step": 11134
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9993578984345672e-05,
+      "loss": 17.5783,
+      "step": 11135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9985020649903416e-05,
+      "loss": 17.0564,
+      "step": 11136
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.997646301369823e-05,
+      "loss": 17.8628,
+      "step": 11137
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.996790607602867e-05,
+      "loss": 17.5175,
+      "step": 11138
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.995934983719323e-05,
+      "loss": 0.6282,
+      "step": 11139
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.995079429749043e-05,
+      "loss": 18.2325,
+      "step": 11140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9942239457218718e-05,
+      "loss": 18.7547,
+      "step": 11141
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.993368531667653e-05,
+      "loss": 17.0482,
+      "step": 11142
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9925131876162306e-05,
+      "loss": 16.1264,
+      "step": 11143
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9916579135974416e-05,
+      "loss": 0.6272,
+      "step": 11144
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.990802709641124e-05,
+      "loss": 17.5254,
+      "step": 11145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.98994757577711e-05,
+      "loss": 18.8866,
+      "step": 11146
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9890925120352354e-05,
+      "loss": 18.8188,
+      "step": 11147
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.988237518445326e-05,
+      "loss": 0.7413,
+      "step": 11148
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.987382595037212e-05,
+      "loss": 18.546,
+      "step": 11149
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9865277418407168e-05,
+      "loss": 0.7217,
+      "step": 11150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9856729588856615e-05,
+      "loss": 17.5052,
+      "step": 11151
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.984818246201867e-05,
+      "loss": 0.6944,
+      "step": 11152
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.98396360381915e-05,
+      "loss": 18.8604,
+      "step": 11153
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.983109031767326e-05,
+      "loss": 17.7463,
+      "step": 11154
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9822545300762078e-05,
+      "loss": 0.6709,
+      "step": 11155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.981400098775603e-05,
+      "loss": 0.7081,
+      "step": 11156
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9805457378953217e-05,
+      "loss": 16.5357,
+      "step": 11157
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.979691447465167e-05,
+      "loss": 0.7106,
+      "step": 11158
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9788372275149435e-05,
+      "loss": 0.6205,
+      "step": 11159
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.977983078074449e-05,
+      "loss": 0.7089,
+      "step": 11160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.977128999173482e-05,
+      "loss": 19.2386,
+      "step": 11161
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9762749908418387e-05,
+      "loss": 17.525,
+      "step": 11162
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.975421053109309e-05,
+      "loss": 17.6986,
+      "step": 11163
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9745671860056868e-05,
+      "loss": 18.0182,
+      "step": 11164
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9737133895607584e-05,
+      "loss": 18.3333,
+      "step": 11165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9728596638043073e-05,
+      "loss": 16.5417,
+      "step": 11166
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9720060087661182e-05,
+      "loss": 0.6907,
+      "step": 11167
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9711524244759715e-05,
+      "loss": 0.681,
+      "step": 11168
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9702989109636448e-05,
+      "loss": 0.6342,
+      "step": 11169
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9694454682589133e-05,
+      "loss": 17.9313,
+      "step": 11170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9685920963915492e-05,
+      "loss": 17.893,
+      "step": 11171
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.967738795391326e-05,
+      "loss": 19.1027,
+      "step": 11172
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.966885565288008e-05,
+      "loss": 16.8592,
+      "step": 11173
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.966032406111363e-05,
+      "loss": 17.2323,
+      "step": 11174
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9651793178911536e-05,
+      "loss": 0.6502,
+      "step": 11175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9643263006571397e-05,
+      "loss": 18.9076,
+      "step": 11176
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9634733544390802e-05,
+      "loss": 17.7547,
+      "step": 11177
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962620479266732e-05,
+      "loss": 17.1401,
+      "step": 11178
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.961767675169844e-05,
+      "loss": 0.6686,
+      "step": 11179
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9609149421781725e-05,
+      "loss": 0.6758,
+      "step": 11180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9600622803214607e-05,
+      "loss": 18.2579,
+      "step": 11181
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.959209689629458e-05,
+      "loss": 17.6625,
+      "step": 11182
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9583571701319062e-05,
+      "loss": 20.0114,
+      "step": 11183
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.957504721858545e-05,
+      "loss": 17.4819,
+      "step": 11184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9566523448391146e-05,
+      "loss": 18.671,
+      "step": 11185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9558000391033493e-05,
+      "loss": 18.0005,
+      "step": 11186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.954947804680984e-05,
+      "loss": 0.7001,
+      "step": 11187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9540956416017484e-05,
+      "loss": 18.5736,
+      "step": 11188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9532435498953697e-05,
+      "loss": 17.704,
+      "step": 11189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.952391529591576e-05,
+      "loss": 18.681,
+      "step": 11190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9515395807200896e-05,
+      "loss": 17.6987,
+      "step": 11191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9506877033106317e-05,
+      "loss": 18.264,
+      "step": 11192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9498358973929196e-05,
+      "loss": 0.6739,
+      "step": 11193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9489841629966703e-05,
+      "loss": 18.3637,
+      "step": 11194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.948132500151597e-05,
+      "loss": 18.3997,
+      "step": 11195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9472809088874097e-05,
+      "loss": 0.6516,
+      "step": 11196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9464293892338184e-05,
+      "loss": 0.6733,
+      "step": 11197
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9455779412205293e-05,
+      "loss": 15.3457,
+      "step": 11198
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.944726564877243e-05,
+      "loss": 17.3344,
+      "step": 11199
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9438752602336617e-05,
+      "loss": 18.1109,
+      "step": 11200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9430240273194842e-05,
+      "loss": 0.709,
+      "step": 11201
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9421728661644066e-05,
+      "loss": 18.3148,
+      "step": 11202
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9413217767981203e-05,
+      "loss": 18.4922,
+      "step": 11203
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9404707592503195e-05,
+      "loss": 0.7148,
+      "step": 11204
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.939619813550689e-05,
+      "loss": 0.671,
+      "step": 11205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9387689397289186e-05,
+      "loss": 17.2594,
+      "step": 11206
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9379181378146885e-05,
+      "loss": 17.3696,
+      "step": 11207
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9370674078376798e-05,
+      "loss": 0.6689,
+      "step": 11208
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9362167498275728e-05,
+      "loss": 18.6108,
+      "step": 11209
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9353661638140406e-05,
+      "loss": 15.464,
+      "step": 11210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.934515649826759e-05,
+      "loss": 0.6249,
+      "step": 11211
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9336652078953986e-05,
+      "loss": 17.7967,
+      "step": 11212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9328148380496252e-05,
+      "loss": 0.7516,
+      "step": 11213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9319645403191077e-05,
+      "loss": 0.6695,
+      "step": 11214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.931114314733506e-05,
+      "loss": 19.2615,
+      "step": 11215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9302641613224845e-05,
+      "loss": 16.5258,
+      "step": 11216
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9294140801157e-05,
+      "loss": 17.9423,
+      "step": 11217
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9285640711428057e-05,
+      "loss": 18.5839,
+      "step": 11218
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.927714134433459e-05,
+      "loss": 0.6932,
+      "step": 11219
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9268642700173065e-05,
+      "loss": 16.7085,
+      "step": 11220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9260144779240006e-05,
+      "loss": 17.4694,
+      "step": 11221
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9251647581831836e-05,
+      "loss": 0.6451,
+      "step": 11222
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9243151108244994e-05,
+      "loss": 0.6568,
+      "step": 11223
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9234655358775892e-05,
+      "loss": 16.2271,
+      "step": 11224
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9226160333720897e-05,
+      "loss": 18.52,
+      "step": 11225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9217666033376385e-05,
+      "loss": 17.3562,
+      "step": 11226
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9209172458038675e-05,
+      "loss": 19.7153,
+      "step": 11227
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9200679608004057e-05,
+      "loss": 17.1414,
+      "step": 11228
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9192187483568832e-05,
+      "loss": 17.3335,
+      "step": 11229
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918369608502925e-05,
+      "loss": 17.7036,
+      "step": 11230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.917520541268154e-05,
+      "loss": 19.7142,
+      "step": 11231
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.916671546682189e-05,
+      "loss": 0.7088,
+      "step": 11232
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.91582262477465e-05,
+      "loss": 0.7383,
+      "step": 11233
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9149737755751487e-05,
+      "loss": 0.7145,
+      "step": 11234
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9141249991133024e-05,
+      "loss": 18.7434,
+      "step": 11235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9132762954187192e-05,
+      "loss": 0.6927,
+      "step": 11236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9124276645210067e-05,
+      "loss": 0.6782,
+      "step": 11237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.91157910644977e-05,
+      "loss": 0.6871,
+      "step": 11238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9107306212346096e-05,
+      "loss": 16.6418,
+      "step": 11239
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.909882208905129e-05,
+      "loss": 18.7888,
+      "step": 11240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9090338694909254e-05,
+      "loss": 17.761,
+      "step": 11241
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9081856030215915e-05,
+      "loss": 18.5816,
+      "step": 11242
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9073374095267213e-05,
+      "loss": 0.6762,
+      "step": 11243
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9064892890359023e-05,
+      "loss": 17.8407,
+      "step": 11244
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9056412415787255e-05,
+      "loss": 0.7074,
+      "step": 11245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9047932671847734e-05,
+      "loss": 0.7059,
+      "step": 11246
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.903945365883629e-05,
+      "loss": 20.8081,
+      "step": 11247
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.903097537704872e-05,
+      "loss": 0.7324,
+      "step": 11248
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.902249782678076e-05,
+      "loss": 17.2834,
+      "step": 11249
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9014021008328212e-05,
+      "loss": 17.3506,
+      "step": 11250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.900554492198677e-05,
+      "loss": 0.6694,
+      "step": 11251
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.899706956805215e-05,
+      "loss": 16.4214,
+      "step": 11252
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8988594946819946e-05,
+      "loss": 18.3257,
+      "step": 11253
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8980121058585886e-05,
+      "loss": 18.0366,
+      "step": 11254
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.897164790364556e-05,
+      "loss": 19.0926,
+      "step": 11255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8963175482294545e-05,
+      "loss": 18.2926,
+      "step": 11256
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.895470379482843e-05,
+      "loss": 0.7472,
+      "step": 11257
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.894623284154272e-05,
+      "loss": 18.5961,
+      "step": 11258
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8937762622732973e-05,
+      "loss": 19.3483,
+      "step": 11259
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8929293138694657e-05,
+      "loss": 0.6546,
+      "step": 11260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.892082438972325e-05,
+      "loss": 0.6568,
+      "step": 11261
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8912356376114176e-05,
+      "loss": 18.2007,
+      "step": 11262
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8903889098162827e-05,
+      "loss": 0.6197,
+      "step": 11263
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8895422556164643e-05,
+      "loss": 16.3503,
+      "step": 11264
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.888695675041496e-05,
+      "loss": 0.7157,
+      "step": 11265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8878491681209107e-05,
+      "loss": 17.0349,
+      "step": 11266
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8870027348842398e-05,
+      "loss": 0.7392,
+      "step": 11267
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.88615637536101e-05,
+      "loss": 18.5699,
+      "step": 11268
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8853100895807506e-05,
+      "loss": 0.658,
+      "step": 11269
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8844638775729838e-05,
+      "loss": 20.8235,
+      "step": 11270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.883617739367229e-05,
+      "loss": 17.7443,
+      "step": 11271
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8827716749930068e-05,
+      "loss": 16.9991,
+      "step": 11272
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8819256844798274e-05,
+      "loss": 0.6547,
+      "step": 11273
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8810797678572098e-05,
+      "loss": 0.688,
+      "step": 11274
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8802339251546627e-05,
+      "loss": 0.7088,
+      "step": 11275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8793881564016932e-05,
+      "loss": 0.6646,
+      "step": 11276
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8785424616278068e-05,
+      "loss": 18.1861,
+      "step": 11277
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8776968408625036e-05,
+      "loss": 17.8018,
+      "step": 11278
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.876851294135289e-05,
+      "loss": 0.6443,
+      "step": 11279
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.876005821475657e-05,
+      "loss": 18.6934,
+      "step": 11280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.875160422913103e-05,
+      "loss": 17.8558,
+      "step": 11281
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.87431509847712e-05,
+      "loss": 17.2589,
+      "step": 11282
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8734698481971954e-05,
+      "loss": 18.3695,
+      "step": 11283
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.87262467210282e-05,
+      "loss": 19.012,
+      "step": 11284
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8717795702234767e-05,
+      "loss": 0.6928,
+      "step": 11285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8709345425886475e-05,
+      "loss": 18.1746,
+      "step": 11286
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8700895892278118e-05,
+      "loss": 17.7534,
+      "step": 11287
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8692447101704444e-05,
+      "loss": 0.7227,
+      "step": 11288
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8683999054460237e-05,
+      "loss": 0.6909,
+      "step": 11289
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.867555175084019e-05,
+      "loss": 18.6623,
+      "step": 11290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8667105191138987e-05,
+      "loss": 15.9188,
+      "step": 11291
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8658659375651308e-05,
+      "loss": 0.6781,
+      "step": 11292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865021430467176e-05,
+      "loss": 16.5918,
+      "step": 11293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8641769978495004e-05,
+      "loss": 17.6317,
+      "step": 11294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8633326397415594e-05,
+      "loss": 0.7053,
+      "step": 11295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.86248835617281e-05,
+      "loss": 18.8062,
+      "step": 11296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8616441471727063e-05,
+      "loss": 16.9559,
+      "step": 11297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8608000127706945e-05,
+      "loss": 17.2771,
+      "step": 11298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.85995595299623e-05,
+      "loss": 17.7338,
+      "step": 11299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.859111967878755e-05,
+      "loss": 17.0128,
+      "step": 11300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8582680574477117e-05,
+      "loss": 15.4259,
+      "step": 11301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8574242217325418e-05,
+      "loss": 17.195,
+      "step": 11302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8565804607626807e-05,
+      "loss": 18.6852,
+      "step": 11303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855736774567568e-05,
+      "loss": 15.8398,
+      "step": 11304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.854893163176633e-05,
+      "loss": 16.4299,
+      "step": 11305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.854049626619307e-05,
+      "loss": 17.1237,
+      "step": 11306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8532061649250175e-05,
+      "loss": 16.9204,
+      "step": 11307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8523627781231887e-05,
+      "loss": 0.6747,
+      "step": 11308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.851519466243242e-05,
+      "loss": 17.1384,
+      "step": 11309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8506762293145983e-05,
+      "loss": 17.0573,
+      "step": 11310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8498330673666746e-05,
+      "loss": 18.0028,
+      "step": 11311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8489899804288822e-05,
+      "loss": 18.0091,
+      "step": 11312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8481469685306372e-05,
+      "loss": 17.424,
+      "step": 11313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8473040317013467e-05,
+      "loss": 0.6626,
+      "step": 11314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.846461169970417e-05,
+      "loss": 0.696,
+      "step": 11315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8456183833672524e-05,
+      "loss": 0.7594,
+      "step": 11316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8447756719212516e-05,
+      "loss": 18.5028,
+      "step": 11317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8439330356618177e-05,
+      "loss": 0.6544,
+      "step": 11318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.843090474618344e-05,
+      "loss": 0.632,
+      "step": 11319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8422479888202245e-05,
+      "loss": 17.4657,
+      "step": 11320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.84140557829685e-05,
+      "loss": 16.0512,
+      "step": 11321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8405632430776063e-05,
+      "loss": 0.7027,
+      "step": 11322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8397209831918824e-05,
+      "loss": 18.2851,
+      "step": 11323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.83887879866906e-05,
+      "loss": 16.5394,
+      "step": 11324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.838036689538519e-05,
+      "loss": 19.7671,
+      "step": 11325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.837194655829637e-05,
+      "loss": 0.6895,
+      "step": 11326
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8363526975717887e-05,
+      "loss": 0.6656,
+      "step": 11327
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8355108147943443e-05,
+      "loss": 0.6733,
+      "step": 11328
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.834669007526678e-05,
+      "loss": 19.7469,
+      "step": 11329
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8338272757981554e-05,
+      "loss": 0.6281,
+      "step": 11330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8329856196381395e-05,
+      "loss": 0.646,
+      "step": 11331
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8321440390759923e-05,
+      "loss": 18.769,
+      "step": 11332
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8313025341410727e-05,
+      "loss": 17.5889,
+      "step": 11333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.830461104862739e-05,
+      "loss": 0.6982,
+      "step": 11334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8296197512703436e-05,
+      "loss": 18.1132,
+      "step": 11335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8287784733932392e-05,
+      "loss": 17.8783,
+      "step": 11336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8279372712607733e-05,
+      "loss": 0.6812,
+      "step": 11337
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8270961449022893e-05,
+      "loss": 0.711,
+      "step": 11338
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.826255094347135e-05,
+      "loss": 0.663,
+      "step": 11339
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.82541411962465e-05,
+      "loss": 18.6974,
+      "step": 11340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8245732207641708e-05,
+      "loss": 18.5776,
+      "step": 11341
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.823732397795033e-05,
+      "loss": 16.6909,
+      "step": 11342
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.822891650746568e-05,
+      "loss": 17.3062,
+      "step": 11343
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.82205097964811e-05,
+      "loss": 0.6575,
+      "step": 11344
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.821210384528984e-05,
+      "loss": 0.6783,
+      "step": 11345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.820369865418514e-05,
+      "loss": 16.9246,
+      "step": 11346
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.819529422346024e-05,
+      "loss": 17.6808,
+      "step": 11347
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8186890553408297e-05,
+      "loss": 16.3015,
+      "step": 11348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8178487644322526e-05,
+      "loss": 0.6645,
+      "step": 11349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.817008549649606e-05,
+      "loss": 18.2675,
+      "step": 11350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8161684110221986e-05,
+      "loss": 16.9075,
+      "step": 11351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8153283485793413e-05,
+      "loss": 16.7472,
+      "step": 11352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.814488362350338e-05,
+      "loss": 16.7833,
+      "step": 11353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8136484523644967e-05,
+      "loss": 16.7719,
+      "step": 11354
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8128086186511154e-05,
+      "loss": 0.7273,
+      "step": 11355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8119688612394922e-05,
+      "loss": 15.8797,
+      "step": 11356
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8111291801589235e-05,
+      "loss": 0.6928,
+      "step": 11357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8102895754387e-05,
+      "loss": 18.7371,
+      "step": 11358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8094500471081155e-05,
+      "loss": 0.6867,
+      "step": 11359
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8086105951964546e-05,
+      "loss": 0.6757,
+      "step": 11360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.807771219733004e-05,
+      "loss": 18.2823,
+      "step": 11361
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8069319207470455e-05,
+      "loss": 0.7047,
+      "step": 11362
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8060926982678588e-05,
+      "loss": 19.1049,
+      "step": 11363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8052535523247192e-05,
+      "loss": 17.4248,
+      "step": 11364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8044144829469033e-05,
+      "loss": 0.7155,
+      "step": 11365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8035754901636802e-05,
+      "loss": 19.7653,
+      "step": 11366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.802736574004319e-05,
+      "loss": 0.6727,
+      "step": 11367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8018977344980877e-05,
+      "loss": 18.7163,
+      "step": 11368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80105897167425e-05,
+      "loss": 17.3449,
+      "step": 11369
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.800220285562066e-05,
+      "loss": 15.9621,
+      "step": 11370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7993816761907925e-05,
+      "loss": 15.8098,
+      "step": 11371
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7985431435896846e-05,
+      "loss": 16.026,
+      "step": 11372
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7977046877879992e-05,
+      "loss": 17.291,
+      "step": 11373
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.796866308814984e-05,
+      "loss": 17.7986,
+      "step": 11374
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7960280066998852e-05,
+      "loss": 0.6509,
+      "step": 11375
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7951897814719495e-05,
+      "loss": 17.948,
+      "step": 11376
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7943516331604158e-05,
+      "loss": 17.4658,
+      "step": 11377
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7935135617945286e-05,
+      "loss": 0.6527,
+      "step": 11378
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.792675567403521e-05,
+      "loss": 17.1447,
+      "step": 11379
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.791837650016629e-05,
+      "loss": 17.7458,
+      "step": 11380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790999809663082e-05,
+      "loss": 16.424,
+      "step": 11381
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7901620463721084e-05,
+      "loss": 18.2444,
+      "step": 11382
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.789324360172937e-05,
+      "loss": 17.0926,
+      "step": 11383
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.78848675109479e-05,
+      "loss": 17.3511,
+      "step": 11384
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.787649219166887e-05,
+      "loss": 17.7238,
+      "step": 11385
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7868117644184465e-05,
+      "loss": 17.3309,
+      "step": 11386
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7859743868786815e-05,
+      "loss": 0.6658,
+      "step": 11387
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.785137086576809e-05,
+      "loss": 17.1417,
+      "step": 11388
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7842998635420358e-05,
+      "loss": 19.6613,
+      "step": 11389
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7834627178035704e-05,
+      "loss": 18.8796,
+      "step": 11390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7826256493906166e-05,
+      "loss": 16.9966,
+      "step": 11391
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.781788658332374e-05,
+      "loss": 0.6905,
+      "step": 11392
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7809517446580447e-05,
+      "loss": 18.3286,
+      "step": 11393
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7801149083968248e-05,
+      "loss": 0.6838,
+      "step": 11394
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7792781495779074e-05,
+      "loss": 0.6936,
+      "step": 11395
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7784414682304832e-05,
+      "loss": 16.3971,
+      "step": 11396
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.777604864383738e-05,
+      "loss": 18.5114,
+      "step": 11397
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.776768338066863e-05,
+      "loss": 18.8513,
+      "step": 11398
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.775931889309037e-05,
+      "loss": 17.115,
+      "step": 11399
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7750955181394412e-05,
+      "loss": 0.6606,
+      "step": 11400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7742592245872522e-05,
+      "loss": 16.9892,
+      "step": 11401
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7734230086816437e-05,
+      "loss": 19.441,
+      "step": 11402
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.772586870451791e-05,
+      "loss": 0.6828,
+      "step": 11403
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7717508099268618e-05,
+      "loss": 19.7893,
+      "step": 11404
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.770914827136023e-05,
+      "loss": 17.1768,
+      "step": 11405
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.770078922108438e-05,
+      "loss": 0.7203,
+      "step": 11406
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.769243094873265e-05,
+      "loss": 17.6591,
+      "step": 11407
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.768407345459668e-05,
+      "loss": 0.6454,
+      "step": 11408
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7675716738968007e-05,
+      "loss": 0.7206,
+      "step": 11409
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7667360802138154e-05,
+      "loss": 18.2036,
+      "step": 11410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7659005644398618e-05,
+      "loss": 19.5874,
+      "step": 11411
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.765065126604086e-05,
+      "loss": 0.6779,
+      "step": 11412
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.764229766735638e-05,
+      "loss": 18.6542,
+      "step": 11413
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7633944848636567e-05,
+      "loss": 0.7041,
+      "step": 11414
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7625592810172818e-05,
+      "loss": 17.6008,
+      "step": 11415
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7617241552256502e-05,
+      "loss": 0.7141,
+      "step": 11416
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7608891075178956e-05,
+      "loss": 17.1674,
+      "step": 11417
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7600541379231487e-05,
+      "loss": 0.7423,
+      "step": 11418
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.759219246470539e-05,
+      "loss": 17.2756,
+      "step": 11419
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7583844331891916e-05,
+      "loss": 0.6413,
+      "step": 11420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7575496981082287e-05,
+      "loss": 17.9642,
+      "step": 11421
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.756715041256773e-05,
+      "loss": 18.0364,
+      "step": 11422
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7558804626639422e-05,
+      "loss": 0.6672,
+      "step": 11423
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7550459623588486e-05,
+      "loss": 17.8333,
+      "step": 11424
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7542115403706063e-05,
+      "loss": 0.6904,
+      "step": 11425
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7533771967283224e-05,
+      "loss": 19.3315,
+      "step": 11426
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7525429314611068e-05,
+      "loss": 16.576,
+      "step": 11427
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7517087445980616e-05,
+      "loss": 16.6607,
+      "step": 11428
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.750874636168289e-05,
+      "loss": 18.0858,
+      "step": 11429
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7500406062008865e-05,
+      "loss": 0.6682,
+      "step": 11430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7492066547249486e-05,
+      "loss": 16.3244,
+      "step": 11431
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7483727817695713e-05,
+      "loss": 17.0058,
+      "step": 11432
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7475389873638436e-05,
+      "loss": 0.6538,
+      "step": 11433
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.746705271536853e-05,
+      "loss": 18.5427,
+      "step": 11434
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7458716343176837e-05,
+      "loss": 0.6575,
+      "step": 11435
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7450380757354167e-05,
+      "loss": 17.908,
+      "step": 11436
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7442045958191343e-05,
+      "loss": 19.4444,
+      "step": 11437
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7433711945979123e-05,
+      "loss": 0.6731,
+      "step": 11438
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.742537872100823e-05,
+      "loss": 0.6574,
+      "step": 11439
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.741704628356938e-05,
+      "loss": 0.6348,
+      "step": 11440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7408714633953247e-05,
+      "loss": 0.656,
+      "step": 11441
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.740038377245051e-05,
+      "loss": 0.6658,
+      "step": 11442
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7392053699351784e-05,
+      "loss": 18.4956,
+      "step": 11443
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7383724414947677e-05,
+      "loss": 0.6686,
+      "step": 11444
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7375395919528747e-05,
+      "loss": 18.8127,
+      "step": 11445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.736706821338554e-05,
+      "loss": 16.4328,
+      "step": 11446
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7358741296808586e-05,
+      "loss": 18.8201,
+      "step": 11447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7350415170088382e-05,
+      "loss": 0.6885,
+      "step": 11448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7342089833515382e-05,
+      "loss": 18.5292,
+      "step": 11449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.733376528738002e-05,
+      "loss": 0.6589,
+      "step": 11450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7325441531972683e-05,
+      "loss": 19.2109,
+      "step": 11451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7317118567583787e-05,
+      "loss": 0.7106,
+      "step": 11452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.730879639450368e-05,
+      "loss": 18.3821,
+      "step": 11453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7300475013022663e-05,
+      "loss": 0.6585,
+      "step": 11454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.729215442343106e-05,
+      "loss": 17.9695,
+      "step": 11455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.72838346260191e-05,
+      "loss": 17.0647,
+      "step": 11456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7275515621077074e-05,
+      "loss": 16.9884,
+      "step": 11457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7267197408895185e-05,
+      "loss": 18.5123,
+      "step": 11458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7258879989763596e-05,
+      "loss": 0.6465,
+      "step": 11459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7250563363972492e-05,
+      "loss": 0.6729,
+      "step": 11460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.724224753181197e-05,
+      "loss": 17.3923,
+      "step": 11461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7233932493572178e-05,
+      "loss": 0.6661,
+      "step": 11462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7225618249543172e-05,
+      "loss": 19.2406,
+      "step": 11463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7217304800014997e-05,
+      "loss": 0.6662,
+      "step": 11464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7208992145277678e-05,
+      "loss": 19.0816,
+      "step": 11465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7200680285621188e-05,
+      "loss": 0.6503,
+      "step": 11466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.719236922133553e-05,
+      "loss": 0.6825,
+      "step": 11467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7184058952710622e-05,
+      "loss": 18.6279,
+      "step": 11468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7175749480036372e-05,
+      "loss": 0.7361,
+      "step": 11469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7167440803602667e-05,
+      "loss": 18.1432,
+      "step": 11470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.715913292369936e-05,
+      "loss": 0.713,
+      "step": 11471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.715082584061627e-05,
+      "loss": 17.6639,
+      "step": 11472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7142519554643203e-05,
+      "loss": 0.6958,
+      "step": 11473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.713421406606993e-05,
+      "loss": 18.4276,
+      "step": 11474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.712590937518617e-05,
+      "loss": 18.3337,
+      "step": 11475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.711760548228168e-05,
+      "loss": 16.3893,
+      "step": 11476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.710930238764613e-05,
+      "loss": 17.418,
+      "step": 11477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7101000091569178e-05,
+      "loss": 18.1791,
+      "step": 11478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7092698594340454e-05,
+      "loss": 0.7192,
+      "step": 11479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7084397896249542e-05,
+      "loss": 16.0012,
+      "step": 11480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.707609799758606e-05,
+      "loss": 0.6714,
+      "step": 11481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.706779889863953e-05,
+      "loss": 17.8855,
+      "step": 11482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7059500599699476e-05,
+      "loss": 18.5576,
+      "step": 11483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7051203101055396e-05,
+      "loss": 0.7248,
+      "step": 11484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7042906402996726e-05,
+      "loss": 15.0456,
+      "step": 11485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7034610505812942e-05,
+      "loss": 18.438,
+      "step": 11486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.702631540979344e-05,
+      "loss": 19.8229,
+      "step": 11487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.70180211152276e-05,
+      "loss": 0.7249,
+      "step": 11488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7009727622404764e-05,
+      "loss": 0.7099,
+      "step": 11489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7001434931614244e-05,
+      "loss": 16.9038,
+      "step": 11490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6993143043145373e-05,
+      "loss": 18.8331,
+      "step": 11491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6984851957287403e-05,
+      "loss": 20.3094,
+      "step": 11492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6976561674329582e-05,
+      "loss": 16.3402,
+      "step": 11493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6968272194561107e-05,
+      "loss": 18.9795,
+      "step": 11494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.695998351827115e-05,
+      "loss": 17.3506,
+      "step": 11495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6951695645748908e-05,
+      "loss": 0.6445,
+      "step": 11496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6943408577283486e-05,
+      "loss": 16.0349,
+      "step": 11497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.693512231316399e-05,
+      "loss": 0.6897,
+      "step": 11498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.692683685367949e-05,
+      "loss": 0.6915,
+      "step": 11499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6918552199119006e-05,
+      "loss": 18.2712,
+      "step": 11500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6910268349771605e-05,
+      "loss": 19.343,
+      "step": 11501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.690198530592624e-05,
+      "loss": 18.243,
+      "step": 11502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6893703067871884e-05,
+      "loss": 17.7627,
+      "step": 11503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6885421635897462e-05,
+      "loss": 18.9763,
+      "step": 11504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6877141010291863e-05,
+      "loss": 0.6892,
+      "step": 11505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.686886119134399e-05,
+      "loss": 0.657,
+      "step": 11506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6860582179342685e-05,
+      "loss": 18.9188,
+      "step": 11507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6852303974576765e-05,
+      "loss": 0.7136,
+      "step": 11508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6844026577335014e-05,
+      "loss": 19.6288,
+      "step": 11509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.68357499879062e-05,
+      "loss": 0.6658,
+      "step": 11510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.682747420657904e-05,
+      "loss": 18.2756,
+      "step": 11511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6819199233642278e-05,
+      "loss": 18.607,
+      "step": 11512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6810925069384568e-05,
+      "loss": 0.6945,
+      "step": 11513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6802651714094568e-05,
+      "loss": 18.4165,
+      "step": 11514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.679437916806089e-05,
+      "loss": 0.7263,
+      "step": 11515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6786107431572123e-05,
+      "loss": 17.8399,
+      "step": 11516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.677783650491686e-05,
+      "loss": 18.2084,
+      "step": 11517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6769566388383616e-05,
+      "loss": 0.7154,
+      "step": 11518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6761297082260915e-05,
+      "loss": 0.6599,
+      "step": 11519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6753028586837226e-05,
+      "loss": 0.6509,
+      "step": 11520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.674476090240098e-05,
+      "loss": 17.9383,
+      "step": 11521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.673649402924065e-05,
+      "loss": 16.4874,
+      "step": 11522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6728227967644604e-05,
+      "loss": 0.6995,
+      "step": 11523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6719962717901214e-05,
+      "loss": 17.6938,
+      "step": 11524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6711698280298823e-05,
+      "loss": 17.4547,
+      "step": 11525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.670343465512573e-05,
+      "loss": 18.2935,
+      "step": 11526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6695171842670232e-05,
+      "loss": 17.6635,
+      "step": 11527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6686909843220565e-05,
+      "loss": 16.635,
+      "step": 11528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6678648657064976e-05,
+      "loss": 16.9965,
+      "step": 11529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.667038828449162e-05,
+      "loss": 0.654,
+      "step": 11530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6662128725788726e-05,
+      "loss": 0.6004,
+      "step": 11531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6653869981244407e-05,
+      "loss": 0.674,
+      "step": 11532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6645612051146774e-05,
+      "loss": 0.6583,
+      "step": 11533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.663735493578391e-05,
+      "loss": 18.8151,
+      "step": 11534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6629098635443856e-05,
+      "loss": 18.4166,
+      "step": 11535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6620843150414676e-05,
+      "loss": 16.9348,
+      "step": 11536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6612588480984346e-05,
+      "loss": 0.6333,
+      "step": 11537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6604334627440837e-05,
+      "loss": 17.3557,
+      "step": 11538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.659608159007209e-05,
+      "loss": 18.6111,
+      "step": 11539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6587829369166007e-05,
+      "loss": 17.3441,
+      "step": 11540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.65795779650105e-05,
+      "loss": 0.7482,
+      "step": 11541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6571327377893407e-05,
+      "loss": 0.7253,
+      "step": 11542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.656307760810257e-05,
+      "loss": 17.2792,
+      "step": 11543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6554828655925772e-05,
+      "loss": 17.7919,
+      "step": 11544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6546580521650773e-05,
+      "loss": 0.6693,
+      "step": 11545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6538333205565345e-05,
+      "loss": 0.6768,
+      "step": 11546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6530086707957192e-05,
+      "loss": 0.6587,
+      "step": 11547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6521841029114003e-05,
+      "loss": 17.8656,
+      "step": 11548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.651359616932342e-05,
+      "loss": 0.6656,
+      "step": 11549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.650535212887305e-05,
+      "loss": 17.4514,
+      "step": 11550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.649710890805055e-05,
+      "loss": 18.3737,
+      "step": 11551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.648886650714346e-05,
+      "loss": 18.9931,
+      "step": 11552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6480624926439322e-05,
+      "loss": 17.6066,
+      "step": 11553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.647238416622565e-05,
+      "loss": 0.6449,
+      "step": 11554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.646414422678991e-05,
+      "loss": 17.85,
+      "step": 11555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.64559051084196e-05,
+      "loss": 16.8052,
+      "step": 11556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6447666811402118e-05,
+      "loss": 0.6689,
+      "step": 11557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6439429336024878e-05,
+      "loss": 17.6272,
+      "step": 11558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.643119268257524e-05,
+      "loss": 17.097,
+      "step": 11559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.642295685134053e-05,
+      "loss": 0.6765,
+      "step": 11560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.641472184260809e-05,
+      "loss": 19.5752,
+      "step": 11561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6406487656665203e-05,
+      "loss": 18.6116,
+      "step": 11562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6398254293799112e-05,
+      "loss": 18.4852,
+      "step": 11563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6390021754297047e-05,
+      "loss": 0.6555,
+      "step": 11564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6381790038446187e-05,
+      "loss": 0.6573,
+      "step": 11565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6373559146533744e-05,
+      "loss": 17.7635,
+      "step": 11566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6365329078846823e-05,
+      "loss": 19.7872,
+      "step": 11567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.635709983567256e-05,
+      "loss": 16.5764,
+      "step": 11568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.634887141729802e-05,
+      "loss": 18.2665,
+      "step": 11569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6340643824010247e-05,
+      "loss": 15.7987,
+      "step": 11570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6332417056096297e-05,
+      "loss": 18.4712,
+      "step": 11571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6324191113843156e-05,
+      "loss": 0.6238,
+      "step": 11572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.631596599753779e-05,
+      "loss": 0.7006,
+      "step": 11573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6307741707467136e-05,
+      "loss": 0.663,
+      "step": 11574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6299518243918086e-05,
+      "loss": 17.9093,
+      "step": 11575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6291295607177557e-05,
+      "loss": 16.911,
+      "step": 11576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6283073797532388e-05,
+      "loss": 18.3403,
+      "step": 11577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6274852815269395e-05,
+      "loss": 0.6971,
+      "step": 11578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6266632660675383e-05,
+      "loss": 18.3956,
+      "step": 11579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6258413334037114e-05,
+      "loss": 0.6691,
+      "step": 11580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.625019483564132e-05,
+      "loss": 0.6516,
+      "step": 11581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6241977165774716e-05,
+      "loss": 18.0338,
+      "step": 11582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.623376032472398e-05,
+      "loss": 0.638,
+      "step": 11583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.622554431277574e-05,
+      "loss": 0.7042,
+      "step": 11584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6217329130216662e-05,
+      "loss": 17.3567,
+      "step": 11585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.620911477733331e-05,
+      "loss": 0.7292,
+      "step": 11586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.620090125441226e-05,
+      "loss": 0.6984,
+      "step": 11587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.619268856174003e-05,
+      "loss": 17.1231,
+      "step": 11588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6184476699603123e-05,
+      "loss": 0.6523,
+      "step": 11589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6176265668288046e-05,
+      "loss": 0.6309,
+      "step": 11590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6168055468081226e-05,
+      "loss": 0.6895,
+      "step": 11591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6159846099269092e-05,
+      "loss": 0.6783,
+      "step": 11592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.615163756213802e-05,
+      "loss": 0.6767,
+      "step": 11593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.614342985697436e-05,
+      "loss": 0.6656,
+      "step": 11594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6135222984064477e-05,
+      "loss": 17.5552,
+      "step": 11595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6127016943694666e-05,
+      "loss": 17.6742,
+      "step": 11596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.611881173615119e-05,
+      "loss": 0.6377,
+      "step": 11597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6110607361720297e-05,
+      "loss": 15.871,
+      "step": 11598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6102403820688177e-05,
+      "loss": 20.3459,
+      "step": 11599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6094201113341066e-05,
+      "loss": 17.8,
+      "step": 11600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6085999239965092e-05,
+      "loss": 0.7532,
+      "step": 11601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6077798200846393e-05,
+      "loss": 17.0795,
+      "step": 11602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6069597996271056e-05,
+      "loss": 0.6458,
+      "step": 11603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6061398626525147e-05,
+      "loss": 18.8803,
+      "step": 11604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6053200091894736e-05,
+      "loss": 0.692,
+      "step": 11605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6045002392665814e-05,
+      "loss": 16.3942,
+      "step": 11606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6036805529124363e-05,
+      "loss": 0.6485,
+      "step": 11607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.602860950155635e-05,
+      "loss": 17.9104,
+      "step": 11608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6020414310247664e-05,
+      "loss": 17.4521,
+      "step": 11609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.601221995548425e-05,
+      "loss": 0.6921,
+      "step": 11610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6004026437551943e-05,
+      "loss": 0.689,
+      "step": 11611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5995833756736592e-05,
+      "loss": 0.6715,
+      "step": 11612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5987641913324003e-05,
+      "loss": 0.7032,
+      "step": 11613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5979450907599924e-05,
+      "loss": 0.7167,
+      "step": 11614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5971260739850144e-05,
+      "loss": 17.0813,
+      "step": 11615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5963071410360384e-05,
+      "loss": 0.7226,
+      "step": 11616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5954882919416312e-05,
+      "loss": 19.5056,
+      "step": 11617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.59466952673036e-05,
+      "loss": 0.7235,
+      "step": 11618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5938508454307864e-05,
+      "loss": 0.6398,
+      "step": 11619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.593032248071473e-05,
+      "loss": 18.4145,
+      "step": 11620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.592213734680977e-05,
+      "loss": 0.6777,
+      "step": 11621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.591395305287852e-05,
+      "loss": 0.6757,
+      "step": 11622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5905769599206496e-05,
+      "loss": 0.6946,
+      "step": 11623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5897586986079165e-05,
+      "loss": 0.6884,
+      "step": 11624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5889405213782016e-05,
+      "loss": 18.8333,
+      "step": 11625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.588122428260047e-05,
+      "loss": 17.9525,
+      "step": 11626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5873044192819913e-05,
+      "loss": 0.6907,
+      "step": 11627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.586486494472572e-05,
+      "loss": 0.6307,
+      "step": 11628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5856686538603204e-05,
+      "loss": 18.4808,
+      "step": 11629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5848508974737716e-05,
+      "loss": 17.9866,
+      "step": 11630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5840332253414522e-05,
+      "loss": 17.9332,
+      "step": 11631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.583215637491887e-05,
+      "loss": 16.9541,
+      "step": 11632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5823981339535973e-05,
+      "loss": 16.8892,
+      "step": 11633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.581580714755102e-05,
+      "loss": 0.6654,
+      "step": 11634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.580763379924922e-05,
+      "loss": 16.8825,
+      "step": 11635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.579946129491565e-05,
+      "loss": 18.2195,
+      "step": 11636
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5791289634835437e-05,
+      "loss": 17.9982,
+      "step": 11637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5783118819293628e-05,
+      "loss": 18.5253,
+      "step": 11638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.577494884857532e-05,
+      "loss": 18.0544,
+      "step": 11639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5766779722965495e-05,
+      "loss": 18.4153,
+      "step": 11640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.575861144274914e-05,
+      "loss": 17.9767,
+      "step": 11641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5750444008211222e-05,
+      "loss": 16.76,
+      "step": 11642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.574227741963664e-05,
+      "loss": 17.3415,
+      "step": 11643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5734111677310335e-05,
+      "loss": 0.6389,
+      "step": 11644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.572594678151715e-05,
+      "loss": 15.8984,
+      "step": 11645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.571778273254193e-05,
+      "loss": 17.9083,
+      "step": 11646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5709619530669476e-05,
+      "loss": 17.4864,
+      "step": 11647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.570145717618455e-05,
+      "loss": 0.6805,
+      "step": 11648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5693295669371954e-05,
+      "loss": 19.5467,
+      "step": 11649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5685135010516366e-05,
+      "loss": 18.5122,
+      "step": 11650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5676975199902486e-05,
+      "loss": 0.7201,
+      "step": 11651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5668816237814984e-05,
+      "loss": 0.6679,
+      "step": 11652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5660658124538463e-05,
+      "loss": 17.2684,
+      "step": 11653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.565250086035756e-05,
+      "loss": 17.9599,
+      "step": 11654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5644344445556835e-05,
+      "loss": 0.7084,
+      "step": 11655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5636188880420832e-05,
+      "loss": 0.6952,
+      "step": 11656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.562803416523405e-05,
+      "loss": 18.04,
+      "step": 11657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5619880300280973e-05,
+      "loss": 18.1722,
+      "step": 11658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5611727285846075e-05,
+      "loss": 17.7702,
+      "step": 11659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5603575122213774e-05,
+      "loss": 19.2141,
+      "step": 11660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5595423809668452e-05,
+      "loss": 0.726,
+      "step": 11661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.558727334849449e-05,
+      "loss": 17.8779,
+      "step": 11662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5579123738976185e-05,
+      "loss": 0.6196,
+      "step": 11663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5570974981397887e-05,
+      "loss": 16.1815,
+      "step": 11664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.556282707604386e-05,
+      "loss": 0.7226,
+      "step": 11665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.555468002319834e-05,
+      "loss": 0.7555,
+      "step": 11666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5546533823145548e-05,
+      "loss": 0.7114,
+      "step": 11667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.553838847616965e-05,
+      "loss": 16.8521,
+      "step": 11668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5530243982554837e-05,
+      "loss": 18.084,
+      "step": 11669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.552210034258522e-05,
+      "loss": 18.6748,
+      "step": 11670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5513957556544898e-05,
+      "loss": 16.3468,
+      "step": 11671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5505815624717932e-05,
+      "loss": 0.6789,
+      "step": 11672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.549767454738835e-05,
+      "loss": 0.6703,
+      "step": 11673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5489534324840185e-05,
+      "loss": 17.9981,
+      "step": 11674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5481394957357406e-05,
+      "loss": 17.0564,
+      "step": 11675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5473256445223948e-05,
+      "loss": 0.6827,
+      "step": 11676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.546511878872374e-05,
+      "loss": 17.772,
+      "step": 11677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.545698198814065e-05,
+      "loss": 16.5687,
+      "step": 11678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5448846043758578e-05,
+      "loss": 0.6842,
+      "step": 11679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.544071095586132e-05,
+      "loss": 0.6659,
+      "step": 11680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.543257672473268e-05,
+      "loss": 0.6347,
+      "step": 11681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5424443350656435e-05,
+      "loss": 18.3318,
+      "step": 11682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.541631083391629e-05,
+      "loss": 18.678,
+      "step": 11683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.540817917479601e-05,
+      "loss": 18.4135,
+      "step": 11684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.540004837357924e-05,
+      "loss": 0.6924,
+      "step": 11685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.539191843054963e-05,
+      "loss": 19.4309,
+      "step": 11686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.538378934599081e-05,
+      "loss": 16.0956,
+      "step": 11687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.537566112018634e-05,
+      "loss": 16.1808,
+      "step": 11688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5367533753419836e-05,
+      "loss": 0.6692,
+      "step": 11689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5359407245974775e-05,
+      "loss": 0.6629,
+      "step": 11690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5351281598134674e-05,
+      "loss": 18.1294,
+      "step": 11691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5343156810182977e-05,
+      "loss": 18.243,
+      "step": 11692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5335032882403174e-05,
+      "loss": 17.7136,
+      "step": 11693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.532690981507864e-05,
+      "loss": 19.6094,
+      "step": 11694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5318787608492765e-05,
+      "loss": 0.6699,
+      "step": 11695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5310666262928895e-05,
+      "loss": 16.8116,
+      "step": 11696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5302545778670328e-05,
+      "loss": 17.8097,
+      "step": 11697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5294426156000396e-05,
+      "loss": 0.6925,
+      "step": 11698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.528630739520233e-05,
+      "loss": 18.4359,
+      "step": 11699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.527818949655937e-05,
+      "loss": 0.6452,
+      "step": 11700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.52700724603547e-05,
+      "loss": 0.6385,
+      "step": 11701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5261956286871485e-05,
+      "loss": 17.064,
+      "step": 11702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5253840976392894e-05,
+      "loss": 0.7248,
+      "step": 11703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5245726529202017e-05,
+      "loss": 0.6841,
+      "step": 11704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.523761294558194e-05,
+      "loss": 18.6037,
+      "step": 11705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.52295002258157e-05,
+      "loss": 19.6728,
+      "step": 11706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5221388370186316e-05,
+      "loss": 0.686,
+      "step": 11707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5213277378976762e-05,
+      "loss": 18.4291,
+      "step": 11708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5205167252470035e-05,
+      "loss": 0.7001,
+      "step": 11709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.519705799094904e-05,
+      "loss": 0.7084,
+      "step": 11710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5188949594696675e-05,
+      "loss": 16.9055,
+      "step": 11711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5180842063995812e-05,
+      "loss": 17.3557,
+      "step": 11712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.517273539912926e-05,
+      "loss": 17.431,
+      "step": 11713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.516462960037987e-05,
+      "loss": 0.7013,
+      "step": 11714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51565246680304e-05,
+      "loss": 0.662,
+      "step": 11715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5148420602363594e-05,
+      "loss": 0.6771,
+      "step": 11716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5140317403662167e-05,
+      "loss": 15.6717,
+      "step": 11717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5132215072208787e-05,
+      "loss": 0.7081,
+      "step": 11718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5124113608286148e-05,
+      "loss": 0.5904,
+      "step": 11719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5116013012176865e-05,
+      "loss": 16.6354,
+      "step": 11720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.510791328416352e-05,
+      "loss": 17.1421,
+      "step": 11721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5099814424528683e-05,
+      "loss": 0.6547,
+      "step": 11722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5091716433554867e-05,
+      "loss": 0.6811,
+      "step": 11723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5083619311524627e-05,
+      "loss": 19.9686,
+      "step": 11724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.50755230587204e-05,
+      "loss": 15.6872,
+      "step": 11725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5067427675424642e-05,
+      "loss": 0.6057,
+      "step": 11726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5059333161919763e-05,
+      "loss": 0.7113,
+      "step": 11727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5051239518488123e-05,
+      "loss": 18.9818,
+      "step": 11728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5043146745412118e-05,
+      "loss": 0.6525,
+      "step": 11729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5035054842974044e-05,
+      "loss": 0.7031,
+      "step": 11730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.50269638114562e-05,
+      "loss": 17.4559,
+      "step": 11731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5018873651140845e-05,
+      "loss": 17.4406,
+      "step": 11732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5010784362310192e-05,
+      "loss": 16.6914,
+      "step": 11733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.500269594524648e-05,
+      "loss": 17.3377,
+      "step": 11734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.499460840023186e-05,
+      "loss": 0.7189,
+      "step": 11735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4986521727548473e-05,
+      "loss": 17.361,
+      "step": 11736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.497843592747842e-05,
+      "loss": 17.6489,
+      "step": 11737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4970351000303776e-05,
+      "loss": 18.3645,
+      "step": 11738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.496226694630662e-05,
+      "loss": 17.5567,
+      "step": 11739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4954183765768944e-05,
+      "loss": 17.7206,
+      "step": 11740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4946101458972742e-05,
+      "loss": 0.7079,
+      "step": 11741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4938020026199975e-05,
+      "loss": 0.654,
+      "step": 11742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.492993946773255e-05,
+      "loss": 15.8084,
+      "step": 11743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4921859783852408e-05,
+      "loss": 18.0189,
+      "step": 11744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4913780974841373e-05,
+      "loss": 0.6392,
+      "step": 11745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4905703040981293e-05,
+      "loss": 0.6751,
+      "step": 11746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4897625982553952e-05,
+      "loss": 0.6777,
+      "step": 11747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.488954979984116e-05,
+      "loss": 0.6877,
+      "step": 11748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4881474493124646e-05,
+      "loss": 0.6144,
+      "step": 11749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4873400062686125e-05,
+      "loss": 0.7045,
+      "step": 11750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4865326508807276e-05,
+      "loss": 17.7916,
+      "step": 11751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4857253831769727e-05,
+      "loss": 0.6918,
+      "step": 11752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.484918203185515e-05,
+      "loss": 18.6981,
+      "step": 11753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4841111109345105e-05,
+      "loss": 0.6542,
+      "step": 11754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4833041064521156e-05,
+      "loss": 0.6018,
+      "step": 11755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4824971897664827e-05,
+      "loss": 0.6583,
+      "step": 11756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4816903609057612e-05,
+      "loss": 17.8111,
+      "step": 11757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4808836198981e-05,
+      "loss": 0.6683,
+      "step": 11758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.480076966771642e-05,
+      "loss": 16.4041,
+      "step": 11759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4792704015545282e-05,
+      "loss": 18.2663,
+      "step": 11760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4784639242748953e-05,
+      "loss": 0.6753,
+      "step": 11761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.477657534960876e-05,
+      "loss": 0.6386,
+      "step": 11762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4768512336406058e-05,
+      "loss": 18.3193,
+      "step": 11763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4760450203422116e-05,
+      "loss": 18.1887,
+      "step": 11764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.475238895093818e-05,
+      "loss": 16.4676,
+      "step": 11765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.474432857923548e-05,
+      "loss": 0.6709,
+      "step": 11766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.473626908859519e-05,
+      "loss": 0.6777,
+      "step": 11767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4728210479298498e-05,
+      "loss": 17.7266,
+      "step": 11768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4720152751626523e-05,
+      "loss": 16.1833,
+      "step": 11769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.471209590586036e-05,
+      "loss": 0.6612,
+      "step": 11770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4704039942281087e-05,
+      "loss": 16.6673,
+      "step": 11771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.469598486116972e-05,
+      "loss": 0.7133,
+      "step": 11772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4687930662807303e-05,
+      "loss": 16.5254,
+      "step": 11773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4679877347474783e-05,
+      "loss": 17.7622,
+      "step": 11774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4671824915453123e-05,
+      "loss": 17.6139,
+      "step": 11775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.466377336702323e-05,
+      "loss": 0.6645,
+      "step": 11776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4655722702465974e-05,
+      "loss": 17.4066,
+      "step": 11777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4647672922062233e-05,
+      "loss": 0.6482,
+      "step": 11778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4639624026092822e-05,
+      "loss": 0.7466,
+      "step": 11779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.463157601483853e-05,
+      "loss": 0.6119,
+      "step": 11780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4623528888580122e-05,
+      "loss": 17.1575,
+      "step": 11781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4615482647598297e-05,
+      "loss": 0.6683,
+      "step": 11782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.460743729217381e-05,
+      "loss": 0.6391,
+      "step": 11783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4599392822587287e-05,
+      "loss": 0.6525,
+      "step": 11784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4591349239119383e-05,
+      "loss": 19.2394,
+      "step": 11785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.45833065420507e-05,
+      "loss": 17.8559,
+      "step": 11786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4575264731661795e-05,
+      "loss": 16.5618,
+      "step": 11787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.456722380823324e-05,
+      "loss": 18.995,
+      "step": 11788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4559183772045548e-05,
+      "loss": 20.197,
+      "step": 11789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4551144623379186e-05,
+      "loss": 17.526,
+      "step": 11790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4543106362514613e-05,
+      "loss": 19.7076,
+      "step": 11791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.453506898973223e-05,
+      "loss": 18.1394,
+      "step": 11792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4527032505312458e-05,
+      "loss": 16.875,
+      "step": 11793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4518996909535647e-05,
+      "loss": 0.7326,
+      "step": 11794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.451096220268212e-05,
+      "loss": 17.0984,
+      "step": 11795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.450292838503217e-05,
+      "loss": 16.8424,
+      "step": 11796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4494895456866046e-05,
+      "loss": 18.0709,
+      "step": 11797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4486863418464023e-05,
+      "loss": 0.6859,
+      "step": 11798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.44788322701063e-05,
+      "loss": 0.6457,
+      "step": 11799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.447080201207302e-05,
+      "loss": 0.6407,
+      "step": 11800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.446277264464431e-05,
+      "loss": 16.8513,
+      "step": 11801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.445474416810033e-05,
+      "loss": 0.6799,
+      "step": 11802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4446716582721135e-05,
+      "loss": 19.104,
+      "step": 11803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.443868988878678e-05,
+      "loss": 0.7172,
+      "step": 11804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4430664086577275e-05,
+      "loss": 16.8003,
+      "step": 11805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4422639176372587e-05,
+      "loss": 0.6647,
+      "step": 11806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4414615158452715e-05,
+      "loss": 17.2178,
+      "step": 11807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4406592033097554e-05,
+      "loss": 17.736,
+      "step": 11808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4398569800587006e-05,
+      "loss": 17.1496,
+      "step": 11809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.439054846120093e-05,
+      "loss": 0.5837,
+      "step": 11810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.438252801521913e-05,
+      "loss": 0.7107,
+      "step": 11811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4374508462921453e-05,
+      "loss": 15.5151,
+      "step": 11812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4366489804587643e-05,
+      "loss": 17.9403,
+      "step": 11813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.435847204049744e-05,
+      "loss": 17.7087,
+      "step": 11814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.435045517093054e-05,
+      "loss": 17.7324,
+      "step": 11815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4342439196166613e-05,
+      "loss": 18.1458,
+      "step": 11816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.433442411648532e-05,
+      "loss": 0.7143,
+      "step": 11817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4326409932166277e-05,
+      "loss": 17.276,
+      "step": 11818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4318396643489056e-05,
+      "loss": 16.6034,
+      "step": 11819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4310384250733197e-05,
+      "loss": 17.1211,
+      "step": 11820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.430237275417821e-05,
+      "loss": 18.5641,
+      "step": 11821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4294362154103613e-05,
+      "loss": 16.867,
+      "step": 11822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4286352450788848e-05,
+      "loss": 0.6819,
+      "step": 11823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.427834364451334e-05,
+      "loss": 19.173,
+      "step": 11824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4270335735556475e-05,
+      "loss": 18.5718,
+      "step": 11825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.42623287241976e-05,
+      "loss": 0.6238,
+      "step": 11826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4254322610716084e-05,
+      "loss": 0.6574,
+      "step": 11827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.424631739539121e-05,
+      "loss": 17.0734,
+      "step": 11828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4238313078502235e-05,
+      "loss": 17.8012,
+      "step": 11829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.423030966032841e-05,
+      "loss": 18.2445,
+      "step": 11830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.422230714114891e-05,
+      "loss": 17.5458,
+      "step": 11831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4214305521242947e-05,
+      "loss": 18.1714,
+      "step": 11832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.420630480088965e-05,
+      "loss": 17.6153,
+      "step": 11833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.419830498036813e-05,
+      "loss": 0.6466,
+      "step": 11834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4190306059957462e-05,
+      "loss": 18.8726,
+      "step": 11835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.418230803993668e-05,
+      "loss": 17.7373,
+      "step": 11836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.417431092058483e-05,
+      "loss": 17.2822,
+      "step": 11837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4166314702180885e-05,
+      "loss": 0.6796,
+      "step": 11838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4158319385003804e-05,
+      "loss": 21.1928,
+      "step": 11839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.41503249693325e-05,
+      "loss": 0.6949,
+      "step": 11840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.414233145544585e-05,
+      "loss": 16.4477,
+      "step": 11841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.413433884362275e-05,
+      "loss": 0.6538,
+      "step": 11842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4126347134142007e-05,
+      "loss": 19.1464,
+      "step": 11843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.411835632728242e-05,
+      "loss": 0.653,
+      "step": 11844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4110366423322756e-05,
+      "loss": 0.7033,
+      "step": 11845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4102377422541724e-05,
+      "loss": 0.7003,
+      "step": 11846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.409438932521807e-05,
+      "loss": 16.9862,
+      "step": 11847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4086402131630444e-05,
+      "loss": 18.3062,
+      "step": 11848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4078415842057476e-05,
+      "loss": 18.505,
+      "step": 11849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4070430456777792e-05,
+      "loss": 17.3918,
+      "step": 11850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4062445976069935e-05,
+      "loss": 17.1363,
+      "step": 11851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.405446240021249e-05,
+      "loss": 0.6343,
+      "step": 11852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.404647972948397e-05,
+      "loss": 0.6982,
+      "step": 11853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4038497964162827e-05,
+      "loss": 18.4633,
+      "step": 11854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4030517104527494e-05,
+      "loss": 0.7101,
+      "step": 11855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.402253715085644e-05,
+      "loss": 16.3726,
+      "step": 11856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4014558103428026e-05,
+      "loss": 16.6943,
+      "step": 11857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4006579962520608e-05,
+      "loss": 17.2052,
+      "step": 11858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.399860272841251e-05,
+      "loss": 0.6247,
+      "step": 11859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.399062640138201e-05,
+      "loss": 17.1952,
+      "step": 11860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39826509817074e-05,
+      "loss": 0.6899,
+      "step": 11861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3974676469666885e-05,
+      "loss": 17.2209,
+      "step": 11862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.396670286553867e-05,
+      "loss": 0.6994,
+      "step": 11863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.395873016960092e-05,
+      "loss": 16.9329,
+      "step": 11864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3950758382131744e-05,
+      "loss": 0.6831,
+      "step": 11865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.394278750340928e-05,
+      "loss": 16.4305,
+      "step": 11866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3934817533711584e-05,
+      "loss": 0.6372,
+      "step": 11867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39268484733167e-05,
+      "loss": 17.8825,
+      "step": 11868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.391888032250262e-05,
+      "loss": 0.6955,
+      "step": 11869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3910913081547308e-05,
+      "loss": 0.6799,
+      "step": 11870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3902946750728744e-05,
+      "loss": 19.1786,
+      "step": 11871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3894981330324823e-05,
+      "loss": 17.5576,
+      "step": 11872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3887016820613418e-05,
+      "loss": 0.6557,
+      "step": 11873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3879053221872384e-05,
+      "loss": 17.329,
+      "step": 11874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.387109053437951e-05,
+      "loss": 18.2878,
+      "step": 11875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3863128758412622e-05,
+      "loss": 0.6756,
+      "step": 11876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.385516789424946e-05,
+      "loss": 0.607,
+      "step": 11877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3847207942167737e-05,
+      "loss": 0.6124,
+      "step": 11878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.383924890244514e-05,
+      "loss": 17.3623,
+      "step": 11879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3831290775359312e-05,
+      "loss": 0.7003,
+      "step": 11880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.382333356118791e-05,
+      "loss": 18.4372,
+      "step": 11881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3815377260208514e-05,
+      "loss": 18.4451,
+      "step": 11882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.380742187269869e-05,
+      "loss": 0.6965,
+      "step": 11883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3799467398935948e-05,
+      "loss": 18.2572,
+      "step": 11884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3791513839197787e-05,
+      "loss": 17.6005,
+      "step": 11885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3783561193761704e-05,
+      "loss": 19.3957,
+      "step": 11886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3775609462905113e-05,
+      "loss": 17.8701,
+      "step": 11887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.376765864690541e-05,
+      "loss": 17.212,
+      "step": 11888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3759708746039976e-05,
+      "loss": 0.6965,
+      "step": 11889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.375175976058613e-05,
+      "loss": 19.5489,
+      "step": 11890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3743811690821205e-05,
+      "loss": 0.7313,
+      "step": 11891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3735864537022462e-05,
+      "loss": 17.2588,
+      "step": 11892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3727918299467146e-05,
+      "loss": 17.2098,
+      "step": 11893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.371997297843246e-05,
+      "loss": 17.2581,
+      "step": 11894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3712028574195587e-05,
+      "loss": 18.5339,
+      "step": 11895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3704085087033663e-05,
+      "loss": 0.6488,
+      "step": 11896
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3696142517223823e-05,
+      "loss": 17.9605,
+      "step": 11897
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3688200865043142e-05,
+      "loss": 18.9183,
+      "step": 11898
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3680260130768668e-05,
+      "loss": 0.6401,
+      "step": 11899
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3672320314677414e-05,
+      "loss": 20.1334,
+      "step": 11900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.366438141704636e-05,
+      "loss": 0.7005,
+      "step": 11901
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.365644343815248e-05,
+      "loss": 18.5558,
+      "step": 11902
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3648506378272695e-05,
+      "loss": 18.2236,
+      "step": 11903
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3640570237683886e-05,
+      "loss": 18.2446,
+      "step": 11904
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3632635016662914e-05,
+      "loss": 15.102,
+      "step": 11905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.362470071548658e-05,
+      "loss": 18.472,
+      "step": 11906
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3616767334431722e-05,
+      "loss": 0.6995,
+      "step": 11907
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.36088348737751e-05,
+      "loss": 16.8029,
+      "step": 11908
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3600903333793402e-05,
+      "loss": 0.6832,
+      "step": 11909
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3592972714763334e-05,
+      "loss": 17.2744,
+      "step": 11910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.358504301696159e-05,
+      "loss": 17.3335,
+      "step": 11911
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3577114240664788e-05,
+      "loss": 18.5275,
+      "step": 11912
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3569186386149533e-05,
+      "loss": 17.4446,
+      "step": 11913
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.356125945369239e-05,
+      "loss": 0.633,
+      "step": 11914
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.355333344356988e-05,
+      "loss": 17.4283,
+      "step": 11915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.354540835605854e-05,
+      "loss": 19.3902,
+      "step": 11916
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.353748419143483e-05,
+      "loss": 0.698,
+      "step": 11917
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3529560949975182e-05,
+      "loss": 0.6838,
+      "step": 11918
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.352163863195601e-05,
+      "loss": 17.9399,
+      "step": 11919
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3513717237653672e-05,
+      "loss": 18.3306,
+      "step": 11920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3505796767344545e-05,
+      "loss": 17.2942,
+      "step": 11921
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3497877221304926e-05,
+      "loss": 16.2494,
+      "step": 11922
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.348995859981109e-05,
+      "loss": 18.5118,
+      "step": 11923
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3482040903139284e-05,
+      "loss": 0.7098,
+      "step": 11924
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3474124131565706e-05,
+      "loss": 0.7683,
+      "step": 11925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3466208285366577e-05,
+      "loss": 17.5254,
+      "step": 11926
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3458293364818025e-05,
+      "loss": 18.0376,
+      "step": 11927
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3450379370196167e-05,
+      "loss": 19.8023,
+      "step": 11928
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.34424663017771e-05,
+      "loss": 17.5456,
+      "step": 11929
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.343455415983684e-05,
+      "loss": 0.7052,
+      "step": 11930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3426642944651457e-05,
+      "loss": 16.5698,
+      "step": 11931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3418732656496915e-05,
+      "loss": 19.431,
+      "step": 11932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3410823295649182e-05,
+      "loss": 0.6767,
+      "step": 11933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3402914862384163e-05,
+      "loss": 0.6654,
+      "step": 11934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3395007356977744e-05,
+      "loss": 17.7241,
+      "step": 11935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3387100779705824e-05,
+      "loss": 0.7059,
+      "step": 11936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.33791951308442e-05,
+      "loss": 19.3141,
+      "step": 11937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.337129041066867e-05,
+      "loss": 0.6748,
+      "step": 11938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3363386619455004e-05,
+      "loss": 18.6692,
+      "step": 11939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3355483757478895e-05,
+      "loss": 15.9567,
+      "step": 11940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3347581825016102e-05,
+      "loss": 17.9788,
+      "step": 11941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3339680822342247e-05,
+      "loss": 0.6793,
+      "step": 11942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.333178074973298e-05,
+      "loss": 0.6687,
+      "step": 11943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3323881607463888e-05,
+      "loss": 0.6421,
+      "step": 11944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3315983395810533e-05,
+      "loss": 16.1605,
+      "step": 11945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3308086115048473e-05,
+      "loss": 0.6768,
+      "step": 11946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3300189765453196e-05,
+      "loss": 19.2225,
+      "step": 11947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3292294347300182e-05,
+      "loss": 0.6272,
+      "step": 11948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3284399860864854e-05,
+      "loss": 0.6634,
+      "step": 11949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3276506306422608e-05,
+      "loss": 16.6279,
+      "step": 11950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3268613684248846e-05,
+      "loss": 17.7756,
+      "step": 11951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3260721994618894e-05,
+      "loss": 17.5757,
+      "step": 11952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3252831237808064e-05,
+      "loss": 18.9484,
+      "step": 11953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3244941414091627e-05,
+      "loss": 16.742,
+      "step": 11954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3237052523744795e-05,
+      "loss": 19.7044,
+      "step": 11955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.322916456704283e-05,
+      "loss": 16.0452,
+      "step": 11956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.322127754426089e-05,
+      "loss": 0.6471,
+      "step": 11957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.321339145567411e-05,
+      "loss": 18.2375,
+      "step": 11958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.320550630155761e-05,
+      "loss": 18.3015,
+      "step": 11959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3197622082186448e-05,
+      "loss": 0.6721,
+      "step": 11960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3189738797835708e-05,
+      "loss": 19.9388,
+      "step": 11961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3181856448780402e-05,
+      "loss": 18.4388,
+      "step": 11962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3173975035295477e-05,
+      "loss": 18.452,
+      "step": 11963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3166094557655883e-05,
+      "loss": 0.6663,
+      "step": 11964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3158215016136563e-05,
+      "loss": 18.7744,
+      "step": 11965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3150336411012396e-05,
+      "loss": 16.5578,
+      "step": 11966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3142458742558232e-05,
+      "loss": 16.9073,
+      "step": 11967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.313458201104888e-05,
+      "loss": 18.2135,
+      "step": 11968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.312670621675911e-05,
+      "loss": 0.6813,
+      "step": 11969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.311883135996371e-05,
+      "loss": 0.7009,
+      "step": 11970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3110957440937392e-05,
+      "loss": 17.86,
+      "step": 11971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3103084459954827e-05,
+      "loss": 0.6946,
+      "step": 11972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3095212417290685e-05,
+      "loss": 18.5491,
+      "step": 11973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3087341313219564e-05,
+      "loss": 19.4507,
+      "step": 11974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3079471148016086e-05,
+      "loss": 16.4732,
+      "step": 11975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3071601921954794e-05,
+      "loss": 0.7429,
+      "step": 11976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.306373363531021e-05,
+      "loss": 0.6585,
+      "step": 11977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.305586628835682e-05,
+      "loss": 0.6988,
+      "step": 11978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3047999881369075e-05,
+      "loss": 16.7268,
+      "step": 11979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3040134414621428e-05,
+      "loss": 18.2001,
+      "step": 11980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.303226988838825e-05,
+      "loss": 18.9418,
+      "step": 11981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3024406302943914e-05,
+      "loss": 0.6536,
+      "step": 11982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.301654365856274e-05,
+      "loss": 16.772,
+      "step": 11983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3008681955518996e-05,
+      "loss": 18.355,
+      "step": 11984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.300082119408699e-05,
+      "loss": 0.6672,
+      "step": 11985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2992961374540932e-05,
+      "loss": 18.3029,
+      "step": 11986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2985102497155014e-05,
+      "loss": 17.5502,
+      "step": 11987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2977244562203392e-05,
+      "loss": 0.6456,
+      "step": 11988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2969387569960195e-05,
+      "loss": 18.938,
+      "step": 11989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2961531520699537e-05,
+      "loss": 17.7437,
+      "step": 11990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.295367641469548e-05,
+      "loss": 18.3866,
+      "step": 11991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2945822252222044e-05,
+      "loss": 16.962,
+      "step": 11992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.293796903355323e-05,
+      "loss": 16.515,
+      "step": 11993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2930116758962984e-05,
+      "loss": 18.3077,
+      "step": 11994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2922265428725282e-05,
+      "loss": 16.1666,
+      "step": 11995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2914415043113997e-05,
+      "loss": 0.6949,
+      "step": 11996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2906565602402998e-05,
+      "loss": 18.2573,
+      "step": 11997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.289871710686612e-05,
+      "loss": 18.3792,
+      "step": 11998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2890869556777138e-05,
+      "loss": 19.4598,
+      "step": 11999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2883022952409864e-05,
+      "loss": 0.6579,
+      "step": 12000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.287517729403802e-05,
+      "loss": 0.6609,
+      "step": 12001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2867332581935292e-05,
+      "loss": 0.6153,
+      "step": 12002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2859488816375356e-05,
+      "loss": 17.1416,
+      "step": 12003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.285164599763183e-05,
+      "loss": 0.6685,
+      "step": 12004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2843804125978357e-05,
+      "loss": 17.8406,
+      "step": 12005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2835963201688477e-05,
+      "loss": 0.6727,
+      "step": 12006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2828123225035735e-05,
+      "loss": 0.5751,
+      "step": 12007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2820284196293633e-05,
+      "loss": 18.0515,
+      "step": 12008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.281244611573562e-05,
+      "loss": 18.4837,
+      "step": 12009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2804608983635174e-05,
+      "loss": 0.7223,
+      "step": 12010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2796772800265675e-05,
+      "loss": 16.5977,
+      "step": 12011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.27889375659005e-05,
+      "loss": 18.4274,
+      "step": 12012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2781103280812983e-05,
+      "loss": 18.841,
+      "step": 12013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.277326994527641e-05,
+      "loss": 18.0285,
+      "step": 12014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.276543755956409e-05,
+      "loss": 17.2641,
+      "step": 12015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.275760612394925e-05,
+      "loss": 18.6821,
+      "step": 12016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.27497756387051e-05,
+      "loss": 16.7781,
+      "step": 12017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2741946104104766e-05,
+      "loss": 17.1007,
+      "step": 12018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2734117520421444e-05,
+      "loss": 0.6691,
+      "step": 12019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2726289887928216e-05,
+      "loss": 17.0207,
+      "step": 12020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2718463206898156e-05,
+      "loss": 18.3367,
+      "step": 12021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2710637477604308e-05,
+      "loss": 17.1373,
+      "step": 12022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2702812700319653e-05,
+      "loss": 0.6856,
+      "step": 12023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2694988875317203e-05,
+      "loss": 0.6617,
+      "step": 12024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2687166002869882e-05,
+      "loss": 0.7016,
+      "step": 12025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.267934408325059e-05,
+      "loss": 15.9767,
+      "step": 12026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2671523116732214e-05,
+      "loss": 18.2434,
+      "step": 12027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2663703103587558e-05,
+      "loss": 17.8952,
+      "step": 12028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2655884044089482e-05,
+      "loss": 17.0271,
+      "step": 12029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2648065938510725e-05,
+      "loss": 17.3868,
+      "step": 12030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2640248787124047e-05,
+      "loss": 0.6984,
+      "step": 12031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2632432590202134e-05,
+      "loss": 0.6821,
+      "step": 12032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2624617348017656e-05,
+      "loss": 0.613,
+      "step": 12033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2616803060843283e-05,
+      "loss": 19.6003,
+      "step": 12034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2608989728951614e-05,
+      "loss": 17.2764,
+      "step": 12035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2601177352615205e-05,
+      "loss": 19.2544,
+      "step": 12036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2593365932106613e-05,
+      "loss": 17.614,
+      "step": 12037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2585555467698315e-05,
+      "loss": 19.31,
+      "step": 12038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2577745959662826e-05,
+      "loss": 16.927,
+      "step": 12039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2569937408272574e-05,
+      "loss": 17.6428,
+      "step": 12040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.256212981379996e-05,
+      "loss": 16.9003,
+      "step": 12041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2554323176517355e-05,
+      "loss": 17.4748,
+      "step": 12042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2546517496697083e-05,
+      "loss": 17.0256,
+      "step": 12043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2538712774611486e-05,
+      "loss": 21.0026,
+      "step": 12044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.253090901053283e-05,
+      "loss": 18.0166,
+      "step": 12045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.252310620473334e-05,
+      "loss": 0.6866,
+      "step": 12046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.251530435748523e-05,
+      "loss": 0.6284,
+      "step": 12047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2507503469060655e-05,
+      "loss": 17.5933,
+      "step": 12048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.249970353973179e-05,
+      "loss": 0.6884,
+      "step": 12049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2491904569770723e-05,
+      "loss": 0.6557,
+      "step": 12050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2484106559449524e-05,
+      "loss": 18.7308,
+      "step": 12051
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.247630950904024e-05,
+      "loss": 0.6198,
+      "step": 12052
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2468513418814847e-05,
+      "loss": 0.669,
+      "step": 12053
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2460718289045364e-05,
+      "loss": 0.668,
+      "step": 12054
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2452924120003703e-05,
+      "loss": 18.8663,
+      "step": 12055
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2445130911961782e-05,
+      "loss": 16.9558,
+      "step": 12056
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.243733866519146e-05,
+      "loss": 17.621,
+      "step": 12057
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.242954737996456e-05,
+      "loss": 0.7485,
+      "step": 12058
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.242175705655293e-05,
+      "loss": 17.2703,
+      "step": 12059
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.241396769522831e-05,
+      "loss": 17.5214,
+      "step": 12060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2406179296262453e-05,
+      "loss": 17.6905,
+      "step": 12061
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.239839185992705e-05,
+      "loss": 18.0976,
+      "step": 12062
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2390605386493757e-05,
+      "loss": 18.8384,
+      "step": 12063
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2382819876234253e-05,
+      "loss": 0.668,
+      "step": 12064
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.237503532942012e-05,
+      "loss": 19.6834,
+      "step": 12065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.236725174632292e-05,
+      "loss": 0.7082,
+      "step": 12066
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2359469127214195e-05,
+      "loss": 0.7097,
+      "step": 12067
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.235168747236543e-05,
+      "loss": 17.6818,
+      "step": 12068
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.234390678204813e-05,
+      "loss": 18.4147,
+      "step": 12069
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.233612705653371e-05,
+      "loss": 18.7479,
+      "step": 12070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2328348296093576e-05,
+      "loss": 16.1353,
+      "step": 12071
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.232057050099909e-05,
+      "loss": 17.5775,
+      "step": 12072
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2312793671521586e-05,
+      "loss": 18.5666,
+      "step": 12073
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2305017807932365e-05,
+      "loss": 0.6617,
+      "step": 12074
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2297242910502697e-05,
+      "loss": 0.6756,
+      "step": 12075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2289468979503814e-05,
+      "loss": 17.2568,
+      "step": 12076
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2281696015206894e-05,
+      "loss": 16.7371,
+      "step": 12077
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.227392401788314e-05,
+      "loss": 0.6922,
+      "step": 12078
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.226615298780367e-05,
+      "loss": 17.1413,
+      "step": 12079
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2258382925239575e-05,
+      "loss": 0.6873,
+      "step": 12080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2250613830461926e-05,
+      "loss": 0.6669,
+      "step": 12081
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2242845703741728e-05,
+      "loss": 17.9721,
+      "step": 12082
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.223507854535002e-05,
+      "loss": 16.8488,
+      "step": 12083
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2227312355557746e-05,
+      "loss": 18.9056,
+      "step": 12084
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.221954713463583e-05,
+      "loss": 18.9559,
+      "step": 12085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2211782882855174e-05,
+      "loss": 18.5693,
+      "step": 12086
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2204019600486618e-05,
+      "loss": 0.7109,
+      "step": 12087
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.219625728780103e-05,
+      "loss": 19.5952,
+      "step": 12088
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2188495945069183e-05,
+      "loss": 18.7505,
+      "step": 12089
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2180735572561833e-05,
+      "loss": 18.102,
+      "step": 12090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2172976170549708e-05,
+      "loss": 15.3153,
+      "step": 12091
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2165217739303508e-05,
+      "loss": 20.2292,
+      "step": 12092
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2157460279093863e-05,
+      "loss": 18.0686,
+      "step": 12093
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.214970379019144e-05,
+      "loss": 0.6647,
+      "step": 12094
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2141948272866812e-05,
+      "loss": 17.4542,
+      "step": 12095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2134193727390533e-05,
+      "loss": 0.6665,
+      "step": 12096
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2126440154033123e-05,
+      "loss": 18.4274,
+      "step": 12097
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2118687553065064e-05,
+      "loss": 17.8297,
+      "step": 12098
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2110935924756837e-05,
+      "loss": 0.7215,
+      "step": 12099
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2103185269378847e-05,
+      "loss": 16.8667,
+      "step": 12100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2095435587201485e-05,
+      "loss": 17.6204,
+      "step": 12101
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2087686878495102e-05,
+      "loss": 0.6801,
+      "step": 12102
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2079939143529992e-05,
+      "loss": 0.6461,
+      "step": 12103
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.207219238257649e-05,
+      "loss": 17.7095,
+      "step": 12104
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.206444659590482e-05,
+      "loss": 15.5846,
+      "step": 12105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2056701783785205e-05,
+      "loss": 0.6631,
+      "step": 12106
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2048957946487826e-05,
+      "loss": 0.7307,
+      "step": 12107
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2041215084282813e-05,
+      "loss": 18.5481,
+      "step": 12108
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.203347319744031e-05,
+      "loss": 18.4945,
+      "step": 12109
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.20257322862304e-05,
+      "loss": 0.6511,
+      "step": 12110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2017992350923117e-05,
+      "loss": 16.5779,
+      "step": 12111
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2010253391788476e-05,
+      "loss": 17.44,
+      "step": 12112
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2002515409096435e-05,
+      "loss": 17.8815,
+      "step": 12113
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1994778403116984e-05,
+      "loss": 19.2584,
+      "step": 12114
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1987042374120005e-05,
+      "loss": 0.6928,
+      "step": 12115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1979307322375387e-05,
+      "loss": 15.9148,
+      "step": 12116
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.197157324815296e-05,
+      "loss": 16.859,
+      "step": 12117
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1963840151722524e-05,
+      "loss": 0.6687,
+      "step": 12118
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.195610803335389e-05,
+      "loss": 17.3761,
+      "step": 12119
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1948376893316775e-05,
+      "loss": 0.6766,
+      "step": 12120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.194064673188089e-05,
+      "loss": 18.062,
+      "step": 12121
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1932917549315897e-05,
+      "loss": 0.6567,
+      "step": 12122
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1925189345891435e-05,
+      "loss": 0.6482,
+      "step": 12123
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1917462121877125e-05,
+      "loss": 17.7532,
+      "step": 12124
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1909735877542532e-05,
+      "loss": 0.659,
+      "step": 12125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1902010613157205e-05,
+      "loss": 17.2349,
+      "step": 12126
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1894286328990588e-05,
+      "loss": 17.1439,
+      "step": 12127
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.188656302531221e-05,
+      "loss": 0.6471,
+      "step": 12128
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1878840702391478e-05,
+      "loss": 0.6859,
+      "step": 12129
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.18711193604978e-05,
+      "loss": 17.8336,
+      "step": 12130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.186339899990053e-05,
+      "loss": 0.6767,
+      "step": 12131
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1855679620868985e-05,
+      "loss": 0.6241,
+      "step": 12132
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1847961223672502e-05,
+      "loss": 16.9391,
+      "step": 12133
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1840243808580317e-05,
+      "loss": 17.8764,
+      "step": 12134
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1832527375861667e-05,
+      "loss": 18.355,
+      "step": 12135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1824811925785732e-05,
+      "loss": 17.4617,
+      "step": 12136
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1817097458621666e-05,
+      "loss": 0.6478,
+      "step": 12137
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.180938397463862e-05,
+      "loss": 18.4908,
+      "step": 12138
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.180167147410567e-05,
+      "loss": 0.6918,
+      "step": 12139
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1793959957291876e-05,
+      "loss": 18.7623,
+      "step": 12140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.178624942446626e-05,
+      "loss": 0.6167,
+      "step": 12141
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1778539875897785e-05,
+      "loss": 17.9056,
+      "step": 12142
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1770831311855444e-05,
+      "loss": 19.1593,
+      "step": 12143
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.176312373260814e-05,
+      "loss": 19.0407,
+      "step": 12144
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1755417138424748e-05,
+      "loss": 0.6594,
+      "step": 12145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174771152957413e-05,
+      "loss": 16.9521,
+      "step": 12146
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174000690632508e-05,
+      "loss": 0.7182,
+      "step": 12147
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1732303268946412e-05,
+      "loss": 16.1042,
+      "step": 12148
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.172460061770686e-05,
+      "loss": 17.7571,
+      "step": 12149
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171689895287513e-05,
+      "loss": 16.6949,
+      "step": 12150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1709198274719907e-05,
+      "loss": 17.2812,
+      "step": 12151
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.170149858350981e-05,
+      "loss": 16.8983,
+      "step": 12152
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.169379987951349e-05,
+      "loss": 0.6476,
+      "step": 12153
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.16861021629995e-05,
+      "loss": 0.6953,
+      "step": 12154
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1678405434236386e-05,
+      "loss": 0.6486,
+      "step": 12155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1670709693492647e-05,
+      "loss": 17.3913,
+      "step": 12156
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1663014941036742e-05,
+      "loss": 17.7339,
+      "step": 12157
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1655321177137144e-05,
+      "loss": 0.635,
+      "step": 12158
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1647628402062233e-05,
+      "loss": 18.3868,
+      "step": 12159
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1639936616080376e-05,
+      "loss": 16.882,
+      "step": 12160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1632245819459913e-05,
+      "loss": 17.5315,
+      "step": 12161
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.162455601246912e-05,
+      "loss": 0.6943,
+      "step": 12162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1616867195376307e-05,
+      "loss": 17.5359,
+      "step": 12163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.160917936844967e-05,
+      "loss": 0.6941,
+      "step": 12164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1601492531957417e-05,
+      "loss": 17.2117,
+      "step": 12165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.15938066861677e-05,
+      "loss": 16.819,
+      "step": 12166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.158612183134864e-05,
+      "loss": 0.6674,
+      "step": 12167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.157843796776835e-05,
+      "loss": 17.946,
+      "step": 12168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.157075509569488e-05,
+      "loss": 0.5971,
+      "step": 12169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1563073215396246e-05,
+      "loss": 17.1625,
+      "step": 12170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.155539232714044e-05,
+      "loss": 17.1626,
+      "step": 12171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1547712431195393e-05,
+      "loss": 0.7095,
+      "step": 12172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1540033527829063e-05,
+      "loss": 16.5851,
+      "step": 12173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.153235561730932e-05,
+      "loss": 17.3253,
+      "step": 12174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1524678699904006e-05,
+      "loss": 18.4011,
+      "step": 12175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1517002775880936e-05,
+      "loss": 16.0215,
+      "step": 12176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1509327845507883e-05,
+      "loss": 16.3131,
+      "step": 12177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.150165390905261e-05,
+      "loss": 0.6909,
+      "step": 12178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.149398096678283e-05,
+      "loss": 0.6232,
+      "step": 12179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1486309018966205e-05,
+      "loss": 0.7315,
+      "step": 12180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1478638065870383e-05,
+      "loss": 16.5174,
+      "step": 12181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1470968107762968e-05,
+      "loss": 18.3974,
+      "step": 12182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.146329914491153e-05,
+      "loss": 17.9279,
+      "step": 12183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.145563117758361e-05,
+      "loss": 18.5408,
+      "step": 12184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1447964206046718e-05,
+      "loss": 16.3856,
+      "step": 12185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.144029823056829e-05,
+      "loss": 17.3371,
+      "step": 12186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.14326332514158e-05,
+      "loss": 18.5183,
+      "step": 12187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.142496926885663e-05,
+      "loss": 17.8045,
+      "step": 12188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1417306283158146e-05,
+      "loss": 18.3338,
+      "step": 12189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1409644294587667e-05,
+      "loss": 16.1043,
+      "step": 12190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1401983303412483e-05,
+      "loss": 18.1846,
+      "step": 12191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1394323309899878e-05,
+      "loss": 17.0168,
+      "step": 12192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1386664314317062e-05,
+      "loss": 17.7564,
+      "step": 12193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.137900631693122e-05,
+      "loss": 17.5007,
+      "step": 12194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1371349318009522e-05,
+      "loss": 16.7595,
+      "step": 12195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1363693317819055e-05,
+      "loss": 16.6931,
+      "step": 12196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.135603831662694e-05,
+      "loss": 17.5491,
+      "step": 12197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.134838431470022e-05,
+      "loss": 19.0784,
+      "step": 12198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1340731312305905e-05,
+      "loss": 16.7465,
+      "step": 12199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1333079309710973e-05,
+      "loss": 19.8941,
+      "step": 12200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1325428307182354e-05,
+      "loss": 17.5237,
+      "step": 12201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1317778304986997e-05,
+      "loss": 0.6053,
+      "step": 12202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1310129303391756e-05,
+      "loss": 0.7136,
+      "step": 12203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1302481302663472e-05,
+      "loss": 16.8213,
+      "step": 12204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.129483430306895e-05,
+      "loss": 0.659,
+      "step": 12205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.128718830487495e-05,
+      "loss": 0.6089,
+      "step": 12206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.127954330834824e-05,
+      "loss": 0.6793,
+      "step": 12207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.12718993137555e-05,
+      "loss": 0.7472,
+      "step": 12208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.12642563213634e-05,
+      "loss": 17.3314,
+      "step": 12209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1256614331438572e-05,
+      "loss": 0.7151,
+      "step": 12210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1248973344247592e-05,
+      "loss": 16.918,
+      "step": 12211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.124133336005706e-05,
+      "loss": 17.3074,
+      "step": 12212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.123369437913348e-05,
+      "loss": 0.697,
+      "step": 12213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122605640174335e-05,
+      "loss": 16.4822,
+      "step": 12214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.121841942815312e-05,
+      "loss": 0.6905,
+      "step": 12215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.121078345862919e-05,
+      "loss": 18.0718,
+      "step": 12216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1203148493437998e-05,
+      "loss": 17.3611,
+      "step": 12217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.119551453284586e-05,
+      "loss": 15.7718,
+      "step": 12218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1187881577119108e-05,
+      "loss": 17.3982,
+      "step": 12219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.118024962652401e-05,
+      "loss": 16.2417,
+      "step": 12220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.11726186813268e-05,
+      "loss": 0.6879,
+      "step": 12221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1164988741793724e-05,
+      "loss": 0.6351,
+      "step": 12222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.115735980819094e-05,
+      "loss": 0.6876,
+      "step": 12223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.114973188078459e-05,
+      "loss": 17.4671,
+      "step": 12224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1142104959840775e-05,
+      "loss": 0.6674,
+      "step": 12225
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1134479045625553e-05,
+      "loss": 16.5538,
+      "step": 12226
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1126854138405e-05,
+      "loss": 0.6498,
+      "step": 12227
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1119230238445085e-05,
+      "loss": 0.691,
+      "step": 12228
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1111607346011785e-05,
+      "loss": 17.9155,
+      "step": 12229
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1103985461371022e-05,
+      "loss": 0.6453,
+      "step": 12230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.109636458478868e-05,
+      "loss": 0.7736,
+      "step": 12231
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1088744716530656e-05,
+      "loss": 16.7056,
+      "step": 12232
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1081125856862743e-05,
+      "loss": 18.0651,
+      "step": 12233
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1073508006050745e-05,
+      "loss": 0.6385,
+      "step": 12234
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1065891164360414e-05,
+      "loss": 19.196,
+      "step": 12235
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.105827533205746e-05,
+      "loss": 0.652,
+      "step": 12236
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.105066050940758e-05,
+      "loss": 17.7692,
+      "step": 12237
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1043046696676404e-05,
+      "loss": 0.6813,
+      "step": 12238
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.103543389412957e-05,
+      "loss": 0.6584,
+      "step": 12239
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.102782210203262e-05,
+      "loss": 17.7347,
+      "step": 12240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1020211320651135e-05,
+      "loss": 17.4408,
+      "step": 12241
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1012601550250606e-05,
+      "loss": 0.681,
+      "step": 12242
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1004992791096512e-05,
+      "loss": 0.6316,
+      "step": 12243
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0997385043454283e-05,
+      "loss": 0.7113,
+      "step": 12244
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.09897783075893e-05,
+      "loss": 17.3235,
+      "step": 12245
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.098217258376698e-05,
+      "loss": 17.8852,
+      "step": 12246
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0974567872252622e-05,
+      "loss": 0.6774,
+      "step": 12247
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0966964173311526e-05,
+      "loss": 0.6644,
+      "step": 12248
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0959361487208955e-05,
+      "loss": 17.9275,
+      "step": 12249
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0951759814210108e-05,
+      "loss": 17.8897,
+      "step": 12250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0944159154580223e-05,
+      "loss": 0.687,
+      "step": 12251
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.093655950858443e-05,
+      "loss": 0.6883,
+      "step": 12252
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0928960876487854e-05,
+      "loss": 18.1573,
+      "step": 12253
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0921363258555565e-05,
+      "loss": 18.5515,
+      "step": 12254
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.09137666550526e-05,
+      "loss": 0.6352,
+      "step": 12255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0906171066244018e-05,
+      "loss": 0.6867,
+      "step": 12256
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0898576492394768e-05,
+      "loss": 0.7411,
+      "step": 12257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0890982933769793e-05,
+      "loss": 0.6862,
+      "step": 12258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0883390390634e-05,
+      "loss": 19.7458,
+      "step": 12259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0875798863252245e-05,
+      "loss": 18.2889,
+      "step": 12260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0868208351889402e-05,
+      "loss": 0.695,
+      "step": 12261
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.086061885681025e-05,
+      "loss": 0.6561,
+      "step": 12262
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0853030378279555e-05,
+      "loss": 17.6518,
+      "step": 12263
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.084544291656204e-05,
+      "loss": 0.7303,
+      "step": 12264
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.083785647192239e-05,
+      "loss": 18.9308,
+      "step": 12265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.08302710446253e-05,
+      "loss": 0.6546,
+      "step": 12266
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0822686634935374e-05,
+      "loss": 15.8619,
+      "step": 12267
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.081510324311719e-05,
+      "loss": 16.7639,
+      "step": 12268
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.080752086943531e-05,
+      "loss": 0.701,
+      "step": 12269
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.079993951415423e-05,
+      "loss": 20.2272,
+      "step": 12270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0792359177538474e-05,
+      "loss": 0.7022,
+      "step": 12271
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0784779859852456e-05,
+      "loss": 16.4773,
+      "step": 12272
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0777201561360597e-05,
+      "loss": 18.7882,
+      "step": 12273
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0769624282327266e-05,
+      "loss": 17.327,
+      "step": 12274
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.07620480230168e-05,
+      "loss": 0.671,
+      "step": 12275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.075447278369349e-05,
+      "loss": 17.9774,
+      "step": 12276
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0746898564621636e-05,
+      "loss": 0.6422,
+      "step": 12277
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.073932536606546e-05,
+      "loss": 18.7237,
+      "step": 12278
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0731753188289144e-05,
+      "loss": 18.6953,
+      "step": 12279
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.072418203155686e-05,
+      "loss": 17.9306,
+      "step": 12280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.071661189613271e-05,
+      "loss": 17.4422,
+      "step": 12281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0709042782280825e-05,
+      "loss": 18.5361,
+      "step": 12282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.070147469026524e-05,
+      "loss": 18.6233,
+      "step": 12283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.069390762034996e-05,
+      "loss": 18.3665,
+      "step": 12284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.068634157279899e-05,
+      "loss": 19.3945,
+      "step": 12285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0678776547876245e-05,
+      "loss": 0.6639,
+      "step": 12286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0671212545845676e-05,
+      "loss": 16.9683,
+      "step": 12287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.066364956697114e-05,
+      "loss": 18.1703,
+      "step": 12288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.065608761151647e-05,
+      "loss": 17.1048,
+      "step": 12289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.064852667974549e-05,
+      "loss": 0.665,
+      "step": 12290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.064096677192195e-05,
+      "loss": 0.64,
+      "step": 12291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0633407888309585e-05,
+      "loss": 18.2602,
+      "step": 12292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.06258500291721e-05,
+      "loss": 0.6849,
+      "step": 12293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0618293194773152e-05,
+      "loss": 18.8011,
+      "step": 12294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.061073738537635e-05,
+      "loss": 0.6914,
+      "step": 12295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0603182601245315e-05,
+      "loss": 0.6377,
+      "step": 12296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.059562884264359e-05,
+      "loss": 0.6899,
+      "step": 12297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0588076109834692e-05,
+      "loss": 0.6332,
+      "step": 12298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0580524403082098e-05,
+      "loss": 17.4892,
+      "step": 12299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0572973722649243e-05,
+      "loss": 0.6846,
+      "step": 12300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.056542406879957e-05,
+      "loss": 17.1754,
+      "step": 12301
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0557875441796444e-05,
+      "loss": 0.6896,
+      "step": 12302
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0550327841903193e-05,
+      "loss": 19.4086,
+      "step": 12303
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0542781269383125e-05,
+      "loss": 0.7239,
+      "step": 12304
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053523572449949e-05,
+      "loss": 0.6378,
+      "step": 12305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.052769120751556e-05,
+      "loss": 18.6401,
+      "step": 12306
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0520147718694505e-05,
+      "loss": 16.4212,
+      "step": 12307
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0512605258299493e-05,
+      "loss": 0.6611,
+      "step": 12308
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.050506382659364e-05,
+      "loss": 16.6672,
+      "step": 12309
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0497523423840024e-05,
+      "loss": 16.025,
+      "step": 12310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.048998405030173e-05,
+      "loss": 19.7658,
+      "step": 12311
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.048244570624176e-05,
+      "loss": 17.279,
+      "step": 12312
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0474908391923088e-05,
+      "loss": 0.6886,
+      "step": 12313
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0467372107608658e-05,
+      "loss": 0.7326,
+      "step": 12314
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.045983685356137e-05,
+      "loss": 17.3968,
+      "step": 12315
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0452302630044122e-05,
+      "loss": 16.2119,
+      "step": 12316
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0444769437319743e-05,
+      "loss": 17.9383,
+      "step": 12317
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.043723727565103e-05,
+      "loss": 18.3025,
+      "step": 12318
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.042970614530075e-05,
+      "loss": 18.8909,
+      "step": 12319
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.04221760465316e-05,
+      "loss": 19.2635,
+      "step": 12320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0414646979606328e-05,
+      "loss": 17.5852,
+      "step": 12321
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0407118944787562e-05,
+      "loss": 17.496,
+      "step": 12322
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0399591942337932e-05,
+      "loss": 0.6613,
+      "step": 12323
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.039206597252001e-05,
+      "loss": 0.6824,
+      "step": 12324
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.038454103559634e-05,
+      "loss": 19.4604,
+      "step": 12325
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0377017131829458e-05,
+      "loss": 18.953,
+      "step": 12326
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.036949426148183e-05,
+      "loss": 0.6448,
+      "step": 12327
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.03619724248159e-05,
+      "loss": 0.6143,
+      "step": 12328
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0354451622094074e-05,
+      "loss": 16.4352,
+      "step": 12329
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0346931853578687e-05,
+      "loss": 0.6525,
+      "step": 12330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0339413119532123e-05,
+      "loss": 17.7876,
+      "step": 12331
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0331895420216657e-05,
+      "loss": 0.6553,
+      "step": 12332
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0324378755894548e-05,
+      "loss": 17.8953,
+      "step": 12333
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0316863126828023e-05,
+      "loss": 18.4128,
+      "step": 12334
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.030934853327924e-05,
+      "loss": 17.4074,
+      "step": 12335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.030183497551041e-05,
+      "loss": 0.7133,
+      "step": 12336
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.029432245378361e-05,
+      "loss": 19.128,
+      "step": 12337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0286810968360924e-05,
+      "loss": 0.7127,
+      "step": 12338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0279300519504406e-05,
+      "loss": 17.4893,
+      "step": 12339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0271791107476036e-05,
+      "loss": 17.572,
+      "step": 12340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0264282732537826e-05,
+      "loss": 17.718,
+      "step": 12341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0256775394951687e-05,
+      "loss": 0.6889,
+      "step": 12342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0249269094979523e-05,
+      "loss": 15.6919,
+      "step": 12343
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.02417638328832e-05,
+      "loss": 17.2557,
+      "step": 12344
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0234259608924533e-05,
+      "loss": 15.4096,
+      "step": 12345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0226756423365323e-05,
+      "loss": 0.7026,
+      "step": 12346
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.021925427646732e-05,
+      "loss": 16.5082,
+      "step": 12347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.021175316849224e-05,
+      "loss": 0.7139,
+      "step": 12348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.020425309970175e-05,
+      "loss": 0.792,
+      "step": 12349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.019675407035754e-05,
+      "loss": 0.6476,
+      "step": 12350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.018925608072118e-05,
+      "loss": 17.1945,
+      "step": 12351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0181759131054257e-05,
+      "loss": 0.617,
+      "step": 12352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0174263221618307e-05,
+      "loss": 18.7983,
+      "step": 12353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0166768352674803e-05,
+      "loss": 18.5046,
+      "step": 12354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0159274524485255e-05,
+      "loss": 17.861,
+      "step": 12355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.015178173731107e-05,
+      "loss": 19.4589,
+      "step": 12356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0144289991413646e-05,
+      "loss": 18.4784,
+      "step": 12357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.013679928705432e-05,
+      "loss": 16.8657,
+      "step": 12358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.012930962449441e-05,
+      "loss": 0.6349,
+      "step": 12359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0121821003995222e-05,
+      "loss": 18.1794,
+      "step": 12360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0114333425817993e-05,
+      "loss": 16.3134,
+      "step": 12361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.010684689022393e-05,
+      "loss": 17.6797,
+      "step": 12362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0099361397474208e-05,
+      "loss": 0.6918,
+      "step": 12363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.009187694782994e-05,
+      "loss": 17.7629,
+      "step": 12364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0084393541552264e-05,
+      "loss": 0.6798,
+      "step": 12365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.007691117890223e-05,
+      "loss": 17.729,
+      "step": 12366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0069429860140864e-05,
+      "loss": 19.0104,
+      "step": 12367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0061949585529156e-05,
+      "loss": 0.6646,
+      "step": 12368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.005447035532804e-05,
+      "loss": 18.0613,
+      "step": 12369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0046992169798486e-05,
+      "loss": 18.5071,
+      "step": 12370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0039515029201333e-05,
+      "loss": 0.5936,
+      "step": 12371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0032038933797447e-05,
+      "loss": 17.9376,
+      "step": 12372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0024563883847625e-05,
+      "loss": 19.4123,
+      "step": 12373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0017089879612628e-05,
+      "loss": 18.3566,
+      "step": 12374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0009616921353223e-05,
+      "loss": 16.8537,
+      "step": 12375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0002145009330098e-05,
+      "loss": 17.3914,
+      "step": 12376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.999467414380391e-05,
+      "loss": 18.1506,
+      "step": 12377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9987204325035292e-05,
+      "loss": 14.9834,
+      "step": 12378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.997973555328481e-05,
+      "loss": 16.6169,
+      "step": 12379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9972267828813057e-05,
+      "loss": 17.9471,
+      "step": 12380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9964801151880524e-05,
+      "loss": 18.0718,
+      "step": 12381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9957335522747707e-05,
+      "loss": 16.923,
+      "step": 12382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9949870941675037e-05,
+      "loss": 19.0923,
+      "step": 12383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.994240740892291e-05,
+      "loss": 16.2972,
+      "step": 12384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9934944924751725e-05,
+      "loss": 16.7096,
+      "step": 12385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9927483489421805e-05,
+      "loss": 16.6088,
+      "step": 12386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9920023103193453e-05,
+      "loss": 0.6926,
+      "step": 12387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9912563766326915e-05,
+      "loss": 0.6811,
+      "step": 12388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9905105479082408e-05,
+      "loss": 0.6598,
+      "step": 12389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9897648241720152e-05,
+      "loss": 15.4263,
+      "step": 12390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9890192054500285e-05,
+      "loss": 17.3466,
+      "step": 12391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.988273691768291e-05,
+      "loss": 17.9663,
+      "step": 12392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9875282831528118e-05,
+      "loss": 18.1076,
+      "step": 12393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9867829796295927e-05,
+      "loss": 17.1261,
+      "step": 12394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9860377812246372e-05,
+      "loss": 16.817,
+      "step": 12395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9852926879639415e-05,
+      "loss": 17.9323,
+      "step": 12396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.984547699873498e-05,
+      "loss": 17.8058,
+      "step": 12397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9838028169792954e-05,
+      "loss": 16.7276,
+      "step": 12398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9830580393073213e-05,
+      "loss": 0.6681,
+      "step": 12399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.982313366883556e-05,
+      "loss": 0.6927,
+      "step": 12400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.981568799733979e-05,
+      "loss": 16.8471,
+      "step": 12401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.980824337884565e-05,
+      "loss": 0.6608,
+      "step": 12402
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9800799813612824e-05,
+      "loss": 0.7115,
+      "step": 12403
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9793357301901033e-05,
+      "loss": 18.4963,
+      "step": 12404
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9785915843969894e-05,
+      "loss": 15.904,
+      "step": 12405
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9778475440079002e-05,
+      "loss": 0.6545,
+      "step": 12406
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9771036090487933e-05,
+      "loss": 0.7049,
+      "step": 12407
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976359779545618e-05,
+      "loss": 18.5176,
+      "step": 12408
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.975616055524328e-05,
+      "loss": 16.5411,
+      "step": 12409
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9748724370108666e-05,
+      "loss": 17.3361,
+      "step": 12410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9741289240311755e-05,
+      "loss": 0.678,
+      "step": 12411
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9733855166111924e-05,
+      "loss": 16.0082,
+      "step": 12412
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9726422147768504e-05,
+      "loss": 18.4248,
+      "step": 12413
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.971899018554083e-05,
+      "loss": 0.6949,
+      "step": 12414
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.971155927968816e-05,
+      "loss": 0.7139,
+      "step": 12415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.970412943046972e-05,
+      "loss": 18.1283,
+      "step": 12416
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.969670063814471e-05,
+      "loss": 16.4476,
+      "step": 12417
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9689272902972272e-05,
+      "loss": 0.7007,
+      "step": 12418
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9681846225211564e-05,
+      "loss": 0.6466,
+      "step": 12419
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9674420605121646e-05,
+      "loss": 16.6901,
+      "step": 12420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9666996042961578e-05,
+      "loss": 18.3613,
+      "step": 12421
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9659572538990355e-05,
+      "loss": 19.3472,
+      "step": 12422
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9652150093466953e-05,
+      "loss": 15.8406,
+      "step": 12423
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.964472870665033e-05,
+      "loss": 18.3479,
+      "step": 12424
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9637308378799373e-05,
+      "loss": 16.8953,
+      "step": 12425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.962988911017295e-05,
+      "loss": 16.8685,
+      "step": 12426
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9622470901029883e-05,
+      "loss": 19.1728,
+      "step": 12427
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9615053751628942e-05,
+      "loss": 0.6738,
+      "step": 12428
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9607637662228918e-05,
+      "loss": 17.6897,
+      "step": 12429
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.960022263308851e-05,
+      "loss": 18.5663,
+      "step": 12430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9592808664466395e-05,
+      "loss": 17.1931,
+      "step": 12431
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9585395756621216e-05,
+      "loss": 16.9131,
+      "step": 12432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9577983909811555e-05,
+      "loss": 0.6314,
+      "step": 12433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9570573124296022e-05,
+      "loss": 16.5302,
+      "step": 12434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.956316340033313e-05,
+      "loss": 0.6653,
+      "step": 12435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9555754738181364e-05,
+      "loss": 19.3152,
+      "step": 12436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9548347138099187e-05,
+      "loss": 17.9481,
+      "step": 12437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9540940600345004e-05,
+      "loss": 17.1295,
+      "step": 12438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9533535125177227e-05,
+      "loss": 17.4466,
+      "step": 12439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9526130712854185e-05,
+      "loss": 0.6623,
+      "step": 12440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9518727363634186e-05,
+      "loss": 17.0846,
+      "step": 12441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9511325077775506e-05,
+      "loss": 18.3729,
+      "step": 12442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950392385553635e-05,
+      "loss": 18.1359,
+      "step": 12443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9496523697174962e-05,
+      "loss": 18.1968,
+      "step": 12444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9489124602949482e-05,
+      "loss": 0.7231,
+      "step": 12445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9481726573118025e-05,
+      "loss": 17.4411,
+      "step": 12446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.947432960793869e-05,
+      "loss": 16.9989,
+      "step": 12447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9466933707669493e-05,
+      "loss": 18.4173,
+      "step": 12448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9459538872568493e-05,
+      "loss": 19.1711,
+      "step": 12449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9452145102893642e-05,
+      "loss": 18.5749,
+      "step": 12450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.944475239890287e-05,
+      "loss": 18.4945,
+      "step": 12451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.943736076085409e-05,
+      "loss": 16.5287,
+      "step": 12452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9429970189005136e-05,
+      "loss": 17.9118,
+      "step": 12453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9422580683613896e-05,
+      "loss": 18.0187,
+      "step": 12454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9415192244938098e-05,
+      "loss": 0.741,
+      "step": 12455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9407804873235507e-05,
+      "loss": 16.5531,
+      "step": 12456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9400418568763827e-05,
+      "loss": 17.313,
+      "step": 12457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9393033331780775e-05,
+      "loss": 0.7291,
+      "step": 12458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9385649162543956e-05,
+      "loss": 0.7339,
+      "step": 12459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.937826606131099e-05,
+      "loss": 17.0023,
+      "step": 12460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.937088402833943e-05,
+      "loss": 0.6781,
+      "step": 12461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936350306388679e-05,
+      "loss": 18.8522,
+      "step": 12462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9356123168210595e-05,
+      "loss": 17.8158,
+      "step": 12463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9348744341568276e-05,
+      "loss": 17.3246,
+      "step": 12464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9341366584217256e-05,
+      "loss": 16.7958,
+      "step": 12465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9333989896414912e-05,
+      "loss": 0.7129,
+      "step": 12466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9326614278418565e-05,
+      "loss": 16.9666,
+      "step": 12467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.931923973048555e-05,
+      "loss": 0.6653,
+      "step": 12468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.931186625287313e-05,
+      "loss": 17.7616,
+      "step": 12469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9304493845838518e-05,
+      "loss": 0.6279,
+      "step": 12470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9297122509638917e-05,
+      "loss": 18.5507,
+      "step": 12471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.928975224453147e-05,
+      "loss": 0.6779,
+      "step": 12472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9282383050773297e-05,
+      "loss": 17.5739,
+      "step": 12473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9275014928621488e-05,
+      "loss": 17.184,
+      "step": 12474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9267647878333078e-05,
+      "loss": 0.6862,
+      "step": 12475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9260281900165078e-05,
+      "loss": 0.6517,
+      "step": 12476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9252916994374454e-05,
+      "loss": 0.7235,
+      "step": 12477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.924555316121811e-05,
+      "loss": 18.4736,
+      "step": 12478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9238190400952977e-05,
+      "loss": 17.3346,
+      "step": 12479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9230828713835897e-05,
+      "loss": 18.0741,
+      "step": 12480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9223468100123688e-05,
+      "loss": 0.6911,
+      "step": 12481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9216108560073125e-05,
+      "loss": 16.3933,
+      "step": 12482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9208750093940936e-05,
+      "loss": 18.6788,
+      "step": 12483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.920139270198386e-05,
+      "loss": 19.8788,
+      "step": 12484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9194036384458553e-05,
+      "loss": 0.6555,
+      "step": 12485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9186681141621637e-05,
+      "loss": 17.6409,
+      "step": 12486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9179326973729712e-05,
+      "loss": 18.4027,
+      "step": 12487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9171973881039313e-05,
+      "loss": 0.6653,
+      "step": 12488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.916462186380699e-05,
+      "loss": 0.6145,
+      "step": 12489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9157270922289217e-05,
+      "loss": 19.0862,
+      "step": 12490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9149921056742427e-05,
+      "loss": 17.6683,
+      "step": 12491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.914257226742302e-05,
+      "loss": 17.0025,
+      "step": 12492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.913522455458736e-05,
+      "loss": 0.6561,
+      "step": 12493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9127877918491805e-05,
+      "loss": 17.7705,
+      "step": 12494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9120532359392636e-05,
+      "loss": 0.6482,
+      "step": 12495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9113187877546096e-05,
+      "loss": 17.8096,
+      "step": 12496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9105844473208417e-05,
+      "loss": 0.6915,
+      "step": 12497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.909850214663575e-05,
+      "loss": 0.6767,
+      "step": 12498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9091160898084282e-05,
+      "loss": 15.5495,
+      "step": 12499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9083820727810085e-05,
+      "loss": 18.0322,
+      "step": 12500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.907648163606925e-05,
+      "loss": 16.8526,
+      "step": 12501
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9069143623117786e-05,
+      "loss": 17.5468,
+      "step": 12502
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9061806689211675e-05,
+      "loss": 17.0997,
+      "step": 12503
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.905447083460691e-05,
+      "loss": 0.6688,
+      "step": 12504
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9047136059559377e-05,
+      "loss": 0.6769,
+      "step": 12505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.903980236432497e-05,
+      "loss": 16.9135,
+      "step": 12506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.903246974915952e-05,
+      "loss": 19.1224,
+      "step": 12507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9025138214318837e-05,
+      "loss": 18.2257,
+      "step": 12508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9017807760058688e-05,
+      "loss": 0.6691,
+      "step": 12509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9010478386634796e-05,
+      "loss": 17.6582,
+      "step": 12510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9003150094302852e-05,
+      "loss": 19.2818,
+      "step": 12511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8995822883318492e-05,
+      "loss": 18.271,
+      "step": 12512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8988496753937367e-05,
+      "loss": 18.9252,
+      "step": 12513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.898117170641504e-05,
+      "loss": 18.3665,
+      "step": 12514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8973847741007045e-05,
+      "loss": 0.6612,
+      "step": 12515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8966524857968886e-05,
+      "loss": 0.6592,
+      "step": 12516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.895920305755601e-05,
+      "loss": 0.7012,
+      "step": 12517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8951882340023875e-05,
+      "loss": 17.8683,
+      "step": 12518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8944562705627855e-05,
+      "loss": 18.0775,
+      "step": 12519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8937244154623306e-05,
+      "loss": 0.6834,
+      "step": 12520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8929926687265533e-05,
+      "loss": 17.6737,
+      "step": 12521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8922610303809795e-05,
+      "loss": 17.0502,
+      "step": 12522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8915295004511363e-05,
+      "loss": 17.5767,
+      "step": 12523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.890798078962543e-05,
+      "loss": 16.9137,
+      "step": 12524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8900667659407147e-05,
+      "loss": 0.6901,
+      "step": 12525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8893355614111634e-05,
+      "loss": 0.7101,
+      "step": 12526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8886044653993968e-05,
+      "loss": 19.0021,
+      "step": 12527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8878734779309233e-05,
+      "loss": 15.4764,
+      "step": 12528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8871425990312418e-05,
+      "loss": 0.6275,
+      "step": 12529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8864118287258492e-05,
+      "loss": 18.7498,
+      "step": 12530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8856811670402402e-05,
+      "loss": 0.6571,
+      "step": 12531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.884950613999901e-05,
+      "loss": 17.8578,
+      "step": 12532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8842201696303224e-05,
+      "loss": 18.3384,
+      "step": 12533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.883489833956984e-05,
+      "loss": 18.1176,
+      "step": 12534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8827596070053645e-05,
+      "loss": 17.3076,
+      "step": 12535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8820294888009376e-05,
+      "loss": 0.6575,
+      "step": 12536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8812994793691736e-05,
+      "loss": 0.662,
+      "step": 12537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8805695787355415e-05,
+      "loss": 18.4574,
+      "step": 12538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8798397869255034e-05,
+      "loss": 16.8621,
+      "step": 12539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8791101039645187e-05,
+      "loss": 17.2411,
+      "step": 12540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8783805298780428e-05,
+      "loss": 17.5982,
+      "step": 12541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8776510646915253e-05,
+      "loss": 0.6526,
+      "step": 12542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.876921708430418e-05,
+      "loss": 18.1682,
+      "step": 12543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.876192461120163e-05,
+      "loss": 17.2444,
+      "step": 12544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8754633227862012e-05,
+      "loss": 19.1716,
+      "step": 12545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8747342934539685e-05,
+      "loss": 17.4276,
+      "step": 12546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8740053731488965e-05,
+      "loss": 17.1421,
+      "step": 12547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8732765618964172e-05,
+      "loss": 0.67,
+      "step": 12548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8725478597219538e-05,
+      "loss": 17.9479,
+      "step": 12549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8718192666509275e-05,
+      "loss": 0.6849,
+      "step": 12550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.871090782708756e-05,
+      "loss": 0.67,
+      "step": 12551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.870362407920852e-05,
+      "loss": 18.7591,
+      "step": 12552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8696341423126274e-05,
+      "loss": 17.8582,
+      "step": 12553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.868905985909488e-05,
+      "loss": 0.6618,
+      "step": 12554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8681779387368354e-05,
+      "loss": 0.6466,
+      "step": 12555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8674500008200674e-05,
+      "loss": 17.9568,
+      "step": 12556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8667221721845778e-05,
+      "loss": 17.5066,
+      "step": 12557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865994452855761e-05,
+      "loss": 0.6223,
+      "step": 12558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8652668428590016e-05,
+      "loss": 18.479,
+      "step": 12559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8645393422196834e-05,
+      "loss": 17.9906,
+      "step": 12560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8638119509631853e-05,
+      "loss": 19.108,
+      "step": 12561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.863084669114882e-05,
+      "loss": 18.4584,
+      "step": 12562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8623574967001488e-05,
+      "loss": 17.7153,
+      "step": 12563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.86163043374435e-05,
+      "loss": 0.6178,
+      "step": 12564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.860903480272851e-05,
+      "loss": 0.7658,
+      "step": 12565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.86017663631101e-05,
+      "loss": 19.1364,
+      "step": 12566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8594499018841872e-05,
+      "loss": 19.6891,
+      "step": 12567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8587232770177336e-05,
+      "loss": 0.6684,
+      "step": 12568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.857996761736998e-05,
+      "loss": 0.7138,
+      "step": 12569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8572703560673253e-05,
+      "loss": 0.6956,
+      "step": 12570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8565440600340557e-05,
+      "loss": 18.6139,
+      "step": 12571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8558178736625287e-05,
+      "loss": 0.709,
+      "step": 12572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8550917969780773e-05,
+      "loss": 0.6704,
+      "step": 12573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8543658300060302e-05,
+      "loss": 0.6994,
+      "step": 12574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.853639972771714e-05,
+      "loss": 17.626,
+      "step": 12575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.852914225300449e-05,
+      "loss": 18.0268,
+      "step": 12576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8521885876175558e-05,
+      "loss": 18.4269,
+      "step": 12577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.851463059748349e-05,
+      "loss": 20.2141,
+      "step": 12578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8507376417181378e-05,
+      "loss": 0.7014,
+      "step": 12579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.850012333552229e-05,
+      "loss": 19.511,
+      "step": 12580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8492871352759245e-05,
+      "loss": 20.2589,
+      "step": 12581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.848562046914526e-05,
+      "loss": 0.6226,
+      "step": 12582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8478370684933278e-05,
+      "loss": 0.6738,
+      "step": 12583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8471122000376205e-05,
+      "loss": 0.6628,
+      "step": 12584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8463874415726918e-05,
+      "loss": 0.6571,
+      "step": 12585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8456627931238247e-05,
+      "loss": 0.6599,
+      "step": 12586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8449382547163007e-05,
+      "loss": 0.6335,
+      "step": 12587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8442138263753962e-05,
+      "loss": 0.6382,
+      "step": 12588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.843489508126382e-05,
+      "loss": 0.6624,
+      "step": 12589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8427652999945266e-05,
+      "loss": 0.6386,
+      "step": 12590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.842041202005093e-05,
+      "loss": 0.6632,
+      "step": 12591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8413172141833463e-05,
+      "loss": 0.6517,
+      "step": 12592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8405933365545396e-05,
+      "loss": 16.4213,
+      "step": 12593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8398695691439277e-05,
+      "loss": 18.8437,
+      "step": 12594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.839145911976759e-05,
+      "loss": 17.7808,
+      "step": 12595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.838422365078277e-05,
+      "loss": 17.9487,
+      "step": 12596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8376989284737268e-05,
+      "loss": 16.117,
+      "step": 12597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8369756021883443e-05,
+      "loss": 0.6247,
+      "step": 12598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8362523862473628e-05,
+      "loss": 0.6455,
+      "step": 12599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8355292806760134e-05,
+      "loss": 0.6626,
+      "step": 12600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.834806285499519e-05,
+      "loss": 0.6803,
+      "step": 12601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8340834007431063e-05,
+      "loss": 0.662,
+      "step": 12602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8333606264319914e-05,
+      "loss": 17.5977,
+      "step": 12603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8326379625913886e-05,
+      "loss": 0.6973,
+      "step": 12604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.831915409246509e-05,
+      "loss": 19.0647,
+      "step": 12605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.831192966422558e-05,
+      "loss": 16.1791,
+      "step": 12606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8304706341447415e-05,
+      "loss": 0.6822,
+      "step": 12607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.829748412438257e-05,
+      "loss": 17.8346,
+      "step": 12608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8290263013282993e-05,
+      "loss": 18.6644,
+      "step": 12609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8283043008400604e-05,
+      "loss": 0.72,
+      "step": 12610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8275824109987256e-05,
+      "loss": 16.8145,
+      "step": 12611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8268606318294822e-05,
+      "loss": 18.7185,
+      "step": 12612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8261389633575082e-05,
+      "loss": 0.6537,
+      "step": 12613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.82541740560798e-05,
+      "loss": 18.8525,
+      "step": 12614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.824695958606069e-05,
+      "loss": 17.2544,
+      "step": 12615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.823974622376942e-05,
+      "loss": 17.8931,
+      "step": 12616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8232533969457667e-05,
+      "loss": 18.8357,
+      "step": 12617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8225322823377035e-05,
+      "loss": 16.5405,
+      "step": 12618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8218112785779055e-05,
+      "loss": 18.5714,
+      "step": 12619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8210903856915264e-05,
+      "loss": 0.6883,
+      "step": 12620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8203696037037172e-05,
+      "loss": 0.6034,
+      "step": 12621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8196489326396222e-05,
+      "loss": 18.3932,
+      "step": 12622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.818928372524381e-05,
+      "loss": 17.3247,
+      "step": 12623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.818207923383133e-05,
+      "loss": 17.0872,
+      "step": 12624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8174875852410085e-05,
+      "loss": 18.7868,
+      "step": 12625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8167673581231405e-05,
+      "loss": 0.6746,
+      "step": 12626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8160472420546525e-05,
+      "loss": 18.7111,
+      "step": 12627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8153272370606677e-05,
+      "loss": 18.1557,
+      "step": 12628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8146073431663035e-05,
+      "loss": 17.817,
+      "step": 12629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8138875603966714e-05,
+      "loss": 20.3417,
+      "step": 12630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.813167888776886e-05,
+      "loss": 16.0017,
+      "step": 12631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.812448328332051e-05,
+      "loss": 0.6894,
+      "step": 12632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8117288790872688e-05,
+      "loss": 16.0995,
+      "step": 12633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8110095410676387e-05,
+      "loss": 17.5574,
+      "step": 12634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.810290314298253e-05,
+      "loss": 17.2469,
+      "step": 12635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8095711988042063e-05,
+      "loss": 16.4842,
+      "step": 12636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8088521946105836e-05,
+      "loss": 16.2663,
+      "step": 12637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8081333017424673e-05,
+      "loss": 16.0054,
+      "step": 12638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8074145202249377e-05,
+      "loss": 17.9526,
+      "step": 12639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8066958500830666e-05,
+      "loss": 17.4545,
+      "step": 12640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8059772913419303e-05,
+      "loss": 15.6173,
+      "step": 12641
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8052588440265938e-05,
+      "loss": 0.6398,
+      "step": 12642
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8045405081621215e-05,
+      "loss": 17.5015,
+      "step": 12643
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8038222837735713e-05,
+      "loss": 17.4672,
+      "step": 12644
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8031041708859992e-05,
+      "loss": 17.555,
+      "step": 12645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8023861695244594e-05,
+      "loss": 18.2908,
+      "step": 12646
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8016682797139983e-05,
+      "loss": 0.6448,
+      "step": 12647
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8009505014796603e-05,
+      "loss": 19.5839,
+      "step": 12648
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8002328348464854e-05,
+      "loss": 0.669,
+      "step": 12649
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.799515279839508e-05,
+      "loss": 17.4656,
+      "step": 12650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7987978364837647e-05,
+      "loss": 17.3074,
+      "step": 12651
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.798080504804282e-05,
+      "loss": 0.6953,
+      "step": 12652
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.797363284826084e-05,
+      "loss": 0.6351,
+      "step": 12653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.796646176574192e-05,
+      "loss": 17.6692,
+      "step": 12654
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.795929180073621e-05,
+      "loss": 0.6029,
+      "step": 12655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7952122953493876e-05,
+      "loss": 0.6739,
+      "step": 12656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7944955224264985e-05,
+      "loss": 0.6789,
+      "step": 12657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7937788613299593e-05,
+      "loss": 0.7252,
+      "step": 12658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7930623120847712e-05,
+      "loss": 18.9595,
+      "step": 12659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.792345874715931e-05,
+      "loss": 17.7765,
+      "step": 12660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7916295492484315e-05,
+      "loss": 17.0281,
+      "step": 12661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.790913335707265e-05,
+      "loss": 0.6375,
+      "step": 12662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7901972341174157e-05,
+      "loss": 16.8913,
+      "step": 12663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7894812445038645e-05,
+      "loss": 19.2589,
+      "step": 12664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7887653668915897e-05,
+      "loss": 0.6577,
+      "step": 12665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7880496013055642e-05,
+      "loss": 18.7561,
+      "step": 12666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7873339477707613e-05,
+      "loss": 17.0181,
+      "step": 12667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.786618406312144e-05,
+      "loss": 20.0057,
+      "step": 12668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.785902976954676e-05,
+      "loss": 17.5248,
+      "step": 12669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7851876597233148e-05,
+      "loss": 0.6353,
+      "step": 12670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.784472454643013e-05,
+      "loss": 17.1812,
+      "step": 12671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7837573617387265e-05,
+      "loss": 18.0513,
+      "step": 12672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.783042381035397e-05,
+      "loss": 0.7127,
+      "step": 12673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.782327512557968e-05,
+      "loss": 17.2781,
+      "step": 12674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7816127563313774e-05,
+      "loss": 0.6871,
+      "step": 12675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.780898112380563e-05,
+      "loss": 0.6783,
+      "step": 12676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7801835807304536e-05,
+      "loss": 18.3534,
+      "step": 12677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.779469161405976e-05,
+      "loss": 18.4951,
+      "step": 12678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7787548544320543e-05,
+      "loss": 0.6528,
+      "step": 12679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7780406598336054e-05,
+      "loss": 17.4592,
+      "step": 12680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.777326577635548e-05,
+      "loss": 0.6807,
+      "step": 12681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.776612607862791e-05,
+      "loss": 18.8817,
+      "step": 12682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7758987505402423e-05,
+      "loss": 18.61,
+      "step": 12683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.775185005692806e-05,
+      "loss": 0.6377,
+      "step": 12684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7744713733453783e-05,
+      "loss": 17.2262,
+      "step": 12685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7737578535228594e-05,
+      "loss": 16.1618,
+      "step": 12686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7730444462501395e-05,
+      "loss": 17.0845,
+      "step": 12687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.772331151552105e-05,
+      "loss": 17.1555,
+      "step": 12688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7716179694536406e-05,
+      "loss": 0.6962,
+      "step": 12689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7709048999796245e-05,
+      "loss": 18.1924,
+      "step": 12690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7701919431549357e-05,
+      "loss": 17.5743,
+      "step": 12691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7694790990044453e-05,
+      "loss": 0.7296,
+      "step": 12692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7687663675530204e-05,
+      "loss": 18.4239,
+      "step": 12693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7680537488255255e-05,
+      "loss": 0.6196,
+      "step": 12694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.767341242846819e-05,
+      "loss": 14.9455,
+      "step": 12695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.766628849641761e-05,
+      "loss": 0.6746,
+      "step": 12696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.765916569235202e-05,
+      "loss": 16.8876,
+      "step": 12697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7652044016519897e-05,
+      "loss": 17.4338,
+      "step": 12698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.76449234691697e-05,
+      "loss": 18.2568,
+      "step": 12699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7637804050549805e-05,
+      "loss": 19.7185,
+      "step": 12700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7630685760908622e-05,
+      "loss": 16.7962,
+      "step": 12701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.762356860049445e-05,
+      "loss": 0.6685,
+      "step": 12702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.761645256955558e-05,
+      "loss": 0.6896,
+      "step": 12703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7609337668340263e-05,
+      "loss": 17.4208,
+      "step": 12704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7602223897096693e-05,
+      "loss": 18.0831,
+      "step": 12705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.759511125607307e-05,
+      "loss": 0.6965,
+      "step": 12706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7587999745517504e-05,
+      "loss": 18.3591,
+      "step": 12707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7580889365678082e-05,
+      "loss": 16.4471,
+      "step": 12708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.757378011680286e-05,
+      "loss": 17.4857,
+      "step": 12709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7566671999139837e-05,
+      "loss": 16.6137,
+      "step": 12710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7559565012937017e-05,
+      "loss": 0.6661,
+      "step": 12711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7552459158442307e-05,
+      "loss": 18.1585,
+      "step": 12712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7545354435903606e-05,
+      "loss": 0.655,
+      "step": 12713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7538250845568767e-05,
+      "loss": 16.1753,
+      "step": 12714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7531148387685588e-05,
+      "loss": 0.6165,
+      "step": 12715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.752404706250187e-05,
+      "loss": 0.6793,
+      "step": 12716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7516946870265344e-05,
+      "loss": 17.4579,
+      "step": 12717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7509847811223696e-05,
+      "loss": 17.1779,
+      "step": 12718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.750274988562458e-05,
+      "loss": 0.6513,
+      "step": 12719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7495653093715604e-05,
+      "loss": 0.7214,
+      "step": 12720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.748855743574437e-05,
+      "loss": 0.6922,
+      "step": 12721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.74814629119584e-05,
+      "loss": 0.6998,
+      "step": 12722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.74743695226052e-05,
+      "loss": 0.6462,
+      "step": 12723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.746727726793222e-05,
+      "loss": 16.8657,
+      "step": 12724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7460186148186863e-05,
+      "loss": 0.6847,
+      "step": 12725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.745309616361654e-05,
+      "loss": 18.3354,
+      "step": 12726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7446007314468593e-05,
+      "loss": 15.5635,
+      "step": 12727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7438919600990288e-05,
+      "loss": 18.2428,
+      "step": 12728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7431833023428886e-05,
+      "loss": 0.671,
+      "step": 12729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7424747582031637e-05,
+      "loss": 0.7054,
+      "step": 12730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.741766327704571e-05,
+      "loss": 0.6387,
+      "step": 12731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7410580108718238e-05,
+      "loss": 18.4234,
+      "step": 12732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7403498077296332e-05,
+      "loss": 0.6425,
+      "step": 12733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.739641718302703e-05,
+      "loss": 16.8492,
+      "step": 12734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.738933742615739e-05,
+      "loss": 0.6478,
+      "step": 12735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.738225880693438e-05,
+      "loss": 0.6875,
+      "step": 12736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7375181325604944e-05,
+      "loss": 17.2839,
+      "step": 12737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7368104982415976e-05,
+      "loss": 17.6193,
+      "step": 12738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7361029777614336e-05,
+      "loss": 17.5129,
+      "step": 12739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7353955711446866e-05,
+      "loss": 18.8789,
+      "step": 12740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7346882784160347e-05,
+      "loss": 18.1022,
+      "step": 12741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7339810996001516e-05,
+      "loss": 18.9509,
+      "step": 12742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7332740347217074e-05,
+      "loss": 17.692,
+      "step": 12743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.732567083805368e-05,
+      "loss": 18.44,
+      "step": 12744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7318602468757983e-05,
+      "loss": 0.6558,
+      "step": 12745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7311535239576555e-05,
+      "loss": 0.6367,
+      "step": 12746
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7304469150755936e-05,
+      "loss": 17.1723,
+      "step": 12747
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7297404202542643e-05,
+      "loss": 0.6355,
+      "step": 12748
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.729034039518312e-05,
+      "loss": 0.7007,
+      "step": 12749
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.728327772892382e-05,
+      "loss": 17.2705,
+      "step": 12750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.727621620401112e-05,
+      "loss": 0.7307,
+      "step": 12751
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7269155820691362e-05,
+      "loss": 0.6669,
+      "step": 12752
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7262096579210858e-05,
+      "loss": 0.6283,
+      "step": 12753
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.725503847981585e-05,
+      "loss": 16.1719,
+      "step": 12754
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7247981522752604e-05,
+      "loss": 18.7386,
+      "step": 12755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.724092570826729e-05,
+      "loss": 17.2586,
+      "step": 12756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.723387103660606e-05,
+      "loss": 17.7791,
+      "step": 12757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.722681750801501e-05,
+      "loss": 17.2124,
+      "step": 12758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.72197651227402e-05,
+      "loss": 0.6686,
+      "step": 12759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7212713881027687e-05,
+      "loss": 17.2023,
+      "step": 12760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7205663783123436e-05,
+      "loss": 0.647,
+      "step": 12761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7198614829273414e-05,
+      "loss": 0.7028,
+      "step": 12762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7191567019723507e-05,
+      "loss": 17.3781,
+      "step": 12763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.718452035471958e-05,
+      "loss": 17.296,
+      "step": 12764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717747483450749e-05,
+      "loss": 18.3101,
+      "step": 12765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7170430459333004e-05,
+      "loss": 0.6651,
+      "step": 12766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716338722944188e-05,
+      "loss": 0.6755,
+      "step": 12767
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7156345145079817e-05,
+      "loss": 0.6343,
+      "step": 12768
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7149304206492473e-05,
+      "loss": 18.5398,
+      "step": 12769
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.71422644139255e-05,
+      "loss": 0.6225,
+      "step": 12770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7135225767624486e-05,
+      "loss": 0.6802,
+      "step": 12771
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7128188267834966e-05,
+      "loss": 18.6438,
+      "step": 12772
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712115191480245e-05,
+      "loss": 0.6557,
+      "step": 12773
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7114116708772393e-05,
+      "loss": 18.3513,
+      "step": 12774
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.710708264999026e-05,
+      "loss": 0.6532,
+      "step": 12775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7100049738701407e-05,
+      "loss": 17.4651,
+      "step": 12776
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7093017975151204e-05,
+      "loss": 0.6874,
+      "step": 12777
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7085987359584944e-05,
+      "loss": 0.6629,
+      "step": 12778
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7078957892247884e-05,
+      "loss": 18.7906,
+      "step": 12779
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.707192957338528e-05,
+      "loss": 17.0754,
+      "step": 12780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7064902403242312e-05,
+      "loss": 0.666,
+      "step": 12781
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7057876382064137e-05,
+      "loss": 0.6206,
+      "step": 12782
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7050851510095817e-05,
+      "loss": 18.4083,
+      "step": 12783
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7043827787582468e-05,
+      "loss": 16.9418,
+      "step": 12784
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7036805214769103e-05,
+      "loss": 16.799,
+      "step": 12785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7029783791900705e-05,
+      "loss": 17.6917,
+      "step": 12786
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.702276351922223e-05,
+      "loss": 17.2203,
+      "step": 12787
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7015744396978556e-05,
+      "loss": 0.6901,
+      "step": 12788
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7008726425414596e-05,
+      "loss": 15.975,
+      "step": 12789
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7001709604775153e-05,
+      "loss": 18.4704,
+      "step": 12790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6994693935305018e-05,
+      "loss": 19.0377,
+      "step": 12791
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6987679417248937e-05,
+      "loss": 17.204,
+      "step": 12792
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6980666050851594e-05,
+      "loss": 20.3671,
+      "step": 12793
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6973653836357694e-05,
+      "loss": 16.0744,
+      "step": 12794
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.696664277401185e-05,
+      "loss": 0.6776,
+      "step": 12795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.695963286405864e-05,
+      "loss": 0.6677,
+      "step": 12796
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6952624106742614e-05,
+      "loss": 0.6578,
+      "step": 12797
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6945616502308264e-05,
+      "loss": 0.6674,
+      "step": 12798
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6938610051000082e-05,
+      "loss": 18.1261,
+      "step": 12799
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.693160475306248e-05,
+      "loss": 0.6482,
+      "step": 12800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6924600608739842e-05,
+      "loss": 18.5137,
+      "step": 12801
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6917597618276515e-05,
+      "loss": 19.2173,
+      "step": 12802
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6910595781916788e-05,
+      "loss": 17.0447,
+      "step": 12803
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.690359509990495e-05,
+      "loss": 20.0427,
+      "step": 12804
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.689659557248521e-05,
+      "loss": 17.405,
+      "step": 12805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6889597199901768e-05,
+      "loss": 0.7057,
+      "step": 12806
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6882599982398743e-05,
+      "loss": 0.6065,
+      "step": 12807
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.687560392022024e-05,
+      "loss": 0.651,
+      "step": 12808
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6868609013610347e-05,
+      "loss": 16.608,
+      "step": 12809
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.686161526281307e-05,
+      "loss": 16.5172,
+      "step": 12810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6854622668072388e-05,
+      "loss": 17.8706,
+      "step": 12811
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.684763122963225e-05,
+      "loss": 0.6468,
+      "step": 12812
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6840640947736537e-05,
+      "loss": 16.6634,
+      "step": 12813
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6833651822629142e-05,
+      "loss": 18.5746,
+      "step": 12814
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6826663854553876e-05,
+      "loss": 16.4177,
+      "step": 12815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6819677043754505e-05,
+      "loss": 17.636,
+      "step": 12816
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6812691390474787e-05,
+      "loss": 0.6508,
+      "step": 12817
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.680570689495839e-05,
+      "loss": 17.5797,
+      "step": 12818
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6798723557449013e-05,
+      "loss": 16.5788,
+      "step": 12819
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6791741378190263e-05,
+      "loss": 0.5903,
+      "step": 12820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6784760357425704e-05,
+      "loss": 17.6745,
+      "step": 12821
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6777780495398894e-05,
+      "loss": 16.9719,
+      "step": 12822
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6770801792353293e-05,
+      "loss": 0.6099,
+      "step": 12823
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6763824248532405e-05,
+      "loss": 18.6,
+      "step": 12824
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6756847864179625e-05,
+      "loss": 16.4018,
+      "step": 12825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6749872639538333e-05,
+      "loss": 0.6557,
+      "step": 12826
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6742898574851862e-05,
+      "loss": 0.6627,
+      "step": 12827
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.673592567036349e-05,
+      "loss": 0.6536,
+      "step": 12828
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6728953926316505e-05,
+      "loss": 0.7122,
+      "step": 12829
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6721983342954106e-05,
+      "loss": 16.7718,
+      "step": 12830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6715013920519466e-05,
+      "loss": 18.0371,
+      "step": 12831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.670804565925572e-05,
+      "loss": 0.6359,
+      "step": 12832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6701078559405947e-05,
+      "loss": 0.7251,
+      "step": 12833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6694112621213224e-05,
+      "loss": 16.0467,
+      "step": 12834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6687147844920552e-05,
+      "loss": 15.3677,
+      "step": 12835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6680184230770917e-05,
+      "loss": 0.6265,
+      "step": 12836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.66732217790072e-05,
+      "loss": 0.6524,
+      "step": 12837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6666260489872348e-05,
+      "loss": 0.6313,
+      "step": 12838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6659300363609182e-05,
+      "loss": 0.7048,
+      "step": 12839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.665234140046052e-05,
+      "loss": 18.228,
+      "step": 12840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6645383600669124e-05,
+      "loss": 0.7022,
+      "step": 12841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.663842696447771e-05,
+      "loss": 0.6779,
+      "step": 12842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6631471492128997e-05,
+      "loss": 17.7218,
+      "step": 12843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6624517183865618e-05,
+      "loss": 0.6742,
+      "step": 12844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6617564039930173e-05,
+      "loss": 16.7919,
+      "step": 12845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6610612060565234e-05,
+      "loss": 18.5662,
+      "step": 12846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.66036612460133e-05,
+      "loss": 16.4916,
+      "step": 12847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6596711596516902e-05,
+      "loss": 17.3648,
+      "step": 12848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6589763112318453e-05,
+      "loss": 16.7638,
+      "step": 12849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6582815793660365e-05,
+      "loss": 17.4378,
+      "step": 12850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6575869640784995e-05,
+      "loss": 16.1591,
+      "step": 12851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.656892465393466e-05,
+      "loss": 17.9016,
+      "step": 12852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6561980833351654e-05,
+      "loss": 0.6168,
+      "step": 12853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6555038179278213e-05,
+      "loss": 17.3876,
+      "step": 12854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.654809669195654e-05,
+      "loss": 0.6386,
+      "step": 12855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6541156371628787e-05,
+      "loss": 18.6153,
+      "step": 12856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6534217218537067e-05,
+      "loss": 18.4,
+      "step": 12857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6527279232923455e-05,
+      "loss": 0.6152,
+      "step": 12858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.652034241503001e-05,
+      "loss": 0.7104,
+      "step": 12859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6513406765098715e-05,
+      "loss": 17.4247,
+      "step": 12860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6506472283371527e-05,
+      "loss": 18.1894,
+      "step": 12861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6499538970090362e-05,
+      "loss": 0.7128,
+      "step": 12862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6492606825497065e-05,
+      "loss": 18.7176,
+      "step": 12863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6485675849833515e-05,
+      "loss": 0.6951,
+      "step": 12864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.647874604334148e-05,
+      "loss": 0.7093,
+      "step": 12865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6471817406262718e-05,
+      "loss": 16.6709,
+      "step": 12866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6464889938838934e-05,
+      "loss": 19.1835,
+      "step": 12867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6457963641311785e-05,
+      "loss": 17.7111,
+      "step": 12868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.645103851392293e-05,
+      "loss": 0.6586,
+      "step": 12869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6444114556913936e-05,
+      "loss": 16.6508,
+      "step": 12870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.643719177052636e-05,
+      "loss": 18.0478,
+      "step": 12871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.643027015500171e-05,
+      "loss": 0.6422,
+      "step": 12872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.642334971058142e-05,
+      "loss": 17.6628,
+      "step": 12873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6416430437506964e-05,
+      "loss": 0.666,
+      "step": 12874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.64095123360197e-05,
+      "loss": 18.5684,
+      "step": 12875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6402595406360977e-05,
+      "loss": 16.6698,
+      "step": 12876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6395679648772088e-05,
+      "loss": 17.6218,
+      "step": 12877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6388765063494283e-05,
+      "loss": 0.6227,
+      "step": 12878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.638185165076882e-05,
+      "loss": 0.6811,
+      "step": 12879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6374939410836853e-05,
+      "loss": 17.6357,
+      "step": 12880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.636802834393953e-05,
+      "loss": 0.6203,
+      "step": 12881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6361118450317942e-05,
+      "loss": 16.3351,
+      "step": 12882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6354209730213126e-05,
+      "loss": 17.9031,
+      "step": 12883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6347302183866144e-05,
+      "loss": 18.5179,
+      "step": 12884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.634039581151794e-05,
+      "loss": 18.118,
+      "step": 12885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.633349061340945e-05,
+      "loss": 0.6661,
+      "step": 12886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.632658658978158e-05,
+      "loss": 0.6836,
+      "step": 12887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6319683740875147e-05,
+      "loss": 0.7092,
+      "step": 12888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.631278206693101e-05,
+      "loss": 0.6536,
+      "step": 12889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.630588156818991e-05,
+      "loss": 0.6859,
+      "step": 12890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.629898224489261e-05,
+      "loss": 17.2991,
+      "step": 12891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6292084097279724e-05,
+      "loss": 17.0362,
+      "step": 12892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6285187125591966e-05,
+      "loss": 16.2259,
+      "step": 12893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6278291330069924e-05,
+      "loss": 18.1862,
+      "step": 12894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6271396710954157e-05,
+      "loss": 20.3642,
+      "step": 12895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6264503268485187e-05,
+      "loss": 17.9354,
+      "step": 12896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.625761100290349e-05,
+      "loss": 16.2354,
+      "step": 12897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.625071991444953e-05,
+      "loss": 18.5856,
+      "step": 12898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6243830003363704e-05,
+      "loss": 18.0805,
+      "step": 12899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6236941269886356e-05,
+      "loss": 18.3701,
+      "step": 12900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.623005371425782e-05,
+      "loss": 0.6802,
+      "step": 12901
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.622316733671834e-05,
+      "loss": 0.6452,
+      "step": 12902
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6216282137508203e-05,
+      "loss": 19.9018,
+      "step": 12903
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6209398116867574e-05,
+      "loss": 0.6803,
+      "step": 12904
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6202515275036616e-05,
+      "loss": 17.5238,
+      "step": 12905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6195633612255433e-05,
+      "loss": 0.6914,
+      "step": 12906
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.618875312876409e-05,
+      "loss": 0.6747,
+      "step": 12907
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6181873824802647e-05,
+      "loss": 16.1105,
+      "step": 12908
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6174995700611074e-05,
+      "loss": 17.3571,
+      "step": 12909
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6168118756429317e-05,
+      "loss": 16.9067,
+      "step": 12910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6161242992497295e-05,
+      "loss": 19.091,
+      "step": 12911
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6154368409054844e-05,
+      "loss": 16.7395,
+      "step": 12912
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6147495006341834e-05,
+      "loss": 17.9492,
+      "step": 12913
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.614062278459802e-05,
+      "loss": 18.4625,
+      "step": 12914
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6133751744063154e-05,
+      "loss": 0.6342,
+      "step": 12915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6126881884976934e-05,
+      "loss": 0.6605,
+      "step": 12916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6120013207579004e-05,
+      "loss": 16.8914,
+      "step": 12917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.611314571210901e-05,
+      "loss": 19.3698,
+      "step": 12918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.610627939880652e-05,
+      "loss": 0.6351,
+      "step": 12919
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.609941426791108e-05,
+      "loss": 17.3076,
+      "step": 12920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.609255031966216e-05,
+      "loss": 17.7041,
+      "step": 12921
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6085687554299216e-05,
+      "loss": 0.6507,
+      "step": 12922
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6078825972061683e-05,
+      "loss": 17.2518,
+      "step": 12923
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.607196557318893e-05,
+      "loss": 0.6512,
+      "step": 12924
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6065106357920273e-05,
+      "loss": 0.6522,
+      "step": 12925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6058248326495005e-05,
+      "loss": 18.139,
+      "step": 12926
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.605139147915236e-05,
+      "loss": 0.7303,
+      "step": 12927
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6044535816131577e-05,
+      "loss": 19.9369,
+      "step": 12928
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.60376813376718e-05,
+      "loss": 17.9481,
+      "step": 12929
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6030828044012153e-05,
+      "loss": 18.5955,
+      "step": 12930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.602397593539172e-05,
+      "loss": 17.5353,
+      "step": 12931
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6017125012049526e-05,
+      "loss": 17.3328,
+      "step": 12932
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6010275274224606e-05,
+      "loss": 0.6232,
+      "step": 12933
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6003426722155894e-05,
+      "loss": 0.6651,
+      "step": 12934
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5996579356082315e-05,
+      "loss": 0.6649,
+      "step": 12935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5989733176242732e-05,
+      "loss": 18.6211,
+      "step": 12936
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5982888182875978e-05,
+      "loss": 17.5686,
+      "step": 12937
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5976044376220866e-05,
+      "loss": 17.1597,
+      "step": 12938
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.596920175651614e-05,
+      "loss": 17.1988,
+      "step": 12939
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5962360324000503e-05,
+      "loss": 0.6561,
+      "step": 12940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5955520078912626e-05,
+      "loss": 17.0532,
+      "step": 12941
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.594868102149112e-05,
+      "loss": 0.6666,
+      "step": 12942
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5941843151974605e-05,
+      "loss": 0.6294,
+      "step": 12943
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5935006470601604e-05,
+      "loss": 17.7203,
+      "step": 12944
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5928170977610634e-05,
+      "loss": 0.653,
+      "step": 12945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5921336673240118e-05,
+      "loss": 19.9679,
+      "step": 12946
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5914503557728515e-05,
+      "loss": 15.9221,
+      "step": 12947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5907671631314192e-05,
+      "loss": 17.1823,
+      "step": 12948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.590084089423548e-05,
+      "loss": 0.6469,
+      "step": 12949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.589401134673068e-05,
+      "loss": 0.6064,
+      "step": 12950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.588718298903803e-05,
+      "loss": 18.5082,
+      "step": 12951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5880355821395765e-05,
+      "loss": 20.7502,
+      "step": 12952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5873529844042053e-05,
+      "loss": 0.6795,
+      "step": 12953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5866705057215014e-05,
+      "loss": 18.7757,
+      "step": 12954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5859881461152738e-05,
+      "loss": 0.673,
+      "step": 12955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5853059056093256e-05,
+      "loss": 17.9519,
+      "step": 12956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5846237842274603e-05,
+      "loss": 17.4623,
+      "step": 12957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5839417819934726e-05,
+      "loss": 19.0357,
+      "step": 12958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5832598989311554e-05,
+      "loss": 0.6832,
+      "step": 12959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5825781350642954e-05,
+      "loss": 17.9782,
+      "step": 12960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5818964904166756e-05,
+      "loss": 0.6744,
+      "step": 12961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5812149650120784e-05,
+      "loss": 17.7514,
+      "step": 12962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5805335588742788e-05,
+      "loss": 0.6846,
+      "step": 12963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5798522720270463e-05,
+      "loss": 0.6483,
+      "step": 12964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5791711044941498e-05,
+      "loss": 0.6519,
+      "step": 12965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5784900562993505e-05,
+      "loss": 0.6858,
+      "step": 12966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5778091274664093e-05,
+      "loss": 0.6654,
+      "step": 12967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5771283180190806e-05,
+      "loss": 18.4564,
+      "step": 12968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5764476279811136e-05,
+      "loss": 15.6523,
+      "step": 12969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5757670573762562e-05,
+      "loss": 0.6788,
+      "step": 12970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5750866062282477e-05,
+      "loss": 0.7096,
+      "step": 12971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.57440627456083e-05,
+      "loss": 0.7153,
+      "step": 12972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5737260623977358e-05,
+      "loss": 18.2183,
+      "step": 12973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5730459697626936e-05,
+      "loss": 17.7414,
+      "step": 12974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5723659966794296e-05,
+      "loss": 0.7026,
+      "step": 12975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5716861431716635e-05,
+      "loss": 0.6731,
+      "step": 12976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.571006409263116e-05,
+      "loss": 17.7599,
+      "step": 12977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.570326794977498e-05,
+      "loss": 17.706,
+      "step": 12978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.569647300338518e-05,
+      "loss": 17.9494,
+      "step": 12979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5689679253698818e-05,
+      "loss": 19.1818,
+      "step": 12980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.568288670095288e-05,
+      "loss": 18.3161,
+      "step": 12981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5676095345384356e-05,
+      "loss": 16.3315,
+      "step": 12982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5669305187230148e-05,
+      "loss": 0.7241,
+      "step": 12983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5662516226727147e-05,
+      "loss": 16.894,
+      "step": 12984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.565572846411218e-05,
+      "loss": 17.5195,
+      "step": 12985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.564894189962204e-05,
+      "loss": 0.6806,
+      "step": 12986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5642156533493497e-05,
+      "loss": 18.8156,
+      "step": 12987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5635372365963258e-05,
+      "loss": 0.6986,
+      "step": 12988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.562858939726799e-05,
+      "loss": 0.6749,
+      "step": 12989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5621807627644325e-05,
+      "loss": 0.6909,
+      "step": 12990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.561502705732883e-05,
+      "loss": 0.6502,
+      "step": 12991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5608247686558082e-05,
+      "loss": 0.6989,
+      "step": 12992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5601469515568574e-05,
+      "loss": 15.2274,
+      "step": 12993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.559469254459676e-05,
+      "loss": 0.6411,
+      "step": 12994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.558791677387907e-05,
+      "loss": 0.6514,
+      "step": 12995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.558114220365185e-05,
+      "loss": 15.9917,
+      "step": 12996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.557436883415148e-05,
+      "loss": 0.6856,
+      "step": 12997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5567596665614237e-05,
+      "loss": 0.6698,
+      "step": 12998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.556082569827637e-05,
+      "loss": 17.4812,
+      "step": 12999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.555405593237409e-05,
+      "loss": 0.6648,
+      "step": 13000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.554728736814356e-05,
+      "loss": 18.0728,
+      "step": 13001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5540520005820915e-05,
+      "loss": 0.7015,
+      "step": 13002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5533753845642236e-05,
+      "loss": 18.2344,
+      "step": 13003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5526988887843563e-05,
+      "loss": 18.3935,
+      "step": 13004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5520225132660883e-05,
+      "loss": 0.6748,
+      "step": 13005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5513462580330186e-05,
+      "loss": 0.6697,
+      "step": 13006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5506701231087374e-05,
+      "loss": 0.6722,
+      "step": 13007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.549994108516832e-05,
+      "loss": 0.6449,
+      "step": 13008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5493182142808854e-05,
+      "loss": 0.6906,
+      "step": 13009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5486424404244753e-05,
+      "loss": 0.6811,
+      "step": 13010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.54796678697118e-05,
+      "loss": 16.5082,
+      "step": 13011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5472912539445684e-05,
+      "loss": 17.7692,
+      "step": 13012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5466158413682065e-05,
+      "loss": 16.9688,
+      "step": 13013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5459405492656575e-05,
+      "loss": 18.9042,
+      "step": 13014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.545265377660477e-05,
+      "loss": 0.6906,
+      "step": 13015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5445903265762224e-05,
+      "loss": 17.3002,
+      "step": 13016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5439153960364415e-05,
+      "loss": 16.2022,
+      "step": 13017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5432405860646805e-05,
+      "loss": 16.9304,
+      "step": 13018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5425658966844798e-05,
+      "loss": 0.6799,
+      "step": 13019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5418913279193746e-05,
+      "loss": 17.0173,
+      "step": 13020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.541216879792902e-05,
+      "loss": 19.4488,
+      "step": 13021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5405425523285884e-05,
+      "loss": 0.7025,
+      "step": 13022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5398683455499586e-05,
+      "loss": 20.0415,
+      "step": 13023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.539194259480532e-05,
+      "loss": 18.6399,
+      "step": 13024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.538520294143823e-05,
+      "loss": 0.6831,
+      "step": 13025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5378464495633475e-05,
+      "loss": 17.8005,
+      "step": 13026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5371727257626112e-05,
+      "loss": 17.2469,
+      "step": 13027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5364991227651172e-05,
+      "loss": 0.6986,
+      "step": 13028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5358256405943638e-05,
+      "loss": 18.0691,
+      "step": 13029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.535152279273846e-05,
+      "loss": 18.6892,
+      "step": 13030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5344790388270564e-05,
+      "loss": 0.5956,
+      "step": 13031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5338059192774805e-05,
+      "loss": 17.1122,
+      "step": 13032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5331329206486005e-05,
+      "loss": 17.6892,
+      "step": 13033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5324600429638942e-05,
+      "loss": 16.387,
+      "step": 13034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5317872862468342e-05,
+      "loss": 0.7275,
+      "step": 13035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5311146505208924e-05,
+      "loss": 16.6664,
+      "step": 13036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5304421358095338e-05,
+      "loss": 17.9426,
+      "step": 13037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.529769742136219e-05,
+      "loss": 0.68,
+      "step": 13038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.529097469524405e-05,
+      "loss": 17.9444,
+      "step": 13039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.528425317997544e-05,
+      "loss": 18.3135,
+      "step": 13040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.527753287579084e-05,
+      "loss": 0.6711,
+      "step": 13041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5270813782924715e-05,
+      "loss": 18.8615,
+      "step": 13042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5264095901611452e-05,
+      "loss": 0.6688,
+      "step": 13043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5257379232085411e-05,
+      "loss": 18.263,
+      "step": 13044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5250663774580903e-05,
+      "loss": 0.5929,
+      "step": 13045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5243949529332197e-05,
+      "loss": 0.6444,
+      "step": 13046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5237236496573542e-05,
+      "loss": 0.7197,
+      "step": 13047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5230524676539121e-05,
+      "loss": 19.1592,
+      "step": 13048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5223814069463078e-05,
+      "loss": 17.4152,
+      "step": 13049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.521710467557952e-05,
+      "loss": 0.6601,
+      "step": 13050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.521039649512248e-05,
+      "loss": 19.0924,
+      "step": 13051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5203689528326026e-05,
+      "loss": 16.158,
+      "step": 13052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5196983775424117e-05,
+      "loss": 0.642,
+      "step": 13053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5190279236650679e-05,
+      "loss": 0.6928,
+      "step": 13054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5183575912239612e-05,
+      "loss": 17.3383,
+      "step": 13055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5176873802424763e-05,
+      "loss": 17.8153,
+      "step": 13056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.517017290743994e-05,
+      "loss": 0.6754,
+      "step": 13057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.516347322751891e-05,
+      "loss": 0.7012,
+      "step": 13058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5156774762895399e-05,
+      "loss": 0.687,
+      "step": 13059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5150077513803069e-05,
+      "loss": 17.3268,
+      "step": 13060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5143381480475583e-05,
+      "loss": 17.9045,
+      "step": 13061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5136686663146538e-05,
+      "loss": 0.6553,
+      "step": 13062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.512999306204947e-05,
+      "loss": 17.909,
+      "step": 13063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5123300677417901e-05,
+      "loss": 16.2955,
+      "step": 13064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.511660950948528e-05,
+      "loss": 17.659,
+      "step": 13065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5109919558485064e-05,
+      "loss": 0.6916,
+      "step": 13066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5103230824650627e-05,
+      "loss": 17.9413,
+      "step": 13067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5096543308215306e-05,
+      "loss": 0.682,
+      "step": 13068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5089857009412394e-05,
+      "loss": 18.1071,
+      "step": 13069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5083171928475142e-05,
+      "loss": 17.8238,
+      "step": 13070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5076488065636784e-05,
+      "loss": 16.2196,
+      "step": 13071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.506980542113049e-05,
+      "loss": 0.7152,
+      "step": 13072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5063123995189382e-05,
+      "loss": 18.566,
+      "step": 13073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5056443788046543e-05,
+      "loss": 16.8466,
+      "step": 13074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5049764799935001e-05,
+      "loss": 18.3941,
+      "step": 13075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5043087031087794e-05,
+      "loss": 16.3667,
+      "step": 13076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5036410481737862e-05,
+      "loss": 0.673,
+      "step": 13077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5029735152118124e-05,
+      "loss": 0.7264,
+      "step": 13078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502306104246145e-05,
+      "loss": 16.961,
+      "step": 13079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501638815300066e-05,
+      "loss": 0.7057,
+      "step": 13080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5009716483968571e-05,
+      "loss": 0.6215,
+      "step": 13081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003046035597913e-05,
+      "loss": 18.8059,
+      "step": 13082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4996376808121387e-05,
+      "loss": 0.6293,
+      "step": 13083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989708801771657e-05,
+      "loss": 18.2054,
+      "step": 13084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983042016781329e-05,
+      "loss": 17.6577,
+      "step": 13085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4976376453383007e-05,
+      "loss": 17.8143,
+      "step": 13086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4969712111809204e-05,
+      "loss": 20.3166,
+      "step": 13087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4963048992292411e-05,
+      "loss": 17.5196,
+      "step": 13088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4956387095065083e-05,
+      "loss": 17.6101,
+      "step": 13089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4949726420359605e-05,
+      "loss": 0.6173,
+      "step": 13090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4943066968408375e-05,
+      "loss": 18.4874,
+      "step": 13091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4936408739443691e-05,
+      "loss": 18.4892,
+      "step": 13092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929751733697828e-05,
+      "loss": 0.7244,
+      "step": 13093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923095951403026e-05,
+      "loss": 18.5155,
+      "step": 13094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916441392791464e-05,
+      "loss": 17.6331,
+      "step": 13095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4909788058095315e-05,
+      "loss": 0.6047,
+      "step": 13096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4903135947546676e-05,
+      "loss": 15.9226,
+      "step": 13097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4896485061377608e-05,
+      "loss": 17.494,
+      "step": 13098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4889835399820129e-05,
+      "loss": 0.6644,
+      "step": 13099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4883186963106206e-05,
+      "loss": 17.8306,
+      "step": 13100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4876539751467806e-05,
+      "loss": 15.7424,
+      "step": 13101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.48698937651368e-05,
+      "loss": 16.1101,
+      "step": 13102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4863249004345048e-05,
+      "loss": 0.6609,
+      "step": 13103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4856605469324347e-05,
+      "loss": 0.6967,
+      "step": 13104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4849963160306451e-05,
+      "loss": 15.0317,
+      "step": 13105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4843322077523114e-05,
+      "loss": 0.6876,
+      "step": 13106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4836682221206e-05,
+      "loss": 0.7231,
+      "step": 13107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4830043591586745e-05,
+      "loss": 16.4155,
+      "step": 13108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4823406188896937e-05,
+      "loss": 17.2165,
+      "step": 13109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4816770013368131e-05,
+      "loss": 0.7222,
+      "step": 13110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4810135065231833e-05,
+      "loss": 18.0518,
+      "step": 13111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4803501344719512e-05,
+      "loss": 0.6656,
+      "step": 13112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4796868852062585e-05,
+      "loss": 0.6645,
+      "step": 13113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4790237587492417e-05,
+      "loss": 18.1732,
+      "step": 13114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.478360755124038e-05,
+      "loss": 15.2518,
+      "step": 13115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4776978743537745e-05,
+      "loss": 0.6983,
+      "step": 13116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.477035116461577e-05,
+      "loss": 18.4369,
+      "step": 13117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.476372481470566e-05,
+      "loss": 0.6728,
+      "step": 13118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4757099694038562e-05,
+      "loss": 0.6648,
+      "step": 13119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.475047580284563e-05,
+      "loss": 17.1745,
+      "step": 13120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4743853141357928e-05,
+      "loss": 17.3536,
+      "step": 13121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4737231709806498e-05,
+      "loss": 0.6697,
+      "step": 13122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4730611508422325e-05,
+      "loss": 17.998,
+      "step": 13123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4723992537436348e-05,
+      "loss": 17.7355,
+      "step": 13124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4717374797079502e-05,
+      "loss": 18.5071,
+      "step": 13125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4710758287582644e-05,
+      "loss": 16.8736,
+      "step": 13126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.470414300917659e-05,
+      "loss": 16.7006,
+      "step": 13127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4697528962092122e-05,
+      "loss": 16.4724,
+      "step": 13128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4690916146559958e-05,
+      "loss": 17.8421,
+      "step": 13129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4684304562810818e-05,
+      "loss": 17.7965,
+      "step": 13130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4677694211075344e-05,
+      "loss": 0.6421,
+      "step": 13131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4671085091584136e-05,
+      "loss": 0.6508,
+      "step": 13132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4664477204567768e-05,
+      "loss": 18.7818,
+      "step": 13133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.465787055025673e-05,
+      "loss": 18.7718,
+      "step": 13134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4651265128881547e-05,
+      "loss": 0.726,
+      "step": 13135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4644660940672627e-05,
+      "loss": 0.6741,
+      "step": 13136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.463805798586037e-05,
+      "loss": 0.6375,
+      "step": 13137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.463145626467512e-05,
+      "loss": 0.693,
+      "step": 13138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4624855777347164e-05,
+      "loss": 18.3387,
+      "step": 13139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4618256524106805e-05,
+      "loss": 0.6616,
+      "step": 13140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4611658505184239e-05,
+      "loss": 0.6896,
+      "step": 13141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4605061720809643e-05,
+      "loss": 15.8075,
+      "step": 13142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459846617121316e-05,
+      "loss": 0.6768,
+      "step": 13143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4591871856624856e-05,
+      "loss": 18.6215,
+      "step": 13144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4585278777274813e-05,
+      "loss": 16.4279,
+      "step": 13145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4578686933393021e-05,
+      "loss": 17.7489,
+      "step": 13146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4572096325209433e-05,
+      "loss": 0.5923,
+      "step": 13147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4565506952953978e-05,
+      "loss": 19.1885,
+      "step": 13148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4558918816856509e-05,
+      "loss": 0.7184,
+      "step": 13149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4552331917146893e-05,
+      "loss": 18.0152,
+      "step": 13150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4545746254054898e-05,
+      "loss": 17.1023,
+      "step": 13151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4539161827810278e-05,
+      "loss": 0.6742,
+      "step": 13152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4532578638642724e-05,
+      "loss": 19.8617,
+      "step": 13153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.452599668678189e-05,
+      "loss": 18.8178,
+      "step": 13154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4519415972457412e-05,
+      "loss": 17.0651,
+      "step": 13155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4512836495898862e-05,
+      "loss": 0.6602,
+      "step": 13156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4506258257335764e-05,
+      "loss": 16.5486,
+      "step": 13157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.44996812569976e-05,
+      "loss": 0.7264,
+      "step": 13158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4493105495113796e-05,
+      "loss": 0.6892,
+      "step": 13159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4486530971913798e-05,
+      "loss": 16.5369,
+      "step": 13160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4479957687626933e-05,
+      "loss": 20.2304,
+      "step": 13161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4473385642482517e-05,
+      "loss": 0.6342,
+      "step": 13162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4466814836709825e-05,
+      "loss": 0.6874,
+      "step": 13163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.446024527053808e-05,
+      "loss": 17.3004,
+      "step": 13164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4453676944196476e-05,
+      "loss": 18.1335,
+      "step": 13165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4447109857914143e-05,
+      "loss": 16.8339,
+      "step": 13166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4440544011920177e-05,
+      "loss": 0.6659,
+      "step": 13167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4433979406443627e-05,
+      "loss": 18.6817,
+      "step": 13168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4427416041713526e-05,
+      "loss": 0.6316,
+      "step": 13169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4420853917958832e-05,
+      "loss": 17.5408,
+      "step": 13170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4414293035408472e-05,
+      "loss": 0.6787,
+      "step": 13171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4407733394291317e-05,
+      "loss": 19.0041,
+      "step": 13172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4401174994836198e-05,
+      "loss": 16.9148,
+      "step": 13173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4394617837271934e-05,
+      "loss": 17.8787,
+      "step": 13174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4388061921827266e-05,
+      "loss": 0.629,
+      "step": 13175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4381507248730902e-05,
+      "loss": 17.8207,
+      "step": 13176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4374953818211501e-05,
+      "loss": 0.7491,
+      "step": 13177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4368401630497674e-05,
+      "loss": 17.6689,
+      "step": 13178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4361850685818029e-05,
+      "loss": 16.8361,
+      "step": 13179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4355300984401083e-05,
+      "loss": 18.4792,
+      "step": 13180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4348752526475328e-05,
+      "loss": 0.7048,
+      "step": 13181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4342205312269208e-05,
+      "loss": 0.6788,
+      "step": 13182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4335659342011121e-05,
+      "loss": 19.0002,
+      "step": 13183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4329114615929452e-05,
+      "loss": 0.6876,
+      "step": 13184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4322571134252504e-05,
+      "loss": 19.2219,
+      "step": 13185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4316028897208555e-05,
+      "loss": 0.7086,
+      "step": 13186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.430948790502583e-05,
+      "loss": 0.6693,
+      "step": 13187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.43029481579325e-05,
+      "loss": 17.7106,
+      "step": 13188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4296409656156745e-05,
+      "loss": 0.6655,
+      "step": 13189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4289872399926652e-05,
+      "loss": 0.6831,
+      "step": 13190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.428333638947027e-05,
+      "loss": 0.6975,
+      "step": 13191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4276801625015618e-05,
+      "loss": 17.8609,
+      "step": 13192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4270268106790647e-05,
+      "loss": 0.6313,
+      "step": 13193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4263735835023317e-05,
+      "loss": 19.7854,
+      "step": 13194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4257204809941494e-05,
+      "loss": 0.6521,
+      "step": 13195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.425067503177302e-05,
+      "loss": 16.986,
+      "step": 13196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4244146500745686e-05,
+      "loss": 0.7098,
+      "step": 13197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4237619217087234e-05,
+      "loss": 0.6712,
+      "step": 13198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.42310931810254e-05,
+      "loss": 0.6733,
+      "step": 13199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.422456839278784e-05,
+      "loss": 17.9791,
+      "step": 13200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4218044852602174e-05,
+      "loss": 0.6999,
+      "step": 13201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4211522560695978e-05,
+      "loss": 0.6414,
+      "step": 13202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4205001517296773e-05,
+      "loss": 18.9266,
+      "step": 13203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4198481722632079e-05,
+      "loss": 16.6363,
+      "step": 13204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4191963176929329e-05,
+      "loss": 0.6636,
+      "step": 13205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4185445880415926e-05,
+      "loss": 0.6278,
+      "step": 13206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4178929833319238e-05,
+      "loss": 0.6183,
+      "step": 13207
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4172415035866554e-05,
+      "loss": 0.6706,
+      "step": 13208
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4165901488285194e-05,
+      "loss": 0.6983,
+      "step": 13209
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4159389190802357e-05,
+      "loss": 18.9109,
+      "step": 13210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4152878143645242e-05,
+      "loss": 18.4168,
+      "step": 13211
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4146368347040984e-05,
+      "loss": 18.9909,
+      "step": 13212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4139859801216665e-05,
+      "loss": 16.249,
+      "step": 13213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4133352506399377e-05,
+      "loss": 17.9767,
+      "step": 13214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4126846462816113e-05,
+      "loss": 15.9137,
+      "step": 13215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4120341670693844e-05,
+      "loss": 17.094,
+      "step": 13216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4113838130259494e-05,
+      "loss": 0.6483,
+      "step": 13217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4107335841739939e-05,
+      "loss": 17.7841,
+      "step": 13218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4100834805362018e-05,
+      "loss": 18.3999,
+      "step": 13219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4094335021352522e-05,
+      "loss": 0.672,
+      "step": 13220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4087836489938212e-05,
+      "loss": 17.6789,
+      "step": 13221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4081339211345768e-05,
+      "loss": 0.6808,
+      "step": 13222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4074843185801883e-05,
+      "loss": 0.6632,
+      "step": 13223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4068348413533167e-05,
+      "loss": 18.177,
+      "step": 13224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4061854894766185e-05,
+      "loss": 0.6859,
+      "step": 13225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.405536262972748e-05,
+      "loss": 17.1126,
+      "step": 13226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4048871618643516e-05,
+      "loss": 18.0344,
+      "step": 13227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4042381861740767e-05,
+      "loss": 16.3873,
+      "step": 13228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4035893359245623e-05,
+      "loss": 18.9867,
+      "step": 13229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4029406111384436e-05,
+      "loss": 17.8847,
+      "step": 13230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.402292011838352e-05,
+      "loss": 0.6717,
+      "step": 13231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4016435380469123e-05,
+      "loss": 16.8305,
+      "step": 13232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4009951897867513e-05,
+      "loss": 17.4231,
+      "step": 13233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4003469670804842e-05,
+      "loss": 0.7026,
+      "step": 13234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3996988699507252e-05,
+      "loss": 17.9665,
+      "step": 13235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3990508984200839e-05,
+      "loss": 16.288,
+      "step": 13236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3984030525111652e-05,
+      "loss": 17.7035,
+      "step": 13237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3977553322465675e-05,
+      "loss": 17.0923,
+      "step": 13238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3971077376488911e-05,
+      "loss": 0.7357,
+      "step": 13239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3964602687407258e-05,
+      "loss": 0.6998,
+      "step": 13240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3958129255446584e-05,
+      "loss": 18.1278,
+      "step": 13241
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.395165708083273e-05,
+      "loss": 0.7017,
+      "step": 13242
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3945186163791462e-05,
+      "loss": 0.6635,
+      "step": 13243
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3938716504548554e-05,
+      "loss": 18.4733,
+      "step": 13244
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3932248103329686e-05,
+      "loss": 16.6791,
+      "step": 13245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3925780960360518e-05,
+      "loss": 0.6407,
+      "step": 13246
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3919315075866657e-05,
+      "loss": 0.6017,
+      "step": 13247
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.391285045007366e-05,
+      "loss": 19.1309,
+      "step": 13248
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3906387083207073e-05,
+      "loss": 18.8902,
+      "step": 13249
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3899924975492368e-05,
+      "loss": 16.8626,
+      "step": 13250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3893464127154977e-05,
+      "loss": 16.351,
+      "step": 13251
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.388700453842029e-05,
+      "loss": 0.664,
+      "step": 13252
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.388054620951364e-05,
+      "loss": 0.6234,
+      "step": 13253
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3874089140660356e-05,
+      "loss": 17.6876,
+      "step": 13254
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3867633332085694e-05,
+      "loss": 15.7321,
+      "step": 13255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3861178784014862e-05,
+      "loss": 18.556,
+      "step": 13256
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3854725496673032e-05,
+      "loss": 0.6575,
+      "step": 13257
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3848273470285317e-05,
+      "loss": 15.7459,
+      "step": 13258
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3841822705076829e-05,
+      "loss": 19.0522,
+      "step": 13259
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3835373201272595e-05,
+      "loss": 0.6468,
+      "step": 13260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3828924959097612e-05,
+      "loss": 0.6933,
+      "step": 13261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3822477978776832e-05,
+      "loss": 16.9952,
+      "step": 13262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.381603226053514e-05,
+      "loss": 0.6732,
+      "step": 13263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380958780459744e-05,
+      "loss": 0.6628,
+      "step": 13264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380314461118853e-05,
+      "loss": 0.6769,
+      "step": 13265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3796702680533185e-05,
+      "loss": 0.6943,
+      "step": 13266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3790262012856136e-05,
+      "loss": 17.6111,
+      "step": 13267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3783822608382063e-05,
+      "loss": 16.7754,
+      "step": 13268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.377738446733563e-05,
+      "loss": 0.6507,
+      "step": 13269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.377094758994143e-05,
+      "loss": 19.2156,
+      "step": 13270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3764511976424005e-05,
+      "loss": 17.7219,
+      "step": 13271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.375807762700788e-05,
+      "loss": 16.8849,
+      "step": 13272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3751644541917518e-05,
+      "loss": 17.0328,
+      "step": 13273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3745212721377337e-05,
+      "loss": 17.9683,
+      "step": 13274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.373878216561172e-05,
+      "loss": 0.6362,
+      "step": 13275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3732352874845001e-05,
+      "loss": 17.9953,
+      "step": 13276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3725924849301453e-05,
+      "loss": 18.4102,
+      "step": 13277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3719498089205357e-05,
+      "loss": 0.7095,
+      "step": 13278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3713072594780895e-05,
+      "loss": 0.6475,
+      "step": 13279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3706648366252228e-05,
+      "loss": 0.6522,
+      "step": 13280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3700225403843469e-05,
+      "loss": 0.6841,
+      "step": 13281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.369380370777868e-05,
+      "loss": 15.7661,
+      "step": 13282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3687383278281901e-05,
+      "loss": 0.6755,
+      "step": 13283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.368096411557711e-05,
+      "loss": 0.6551,
+      "step": 13284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.367454621988824e-05,
+      "loss": 17.7898,
+      "step": 13285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3668129591439188e-05,
+      "loss": 16.4239,
+      "step": 13286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3661714230453782e-05,
+      "loss": 17.561,
+      "step": 13287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3655300137155857e-05,
+      "loss": 18.719,
+      "step": 13288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3648887311769165e-05,
+      "loss": 17.3671,
+      "step": 13289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.364247575451742e-05,
+      "loss": 18.3785,
+      "step": 13290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3636065465624286e-05,
+      "loss": 17.9748,
+      "step": 13291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3629656445313383e-05,
+      "loss": 17.7874,
+      "step": 13292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.362324869380832e-05,
+      "loss": 17.9998,
+      "step": 13293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3616842211332626e-05,
+      "loss": 18.1584,
+      "step": 13294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3610436998109794e-05,
+      "loss": 20.2868,
+      "step": 13295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3604033054363275e-05,
+      "loss": 17.0468,
+      "step": 13296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3597630380316455e-05,
+      "loss": 0.6899,
+      "step": 13297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3591228976192732e-05,
+      "loss": 18.1472,
+      "step": 13298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3584828842215407e-05,
+      "loss": 0.656,
+      "step": 13299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3578429978607753e-05,
+      "loss": 0.6539,
+      "step": 13300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3572032385592998e-05,
+      "loss": 0.6081,
+      "step": 13301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3565636063394315e-05,
+      "loss": 0.667,
+      "step": 13302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3559241012234869e-05,
+      "loss": 17.0249,
+      "step": 13303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3552847232337745e-05,
+      "loss": 17.2227,
+      "step": 13304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3546454723925994e-05,
+      "loss": 17.6713,
+      "step": 13305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3540063487222621e-05,
+      "loss": 0.7205,
+      "step": 13306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.353367352245058e-05,
+      "loss": 18.7244,
+      "step": 13307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.352728482983281e-05,
+      "loss": 20.7506,
+      "step": 13308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3520897409592177e-05,
+      "loss": 0.6369,
+      "step": 13309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3514511261951513e-05,
+      "loss": 18.6436,
+      "step": 13310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3508126387133596e-05,
+      "loss": 18.827,
+      "step": 13311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3501742785361154e-05,
+      "loss": 18.0818,
+      "step": 13312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.349536045685692e-05,
+      "loss": 18.0707,
+      "step": 13313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3488979401843521e-05,
+      "loss": 16.5143,
+      "step": 13314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3482599620543579e-05,
+      "loss": 16.9289,
+      "step": 13315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3476221113179638e-05,
+      "loss": 17.655,
+      "step": 13316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.346984387997422e-05,
+      "loss": 0.6954,
+      "step": 13317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.346346792114982e-05,
+      "loss": 0.6614,
+      "step": 13318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3457093236928852e-05,
+      "loss": 0.6655,
+      "step": 13319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3450719827533709e-05,
+      "loss": 0.6264,
+      "step": 13320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.344434769318672e-05,
+      "loss": 17.0013,
+      "step": 13321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3437976834110177e-05,
+      "loss": 16.1307,
+      "step": 13322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3431607250526362e-05,
+      "loss": 0.6659,
+      "step": 13323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3425238942657465e-05,
+      "loss": 17.9216,
+      "step": 13324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3418871910725644e-05,
+      "loss": 0.6898,
+      "step": 13325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.341250615495302e-05,
+      "loss": 18.1744,
+      "step": 13326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3406141675561656e-05,
+      "loss": 18.6402,
+      "step": 13327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3399778472773628e-05,
+      "loss": 17.1433,
+      "step": 13328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3393416546810872e-05,
+      "loss": 18.2516,
+      "step": 13329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3387055897895339e-05,
+      "loss": 16.109,
+      "step": 13330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3380696526248915e-05,
+      "loss": 19.6061,
+      "step": 13331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3374338432093486e-05,
+      "loss": 17.5726,
+      "step": 13332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3367981615650837e-05,
+      "loss": 18.7305,
+      "step": 13333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.336162607714273e-05,
+      "loss": 18.1188,
+      "step": 13334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3355271816790882e-05,
+      "loss": 0.7452,
+      "step": 13335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.334891883481696e-05,
+      "loss": 16.0413,
+      "step": 13336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3342567131442606e-05,
+      "loss": 0.6287,
+      "step": 13337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3336216706889405e-05,
+      "loss": 0.7037,
+      "step": 13338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.332986756137889e-05,
+      "loss": 17.0412,
+      "step": 13339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3323519695132553e-05,
+      "loss": 18.2846,
+      "step": 13340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3317173108371833e-05,
+      "loss": 18.8206,
+      "step": 13341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.331082780131816e-05,
+      "loss": 16.4236,
+      "step": 13342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3304483774192889e-05,
+      "loss": 0.6781,
+      "step": 13343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3298141027217326e-05,
+      "loss": 19.3551,
+      "step": 13344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3291799560612745e-05,
+      "loss": 18.3716,
+      "step": 13345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3285459374600361e-05,
+      "loss": 19.5539,
+      "step": 13346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3279120469401384e-05,
+      "loss": 18.1995,
+      "step": 13347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3272782845236936e-05,
+      "loss": 16.0037,
+      "step": 13348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3266446502328112e-05,
+      "loss": 18.8144,
+      "step": 13349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3260111440895955e-05,
+      "loss": 18.2018,
+      "step": 13350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3253777661161459e-05,
+      "loss": 16.6453,
+      "step": 13351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3247445163345607e-05,
+      "loss": 0.6335,
+      "step": 13352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3241113947669298e-05,
+      "loss": 0.657,
+      "step": 13353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3234784014353407e-05,
+      "loss": 17.3005,
+      "step": 13354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3228455363618758e-05,
+      "loss": 17.789,
+      "step": 13355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3222127995686107e-05,
+      "loss": 0.6187,
+      "step": 13356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3215801910776226e-05,
+      "loss": 17.2557,
+      "step": 13357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3209477109109785e-05,
+      "loss": 0.6919,
+      "step": 13358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3203153590907435e-05,
+      "loss": 15.834,
+      "step": 13359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3196831356389777e-05,
+      "loss": 17.9689,
+      "step": 13360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3190510405777345e-05,
+      "loss": 18.4408,
+      "step": 13361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3184190739290686e-05,
+      "loss": 16.8439,
+      "step": 13362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3177872357150244e-05,
+      "loss": 18.5174,
+      "step": 13363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3171555259576451e-05,
+      "loss": 18.3291,
+      "step": 13364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3165239446789673e-05,
+      "loss": 0.7046,
+      "step": 13365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3158924919010235e-05,
+      "loss": 16.6343,
+      "step": 13366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3152611676458449e-05,
+      "loss": 17.52,
+      "step": 13367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3146299719354544e-05,
+      "loss": 18.2771,
+      "step": 13368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3139989047918717e-05,
+      "loss": 18.9083,
+      "step": 13369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3133679662371118e-05,
+      "loss": 0.6733,
+      "step": 13370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3127371562931845e-05,
+      "loss": 17.8421,
+      "step": 13371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3121064749820982e-05,
+      "loss": 17.7953,
+      "step": 13372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3114759223258543e-05,
+      "loss": 16.1775,
+      "step": 13373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.310845498346449e-05,
+      "loss": 19.0251,
+      "step": 13374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3102152030658754e-05,
+      "loss": 16.2688,
+      "step": 13375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3095850365061208e-05,
+      "loss": 0.7021,
+      "step": 13376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3089549986891719e-05,
+      "loss": 0.6957,
+      "step": 13377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3083250896370059e-05,
+      "loss": 0.6641,
+      "step": 13378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3076953093715983e-05,
+      "loss": 0.6656,
+      "step": 13379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3070656579149187e-05,
+      "loss": 16.9923,
+      "step": 13380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3064361352889326e-05,
+      "loss": 0.7124,
+      "step": 13381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.305806741515605e-05,
+      "loss": 16.4742,
+      "step": 13382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3051774766168883e-05,
+      "loss": 16.2102,
+      "step": 13383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.304548340614737e-05,
+      "loss": 0.6834,
+      "step": 13384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3039193335310962e-05,
+      "loss": 18.6618,
+      "step": 13385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3032904553879133e-05,
+      "loss": 17.383,
+      "step": 13386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3026617062071255e-05,
+      "loss": 0.6573,
+      "step": 13387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3020330860106672e-05,
+      "loss": 18.926,
+      "step": 13388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3014045948204684e-05,
+      "loss": 0.6396,
+      "step": 13389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3007762326584522e-05,
+      "loss": 0.7182,
+      "step": 13390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3001479995465432e-05,
+      "loss": 17.4477,
+      "step": 13391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2995198955066562e-05,
+      "loss": 17.0875,
+      "step": 13392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2988919205607032e-05,
+      "loss": 0.7165,
+      "step": 13393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2982640747305908e-05,
+      "loss": 0.6928,
+      "step": 13394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2976363580382205e-05,
+      "loss": 16.607,
+      "step": 13395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.297008770505495e-05,
+      "loss": 16.0217,
+      "step": 13396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.296381312154305e-05,
+      "loss": 0.6509,
+      "step": 13397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2957539830065408e-05,
+      "loss": 17.0957,
+      "step": 13398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2951267830840868e-05,
+      "loss": 0.6862,
+      "step": 13399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.294499712408822e-05,
+      "loss": 0.606,
+      "step": 13400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.293872771002625e-05,
+      "loss": 16.9282,
+      "step": 13401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2932459588873658e-05,
+      "loss": 0.6237,
+      "step": 13402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2926192760849115e-05,
+      "loss": 15.7574,
+      "step": 13403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2919927226171236e-05,
+      "loss": 18.9977,
+      "step": 13404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2913662985058588e-05,
+      "loss": 17.2697,
+      "step": 13405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2907400037729739e-05,
+      "loss": 16.833,
+      "step": 13406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2901138384403155e-05,
+      "loss": 16.7532,
+      "step": 13407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2894878025297275e-05,
+      "loss": 0.7014,
+      "step": 13408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2888618960630506e-05,
+      "loss": 17.1157,
+      "step": 13409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2882361190621179e-05,
+      "loss": 0.6756,
+      "step": 13410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.287610471548763e-05,
+      "loss": 17.5618,
+      "step": 13411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2869849535448109e-05,
+      "loss": 0.6796,
+      "step": 13412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2863595650720828e-05,
+      "loss": 18.05,
+      "step": 13413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2857343061523963e-05,
+      "loss": 17.1119,
+      "step": 13414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2851091768075624e-05,
+      "loss": 19.2384,
+      "step": 13415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2844841770593919e-05,
+      "loss": 18.4656,
+      "step": 13416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2838593069296872e-05,
+      "loss": 0.643,
+      "step": 13417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2832345664402473e-05,
+      "loss": 17.2895,
+      "step": 13418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2826099556128668e-05,
+      "loss": 16.4222,
+      "step": 13419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2819854744693349e-05,
+      "loss": 0.6224,
+      "step": 13420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2813611230314371e-05,
+      "loss": 0.6469,
+      "step": 13421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2807369013209557e-05,
+      "loss": 0.6891,
+      "step": 13422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.280112809359667e-05,
+      "loss": 0.6404,
+      "step": 13423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2794888471693423e-05,
+      "loss": 17.6326,
+      "step": 13424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2788650147717491e-05,
+      "loss": 17.8641,
+      "step": 13425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2782413121886483e-05,
+      "loss": 17.0461,
+      "step": 13426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2776177394418016e-05,
+      "loss": 18.0533,
+      "step": 13427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2769942965529614e-05,
+      "loss": 0.7354,
+      "step": 13428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2763709835438764e-05,
+      "loss": 18.2049,
+      "step": 13429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2757478004362922e-05,
+      "loss": 16.7775,
+      "step": 13430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2751247472519467e-05,
+      "loss": 17.3794,
+      "step": 13431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2745018240125789e-05,
+      "loss": 17.229,
+      "step": 13432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.273879030739919e-05,
+      "loss": 0.6399,
+      "step": 13433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2732563674556925e-05,
+      "loss": 17.4389,
+      "step": 13434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2726338341816218e-05,
+      "loss": 0.6702,
+      "step": 13435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2720114309394233e-05,
+      "loss": 0.6435,
+      "step": 13436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2713891577508147e-05,
+      "loss": 16.3629,
+      "step": 13437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.270767014637499e-05,
+      "loss": 16.7678,
+      "step": 13438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2701450016211825e-05,
+      "loss": 0.705,
+      "step": 13439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2695231187235623e-05,
+      "loss": 0.6794,
+      "step": 13440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2689013659663368e-05,
+      "loss": 0.6023,
+      "step": 13441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2682797433711947e-05,
+      "loss": 17.7596,
+      "step": 13442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2676582509598217e-05,
+      "loss": 17.8993,
+      "step": 13443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2670368887538992e-05,
+      "loss": 0.7095,
+      "step": 13444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2664156567751024e-05,
+      "loss": 18.4714,
+      "step": 13445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2657945550451055e-05,
+      "loss": 0.6493,
+      "step": 13446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.265173583585576e-05,
+      "loss": 19.164,
+      "step": 13447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2645527424181764e-05,
+      "loss": 0.6743,
+      "step": 13448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2639320315645641e-05,
+      "loss": 18.6064,
+      "step": 13449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2633114510463934e-05,
+      "loss": 16.8099,
+      "step": 13450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2626910008853155e-05,
+      "loss": 0.6732,
+      "step": 13451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2620706811029736e-05,
+      "loss": 0.6881,
+      "step": 13452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.261450491721009e-05,
+      "loss": 18.4902,
+      "step": 13453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260830432761057e-05,
+      "loss": 17.5397,
+      "step": 13454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260210504244747e-05,
+      "loss": 18.2907,
+      "step": 13455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259590706193709e-05,
+      "loss": 0.6865,
+      "step": 13456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2589710386295633e-05,
+      "loss": 15.1432,
+      "step": 13457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2583515015739272e-05,
+      "loss": 17.8379,
+      "step": 13458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2577320950484151e-05,
+      "loss": 0.6909,
+      "step": 13459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2571128190746323e-05,
+      "loss": 20.2124,
+      "step": 13460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2564936736741867e-05,
+      "loss": 0.6336,
+      "step": 13461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2558746588686755e-05,
+      "loss": 0.7008,
+      "step": 13462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2552557746796938e-05,
+      "loss": 17.2967,
+      "step": 13463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2546370211288322e-05,
+      "loss": 0.7098,
+      "step": 13464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2540183982376741e-05,
+      "loss": 0.5701,
+      "step": 13465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2533999060278046e-05,
+      "loss": 18.3267,
+      "step": 13466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2527815445207974e-05,
+      "loss": 18.1347,
+      "step": 13467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2521633137382255e-05,
+      "loss": 17.5555,
+      "step": 13468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2515452137016564e-05,
+      "loss": 0.6452,
+      "step": 13469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2509272444326509e-05,
+      "loss": 0.6327,
+      "step": 13470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2503094059527703e-05,
+      "loss": 18.0308,
+      "step": 13471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.249691698283567e-05,
+      "loss": 0.6912,
+      "step": 13472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2490741214465906e-05,
+      "loss": 0.636,
+      "step": 13473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2484566754633853e-05,
+      "loss": 0.6508,
+      "step": 13474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2478393603554895e-05,
+      "loss": 0.6372,
+      "step": 13475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.247222176144442e-05,
+      "loss": 18.4997,
+      "step": 13476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2466051228517722e-05,
+      "loss": 18.1129,
+      "step": 13477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2459882004990064e-05,
+      "loss": 19.0508,
+      "step": 13478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2453714091076663e-05,
+      "loss": 0.6749,
+      "step": 13479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2447547486992678e-05,
+      "loss": 17.1635,
+      "step": 13480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2441382192953265e-05,
+      "loss": 16.265,
+      "step": 13481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2435218209173489e-05,
+      "loss": 17.2249,
+      "step": 13482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2429055535868384e-05,
+      "loss": 0.6507,
+      "step": 13483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2422894173252935e-05,
+      "loss": 17.3653,
+      "step": 13484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2416734121542085e-05,
+      "loss": 17.2986,
+      "step": 13485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2410575380950752e-05,
+      "loss": 17.4731,
+      "step": 13486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.240441795169377e-05,
+      "loss": 15.7311,
+      "step": 13487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2398261833985948e-05,
+      "loss": 0.654,
+      "step": 13488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.239210702804205e-05,
+      "loss": 0.6827,
+      "step": 13489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2385953534076772e-05,
+      "loss": 0.6341,
+      "step": 13490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2379801352304832e-05,
+      "loss": 18.7583,
+      "step": 13491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2373650482940807e-05,
+      "loss": 18.3184,
+      "step": 13492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2367500926199287e-05,
+      "loss": 0.6886,
+      "step": 13493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2361352682294797e-05,
+      "loss": 0.6829,
+      "step": 13494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2355205751441839e-05,
+      "loss": 0.702,
+      "step": 13495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.234906013385485e-05,
+      "loss": 0.6542,
+      "step": 13496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.234291582974822e-05,
+      "loss": 0.6032,
+      "step": 13497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2336772839336291e-05,
+      "loss": 17.4971,
+      "step": 13498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2330631162833366e-05,
+      "loss": 15.8859,
+      "step": 13499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2324490800453719e-05,
+      "loss": 0.6509,
+      "step": 13500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.231835175241155e-05,
+      "loss": 17.4362,
+      "step": 13501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2312214018921026e-05,
+      "loss": 17.2396,
+      "step": 13502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2306077600196258e-05,
+      "loss": 17.973,
+      "step": 13503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2299942496451317e-05,
+      "loss": 17.1641,
+      "step": 13504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2293808707900257e-05,
+      "loss": 19.7827,
+      "step": 13505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2287676234757033e-05,
+      "loss": 0.6345,
+      "step": 13506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2281545077235595e-05,
+      "loss": 0.6351,
+      "step": 13507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.227541523554983e-05,
+      "loss": 16.7841,
+      "step": 13508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2269286709913563e-05,
+      "loss": 0.7088,
+      "step": 13509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2263159500540621e-05,
+      "loss": 0.6985,
+      "step": 13510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2257033607644751e-05,
+      "loss": 17.9447,
+      "step": 13511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2250909031439645e-05,
+      "loss": 16.6102,
+      "step": 13512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2244785772138972e-05,
+      "loss": 0.6454,
+      "step": 13513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2238663829956337e-05,
+      "loss": 0.6652,
+      "step": 13514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2232543205105323e-05,
+      "loss": 17.7825,
+      "step": 13515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.222642389779945e-05,
+      "loss": 18.4968,
+      "step": 13516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2220305908252183e-05,
+      "loss": 0.6224,
+      "step": 13517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2214189236676964e-05,
+      "loss": 18.3265,
+      "step": 13518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2208073883287164e-05,
+      "loss": 0.6437,
+      "step": 13519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.220195984829614e-05,
+      "loss": 0.6633,
+      "step": 13520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2195847131917177e-05,
+      "loss": 0.6919,
+      "step": 13521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2189735734363517e-05,
+      "loss": 0.6774,
+      "step": 13522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2183625655848363e-05,
+      "loss": 18.9707,
+      "step": 13523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2177516896584856e-05,
+      "loss": 17.6227,
+      "step": 13524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2171409456786137e-05,
+      "loss": 20.5914,
+      "step": 13525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2165303336665246e-05,
+      "loss": 16.8044,
+      "step": 13526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2159198536435207e-05,
+      "loss": 0.582,
+      "step": 13527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2153095056308984e-05,
+      "loss": 0.6517,
+      "step": 13528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2146992896499487e-05,
+      "loss": 16.5612,
+      "step": 13529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2140892057219628e-05,
+      "loss": 0.6887,
+      "step": 13530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2134792538682222e-05,
+      "loss": 19.0228,
+      "step": 13531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2128694341100056e-05,
+      "loss": 0.612,
+      "step": 13532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.212259746468587e-05,
+      "loss": 16.9911,
+      "step": 13533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2116501909652339e-05,
+      "loss": 17.4911,
+      "step": 13534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.211040767621215e-05,
+      "loss": 16.8562,
+      "step": 13535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2104314764577885e-05,
+      "loss": 0.701,
+      "step": 13536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2098223174962092e-05,
+      "loss": 18.0606,
+      "step": 13537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2092132907577292e-05,
+      "loss": 15.8444,
+      "step": 13538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.208604396263593e-05,
+      "loss": 0.6592,
+      "step": 13539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2079956340350446e-05,
+      "loss": 18.0069,
+      "step": 13540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.207387004093321e-05,
+      "loss": 15.3283,
+      "step": 13541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2067785064596532e-05,
+      "loss": 17.4541,
+      "step": 13542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.20617014115527e-05,
+      "loss": 0.6635,
+      "step": 13543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2055619082013935e-05,
+      "loss": 16.6385,
+      "step": 13544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2049538076192441e-05,
+      "loss": 15.4534,
+      "step": 13545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2043458394300367e-05,
+      "loss": 16.1941,
+      "step": 13546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2037380036549778e-05,
+      "loss": 0.71,
+      "step": 13547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2031303003152717e-05,
+      "loss": 0.6763,
+      "step": 13548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2025227294321223e-05,
+      "loss": 17.6818,
+      "step": 13549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2019152910267223e-05,
+      "loss": 17.6186,
+      "step": 13550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2013079851202641e-05,
+      "loss": 17.2382,
+      "step": 13551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2007008117339336e-05,
+      "loss": 18.8186,
+      "step": 13552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2000937708889104e-05,
+      "loss": 0.7072,
+      "step": 13553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1994868626063749e-05,
+      "loss": 17.7572,
+      "step": 13554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1988800869074978e-05,
+      "loss": 0.6388,
+      "step": 13555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.198273443813448e-05,
+      "loss": 0.6478,
+      "step": 13556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1976669333453871e-05,
+      "loss": 17.8074,
+      "step": 13557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1970605555244734e-05,
+      "loss": 0.6474,
+      "step": 13558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1964543103718633e-05,
+      "loss": 19.1333,
+      "step": 13559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.195848197908705e-05,
+      "loss": 0.7205,
+      "step": 13560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1952422181561424e-05,
+      "loss": 0.6933,
+      "step": 13561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1946363711353158e-05,
+      "loss": 0.6803,
+      "step": 13562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1940306568673599e-05,
+      "loss": 17.8293,
+      "step": 13563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1934250753734072e-05,
+      "loss": 0.6492,
+      "step": 13564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1928196266745833e-05,
+      "loss": 15.6483,
+      "step": 13565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1922143107920097e-05,
+      "loss": 0.6311,
+      "step": 13566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1916091277468027e-05,
+      "loss": 0.6722,
+      "step": 13567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1910040775600733e-05,
+      "loss": 18.4107,
+      "step": 13568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1903991602529323e-05,
+      "loss": 16.705,
+      "step": 13569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1897943758464813e-05,
+      "loss": 16.8323,
+      "step": 13570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1891897243618182e-05,
+      "loss": 16.747,
+      "step": 13571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.188585205820037e-05,
+      "loss": 17.97,
+      "step": 13572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1879808202422254e-05,
+      "loss": 16.5351,
+      "step": 13573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1873765676494702e-05,
+      "loss": 18.7985,
+      "step": 13574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1867724480628506e-05,
+      "loss": 16.3902,
+      "step": 13575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1861684615034407e-05,
+      "loss": 0.6816,
+      "step": 13576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1855646079923116e-05,
+      "loss": 16.9571,
+      "step": 13577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.184960887550528e-05,
+      "loss": 0.6904,
+      "step": 13578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1843573001991536e-05,
+      "loss": 17.504,
+      "step": 13579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.183753845959244e-05,
+      "loss": 15.5953,
+      "step": 13580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1831505248518503e-05,
+      "loss": 18.0169,
+      "step": 13581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1825473368980205e-05,
+      "loss": 17.5658,
+      "step": 13582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1819442821187953e-05,
+      "loss": 18.1077,
+      "step": 13583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1813413605352164e-05,
+      "loss": 16.6309,
+      "step": 13584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1807385721683146e-05,
+      "loss": 18.467,
+      "step": 13585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1801359170391196e-05,
+      "loss": 0.6644,
+      "step": 13586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1795333951686549e-05,
+      "loss": 20.1759,
+      "step": 13587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1789310065779386e-05,
+      "loss": 0.6428,
+      "step": 13588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1783287512879887e-05,
+      "loss": 0.7122,
+      "step": 13589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.177726629319813e-05,
+      "loss": 17.1158,
+      "step": 13590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.177124640694418e-05,
+      "loss": 0.6979,
+      "step": 13591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1765227854328036e-05,
+      "loss": 0.6284,
+      "step": 13592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1759210635559654e-05,
+      "loss": 17.8092,
+      "step": 13593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1753194750848968e-05,
+      "loss": 19.0151,
+      "step": 13594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1747180200405839e-05,
+      "loss": 17.5186,
+      "step": 13595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1741166984440088e-05,
+      "loss": 0.6349,
+      "step": 13596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1735155103161494e-05,
+      "loss": 17.0797,
+      "step": 13597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1729144556779759e-05,
+      "loss": 15.5867,
+      "step": 13598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.172313534550461e-05,
+      "loss": 18.6761,
+      "step": 13599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1717127469545657e-05,
+      "loss": 18.4009,
+      "step": 13600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1711120929112507e-05,
+      "loss": 0.6491,
+      "step": 13601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1705115724414655e-05,
+      "loss": 18.6776,
+      "step": 13602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1699111855661648e-05,
+      "loss": 0.6504,
+      "step": 13603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1693109323062923e-05,
+      "loss": 0.6501,
+      "step": 13604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1687108126827867e-05,
+      "loss": 0.6617,
+      "step": 13605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1681108267165853e-05,
+      "loss": 0.6388,
+      "step": 13606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1675109744286168e-05,
+      "loss": 17.0166,
+      "step": 13607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1669112558398104e-05,
+      "loss": 0.6219,
+      "step": 13608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1663116709710864e-05,
+      "loss": 18.3383,
+      "step": 13609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1657122198433618e-05,
+      "loss": 0.7191,
+      "step": 13610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1651129024775487e-05,
+      "loss": 0.6928,
+      "step": 13611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1645137188945531e-05,
+      "loss": 20.2711,
+      "step": 13612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1639146691152814e-05,
+      "loss": 18.4623,
+      "step": 13613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1633157531606304e-05,
+      "loss": 17.4417,
+      "step": 13614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1627169710514929e-05,
+      "loss": 0.5957,
+      "step": 13615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.162118322808759e-05,
+      "loss": 0.6419,
+      "step": 13616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1615198084533124e-05,
+      "loss": 0.7002,
+      "step": 13617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.160921428006031e-05,
+      "loss": 17.78,
+      "step": 13618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1603231814877929e-05,
+      "loss": 0.7174,
+      "step": 13619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1597250689194666e-05,
+      "loss": 17.2281,
+      "step": 13620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1591270903219187e-05,
+      "loss": 0.6871,
+      "step": 13621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.158529245716009e-05,
+      "loss": 0.6352,
+      "step": 13622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1579315351225927e-05,
+      "loss": 17.778,
+      "step": 13623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1573339585625247e-05,
+      "loss": 17.0428,
+      "step": 13624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1567365160566496e-05,
+      "loss": 17.6654,
+      "step": 13625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1561392076258104e-05,
+      "loss": 15.8064,
+      "step": 13626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.155542033290844e-05,
+      "loss": 0.6464,
+      "step": 13627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.154944993072582e-05,
+      "loss": 0.6803,
+      "step": 13628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1543480869918555e-05,
+      "loss": 17.7558,
+      "step": 13629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1537513150694873e-05,
+      "loss": 0.6727,
+      "step": 13630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1531546773262946e-05,
+      "loss": 17.5482,
+      "step": 13631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1525581737830932e-05,
+      "loss": 17.9121,
+      "step": 13632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.15196180446069e-05,
+      "loss": 17.5251,
+      "step": 13633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1513655693798936e-05,
+      "loss": 0.7292,
+      "step": 13634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1507694685615017e-05,
+      "loss": 18.7466,
+      "step": 13635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.15017350202631e-05,
+      "loss": 0.6399,
+      "step": 13636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1495776697951093e-05,
+      "loss": 19.0802,
+      "step": 13637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1489819718886845e-05,
+      "loss": 16.0917,
+      "step": 13638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1483864083278196e-05,
+      "loss": 0.6741,
+      "step": 13639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1477909791332898e-05,
+      "loss": 0.6731,
+      "step": 13640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1471956843258674e-05,
+      "loss": 18.1099,
+      "step": 13641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.146600523926319e-05,
+      "loss": 17.2993,
+      "step": 13642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1460054979554059e-05,
+      "loss": 18.0221,
+      "step": 13643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1454106064338894e-05,
+      "loss": 16.1965,
+      "step": 13644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1448158493825213e-05,
+      "loss": 0.6313,
+      "step": 13645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1442212268220492e-05,
+      "loss": 17.3435,
+      "step": 13646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.143626738773218e-05,
+      "loss": 16.8543,
+      "step": 13647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1430323852567653e-05,
+      "loss": 0.6347,
+      "step": 13648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1424381662934275e-05,
+      "loss": 0.6548,
+      "step": 13649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1418440819039339e-05,
+      "loss": 18.6717,
+      "step": 13650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1412501321090091e-05,
+      "loss": 0.6798,
+      "step": 13651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1406563169293738e-05,
+      "loss": 18.3918,
+      "step": 13652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1400626363857419e-05,
+      "loss": 16.7709,
+      "step": 13653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1394690904988275e-05,
+      "loss": 17.1395,
+      "step": 13654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1388756792893362e-05,
+      "loss": 0.653,
+      "step": 13655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1382824027779677e-05,
+      "loss": 17.1971,
+      "step": 13656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1376892609854184e-05,
+      "loss": 18.2083,
+      "step": 13657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1370962539323837e-05,
+      "loss": 16.1484,
+      "step": 13658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1365033816395487e-05,
+      "loss": 0.6376,
+      "step": 13659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1359106441275974e-05,
+      "loss": 16.403,
+      "step": 13660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.135318041417207e-05,
+      "loss": 17.5147,
+      "step": 13661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1347255735290497e-05,
+      "loss": 18.4163,
+      "step": 13662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1341332404837973e-05,
+      "loss": 0.6926,
+      "step": 13663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.133541042302112e-05,
+      "loss": 16.3548,
+      "step": 13664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1329489790046528e-05,
+      "loss": 15.9763,
+      "step": 13665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.132357050612075e-05,
+      "loss": 0.6841,
+      "step": 13666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1317652571450265e-05,
+      "loss": 16.7937,
+      "step": 13667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1311735986241556e-05,
+      "loss": 16.8303,
+      "step": 13668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1305820750701012e-05,
+      "loss": 0.7031,
+      "step": 13669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1299906865034992e-05,
+      "loss": 0.6285,
+      "step": 13670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1293994329449803e-05,
+      "loss": 0.6783,
+      "step": 13671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.128808314415169e-05,
+      "loss": 17.8749,
+      "step": 13672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1282173309346905e-05,
+      "loss": 0.6461,
+      "step": 13673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1276264825241606e-05,
+      "loss": 0.6521,
+      "step": 13674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.12703576920419e-05,
+      "loss": 17.2028,
+      "step": 13675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1264451909953876e-05,
+      "loss": 17.5422,
+      "step": 13676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.125854747918354e-05,
+      "loss": 18.1678,
+      "step": 13677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1252644399936901e-05,
+      "loss": 17.3689,
+      "step": 13678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1246742672419885e-05,
+      "loss": 0.7027,
+      "step": 13679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1240842296838367e-05,
+      "loss": 0.6533,
+      "step": 13680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1234943273398196e-05,
+      "loss": 16.1263,
+      "step": 13681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1229045602305139e-05,
+      "loss": 18.4352,
+      "step": 13682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1223149283764977e-05,
+      "loss": 17.5369,
+      "step": 13683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1217254317983395e-05,
+      "loss": 18.6053,
+      "step": 13684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1211360705166035e-05,
+      "loss": 0.6453,
+      "step": 13685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1205468445518507e-05,
+      "loss": 17.1991,
+      "step": 13686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1199577539246347e-05,
+      "loss": 17.1755,
+      "step": 13687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1193687986555102e-05,
+      "loss": 0.685,
+      "step": 13688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1187799787650205e-05,
+      "loss": 0.6707,
+      "step": 13689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1181912942737077e-05,
+      "loss": 0.6414,
+      "step": 13690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1176027452021087e-05,
+      "loss": 18.22,
+      "step": 13691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1170143315707537e-05,
+      "loss": 17.7382,
+      "step": 13692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1164260534001725e-05,
+      "loss": 16.9722,
+      "step": 13693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1158379107108874e-05,
+      "loss": 0.6404,
+      "step": 13694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1152499035234149e-05,
+      "loss": 18.4026,
+      "step": 13695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1146620318582685e-05,
+      "loss": 18.0209,
+      "step": 13696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1140742957359551e-05,
+      "loss": 16.9952,
+      "step": 13697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1134866951769812e-05,
+      "loss": 17.5672,
+      "step": 13698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1128992302018443e-05,
+      "loss": 18.3811,
+      "step": 13699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1123119008310385e-05,
+      "loss": 17.4116,
+      "step": 13700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1117247070850534e-05,
+      "loss": 16.4537,
+      "step": 13701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1111376489843717e-05,
+      "loss": 0.6922,
+      "step": 13702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1105507265494764e-05,
+      "loss": 0.6795,
+      "step": 13703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1099639398008417e-05,
+      "loss": 0.6733,
+      "step": 13704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1093772887589377e-05,
+      "loss": 17.7642,
+      "step": 13705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1087907734442304e-05,
+      "loss": 17.2275,
+      "step": 13706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1082043938771792e-05,
+      "loss": 16.4875,
+      "step": 13707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1076181500782435e-05,
+      "loss": 16.9715,
+      "step": 13708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1070320420678726e-05,
+      "loss": 17.4208,
+      "step": 13709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1064460698665158e-05,
+      "loss": 16.667,
+      "step": 13710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1058602334946099e-05,
+      "loss": 17.5553,
+      "step": 13711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1052745329725972e-05,
+      "loss": 18.1444,
+      "step": 13712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1046889683209088e-05,
+      "loss": 18.5922,
+      "step": 13713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1041035395599725e-05,
+      "loss": 16.3475,
+      "step": 13714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1035182467102112e-05,
+      "loss": 0.7077,
+      "step": 13715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.102933089792042e-05,
+      "loss": 17.8438,
+      "step": 13716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1023480688258814e-05,
+      "loss": 0.6371,
+      "step": 13717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1017631838321368e-05,
+      "loss": 17.3871,
+      "step": 13718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1011784348312121e-05,
+      "loss": 0.636,
+      "step": 13719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1005938218435075e-05,
+      "loss": 16.7909,
+      "step": 13720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1000093448894149e-05,
+      "loss": 17.6688,
+      "step": 13721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0994250039893284e-05,
+      "loss": 18.205,
+      "step": 13722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0988407991636307e-05,
+      "loss": 0.6444,
+      "step": 13723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0982567304327024e-05,
+      "loss": 18.9417,
+      "step": 13724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0976727978169199e-05,
+      "loss": 16.9873,
+      "step": 13725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.097089001336652e-05,
+      "loss": 18.69,
+      "step": 13726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0965053410122683e-05,
+      "loss": 0.6711,
+      "step": 13727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0959218168641278e-05,
+      "loss": 0.6126,
+      "step": 13728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.095338428912588e-05,
+      "loss": 16.9441,
+      "step": 13729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0947551771780007e-05,
+      "loss": 17.0085,
+      "step": 13730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0941720616807116e-05,
+      "loss": 18.426,
+      "step": 13731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0935890824410655e-05,
+      "loss": 0.6072,
+      "step": 13732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0930062394793988e-05,
+      "loss": 17.7261,
+      "step": 13733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0924235328160454e-05,
+      "loss": 18.7775,
+      "step": 13734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0918409624713321e-05,
+      "loss": 0.647,
+      "step": 13735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0912585284655813e-05,
+      "loss": 17.281,
+      "step": 13736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0906762308191149e-05,
+      "loss": 0.7187,
+      "step": 13737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.090094069552245e-05,
+      "loss": 15.1171,
+      "step": 13738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0895120446852802e-05,
+      "loss": 0.665,
+      "step": 13739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0889301562385257e-05,
+      "loss": 0.6408,
+      "step": 13740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0883484042322795e-05,
+      "loss": 16.6826,
+      "step": 13741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0877667886868392e-05,
+      "loss": 0.7166,
+      "step": 13742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0871853096224938e-05,
+      "loss": 17.5982,
+      "step": 13743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0866039670595274e-05,
+      "loss": 0.6725,
+      "step": 13744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0860227610182222e-05,
+      "loss": 19.334,
+      "step": 13745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0854416915188514e-05,
+      "loss": 0.6532,
+      "step": 13746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.084860758581689e-05,
+      "loss": 17.2306,
+      "step": 13747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0842799622270006e-05,
+      "loss": 0.6558,
+      "step": 13748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.083699302475047e-05,
+      "loss": 0.6228,
+      "step": 13749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0831187793460851e-05,
+      "loss": 17.9134,
+      "step": 13750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0825383928603655e-05,
+      "loss": 0.7122,
+      "step": 13751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0819581430381381e-05,
+      "loss": 16.9742,
+      "step": 13752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0813780298996445e-05,
+      "loss": 16.7166,
+      "step": 13753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0807980534651214e-05,
+      "loss": 18.1215,
+      "step": 13754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0802182137548022e-05,
+      "loss": 20.0203,
+      "step": 13755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.079638510788914e-05,
+      "loss": 17.4162,
+      "step": 13756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0790589445876825e-05,
+      "loss": 0.6574,
+      "step": 13757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.078479515171325e-05,
+      "loss": 18.4314,
+      "step": 13758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0779002225600554e-05,
+      "loss": 16.7881,
+      "step": 13759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0773210667740824e-05,
+      "loss": 17.5663,
+      "step": 13760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0767420478336093e-05,
+      "loss": 18.6062,
+      "step": 13761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0761631657588384e-05,
+      "loss": 17.7929,
+      "step": 13762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0755844205699622e-05,
+      "loss": 0.7273,
+      "step": 13763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.075005812287172e-05,
+      "loss": 16.6536,
+      "step": 13764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0744273409306516e-05,
+      "loss": 17.8869,
+      "step": 13765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0738490065205824e-05,
+      "loss": 0.6732,
+      "step": 13766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0732708090771398e-05,
+      "loss": 16.8973,
+      "step": 13767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0726927486204941e-05,
+      "loss": 18.3047,
+      "step": 13768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0721148251708123e-05,
+      "loss": 17.2374,
+      "step": 13769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0715370387482532e-05,
+      "loss": 17.8373,
+      "step": 13770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.070959389372977e-05,
+      "loss": 17.2704,
+      "step": 13771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0703818770651336e-05,
+      "loss": 0.6885,
+      "step": 13772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0698045018448699e-05,
+      "loss": 0.6624,
+      "step": 13773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0692272637323281e-05,
+      "loss": 18.0117,
+      "step": 13774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0686501627476442e-05,
+      "loss": 17.2825,
+      "step": 13775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0680731989109538e-05,
+      "loss": 18.0317,
+      "step": 13776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0674963722423825e-05,
+      "loss": 18.6872,
+      "step": 13777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0669196827620548e-05,
+      "loss": 17.8218,
+      "step": 13778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0663431304900873e-05,
+      "loss": 0.6983,
+      "step": 13779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0657667154465934e-05,
+      "loss": 16.9,
+      "step": 13780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0651904376516836e-05,
+      "loss": 0.6558,
+      "step": 13781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0646142971254608e-05,
+      "loss": 17.0152,
+      "step": 13782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.064038293888024e-05,
+      "loss": 0.667,
+      "step": 13783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0634624279594673e-05,
+      "loss": 0.6583,
+      "step": 13784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0628866993598791e-05,
+      "loss": 18.5663,
+      "step": 13785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0623111081093467e-05,
+      "loss": 0.7122,
+      "step": 13786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0617356542279494e-05,
+      "loss": 18.433,
+      "step": 13787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.061160337735761e-05,
+      "loss": 16.4825,
+      "step": 13788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0605851586528526e-05,
+      "loss": 0.6928,
+      "step": 13789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0600101169992882e-05,
+      "loss": 15.138,
+      "step": 13790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0594352127951313e-05,
+      "loss": 0.6441,
+      "step": 13791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0588604460604368e-05,
+      "loss": 18.2102,
+      "step": 13792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0582858168152554e-05,
+      "loss": 17.2938,
+      "step": 13793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0577113250796333e-05,
+      "loss": 0.6375,
+      "step": 13794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0571369708736111e-05,
+      "loss": 0.6591,
+      "step": 13795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0565627542172284e-05,
+      "loss": 0.6604,
+      "step": 13796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.055988675130516e-05,
+      "loss": 0.6508,
+      "step": 13797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0554147336335002e-05,
+      "loss": 15.9044,
+      "step": 13798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0548409297462042e-05,
+      "loss": 17.2211,
+      "step": 13799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0542672634886436e-05,
+      "loss": 0.7275,
+      "step": 13800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.053693734880834e-05,
+      "loss": 0.7027,
+      "step": 13801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0531203439427823e-05,
+      "loss": 17.5989,
+      "step": 13802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0525470906944918e-05,
+      "loss": 0.6392,
+      "step": 13803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0519739751559598e-05,
+      "loss": 0.6302,
+      "step": 13804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0514009973471811e-05,
+      "loss": 16.5957,
+      "step": 13805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0508281572881429e-05,
+      "loss": 16.9289,
+      "step": 13806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0502554549988309e-05,
+      "loss": 19.6442,
+      "step": 13807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.049682890499224e-05,
+      "loss": 0.6862,
+      "step": 13808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0491104638092958e-05,
+      "loss": 16.8649,
+      "step": 13809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0485381749490164e-05,
+      "loss": 0.7236,
+      "step": 13810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0479660239383482e-05,
+      "loss": 16.3932,
+      "step": 13811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0473940107972551e-05,
+      "loss": 17.9881,
+      "step": 13812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0468221355456897e-05,
+      "loss": 0.6454,
+      "step": 13813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0462503982036032e-05,
+      "loss": 0.6853,
+      "step": 13814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0456787987909406e-05,
+      "loss": 0.6201,
+      "step": 13815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.045107337327641e-05,
+      "loss": 19.1225,
+      "step": 13816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.044536013833643e-05,
+      "loss": 17.4208,
+      "step": 13817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0439648283288767e-05,
+      "loss": 0.6108,
+      "step": 13818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0433937808332694e-05,
+      "loss": 0.6779,
+      "step": 13819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0428228713667387e-05,
+      "loss": 16.6331,
+      "step": 13820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0422520999492053e-05,
+      "loss": 18.0617,
+      "step": 13821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0416814666005786e-05,
+      "loss": 17.2504,
+      "step": 13822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0411109713407668e-05,
+      "loss": 0.6741,
+      "step": 13823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0405406141896718e-05,
+      "loss": 0.7137,
+      "step": 13824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.039970395167189e-05,
+      "loss": 17.7211,
+      "step": 13825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.039400314293214e-05,
+      "loss": 17.5366,
+      "step": 13826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.038830371587633e-05,
+      "loss": 18.7904,
+      "step": 13827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0382605670703293e-05,
+      "loss": 17.4751,
+      "step": 13828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.03769090076118e-05,
+      "loss": 17.5197,
+      "step": 13829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0371213726800583e-05,
+      "loss": 17.3403,
+      "step": 13830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.036551982846834e-05,
+      "loss": 18.204,
+      "step": 13831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359827312813702e-05,
+      "loss": 18.4282,
+      "step": 13832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0354136180035257e-05,
+      "loss": 17.2778,
+      "step": 13833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348446430331533e-05,
+      "loss": 15.6187,
+      "step": 13834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034275806390102e-05,
+      "loss": 15.9092,
+      "step": 13835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0337071080942179e-05,
+      "loss": 0.6986,
+      "step": 13836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.03313854816534e-05,
+      "loss": 0.6401,
+      "step": 13837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0325701266233024e-05,
+      "loss": 0.6176,
+      "step": 13838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0320018434879346e-05,
+      "loss": 0.7011,
+      "step": 13839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0314336987790606e-05,
+      "loss": 16.0894,
+      "step": 13840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0308656925165033e-05,
+      "loss": 0.6592,
+      "step": 13841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0302978247200768e-05,
+      "loss": 16.8818,
+      "step": 13842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0297300954095912e-05,
+      "loss": 17.5128,
+      "step": 13843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0291625046048526e-05,
+      "loss": 17.6907,
+      "step": 13844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0285950523256593e-05,
+      "loss": 16.8595,
+      "step": 13845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0280277385918113e-05,
+      "loss": 15.889,
+      "step": 13846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0274605634230983e-05,
+      "loss": 17.8947,
+      "step": 13847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.026893526839306e-05,
+      "loss": 0.6511,
+      "step": 13848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0263266288602158e-05,
+      "loss": 18.5427,
+      "step": 13849
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0257598695056037e-05,
+      "loss": 0.6842,
+      "step": 13850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0251932487952437e-05,
+      "loss": 18.1079,
+      "step": 13851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0246267667489019e-05,
+      "loss": 18.9444,
+      "step": 13852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0240604233863399e-05,
+      "loss": 17.4894,
+      "step": 13853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.023494218727315e-05,
+      "loss": 17.7248,
+      "step": 13854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.022928152791579e-05,
+      "loss": 17.2722,
+      "step": 13855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0223622255988813e-05,
+      "loss": 0.7,
+      "step": 13856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0217964371689642e-05,
+      "loss": 16.8459,
+      "step": 13857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0212307875215649e-05,
+      "loss": 17.05,
+      "step": 13858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0206652766764169e-05,
+      "loss": 17.283,
+      "step": 13859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0200999046532466e-05,
+      "loss": 17.7518,
+      "step": 13860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0195346714717813e-05,
+      "loss": 0.7169,
+      "step": 13861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0189695771517371e-05,
+      "loss": 0.6494,
+      "step": 13862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0184046217128279e-05,
+      "loss": 0.6251,
+      "step": 13863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0178398051747628e-05,
+      "loss": 0.626,
+      "step": 13864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0172751275572445e-05,
+      "loss": 18.6869,
+      "step": 13865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0167105888799749e-05,
+      "loss": 16.7948,
+      "step": 13866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0161461891626466e-05,
+      "loss": 0.6719,
+      "step": 13867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.01558192842495e-05,
+      "loss": 17.7723,
+      "step": 13868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.015017806686569e-05,
+      "loss": 18.2029,
+      "step": 13869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.014453823967182e-05,
+      "loss": 17.6314,
+      "step": 13870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0138899802864671e-05,
+      "loss": 15.274,
+      "step": 13871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0133262756640926e-05,
+      "loss": 0.6579,
+      "step": 13872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012762710119724e-05,
+      "loss": 17.7283,
+      "step": 13873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0121992836730215e-05,
+      "loss": 18.57,
+      "step": 13874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.011635996343641e-05,
+      "loss": 0.6595,
+      "step": 13875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0110728481512328e-05,
+      "loss": 0.658,
+      "step": 13876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0105098391154433e-05,
+      "loss": 0.6752,
+      "step": 13877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0099469692559128e-05,
+      "loss": 19.086,
+      "step": 13878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0093842385922757e-05,
+      "loss": 17.3935,
+      "step": 13879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0088216471441676e-05,
+      "loss": 16.3676,
+      "step": 13880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.008259194931212e-05,
+      "loss": 16.5816,
+      "step": 13881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0076968819730315e-05,
+      "loss": 17.3052,
+      "step": 13882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0071347082892418e-05,
+      "loss": 17.6395,
+      "step": 13883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006572673899454e-05,
+      "loss": 17.5123,
+      "step": 13884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006010778823278e-05,
+      "loss": 16.2917,
+      "step": 13885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0054490230803143e-05,
+      "loss": 16.8406,
+      "step": 13886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0048874066901598e-05,
+      "loss": 19.3634,
+      "step": 13887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0043259296724078e-05,
+      "loss": 0.6302,
+      "step": 13888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0037645920466431e-05,
+      "loss": 0.6743,
+      "step": 13889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0032033938324525e-05,
+      "loss": 18.9493,
+      "step": 13890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0026423350494113e-05,
+      "loss": 18.7221,
+      "step": 13891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0020814157170933e-05,
+      "loss": 0.6446,
+      "step": 13892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0015206358550661e-05,
+      "loss": 17.0861,
+      "step": 13893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0009599954828919e-05,
+      "loss": 0.656,
+      "step": 13894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0003994946201316e-05,
+      "loss": 0.6754,
+      "step": 13895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.998391332863372e-06,
+      "loss": 0.6265,
+      "step": 13896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.99278911501057e-06,
+      "loss": 0.7302,
+      "step": 13897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.987188292838357e-06,
+      "loss": 18.1987,
+      "step": 13898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.981588866542096e-06,
+      "loss": 0.6379,
+      "step": 13899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.975990836317167e-06,
+      "loss": 17.4091,
+      "step": 13900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.97039420235884e-06,
+      "loss": 18.2631,
+      "step": 13901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.96479896486236e-06,
+      "loss": 0.6628,
+      "step": 13902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.95920512402292e-06,
+      "loss": 17.3449,
+      "step": 13903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.953612680035645e-06,
+      "loss": 0.7082,
+      "step": 13904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.948021633095672e-06,
+      "loss": 18.7526,
+      "step": 13905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.942431983398026e-06,
+      "loss": 18.1839,
+      "step": 13906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.936843731137712e-06,
+      "loss": 16.9319,
+      "step": 13907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.931256876509676e-06,
+      "loss": 19.4387,
+      "step": 13908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.925671419708804e-06,
+      "loss": 17.6638,
+      "step": 13909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.920087360929981e-06,
+      "loss": 19.0989,
+      "step": 13910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.914504700367989e-06,
+      "loss": 16.7898,
+      "step": 13911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.908923438217593e-06,
+      "loss": 17.2361,
+      "step": 13912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903343574673496e-06,
+      "loss": 0.7153,
+      "step": 13913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.897765109930341e-06,
+      "loss": 16.9193,
+      "step": 13914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.892188044182765e-06,
+      "loss": 19.5551,
+      "step": 13915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.886612377625315e-06,
+      "loss": 17.597,
+      "step": 13916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.881038110452495e-06,
+      "loss": 0.6054,
+      "step": 13917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.875465242858778e-06,
+      "loss": 18.6606,
+      "step": 13918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.869893775038558e-06,
+      "loss": 17.4711,
+      "step": 13919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.86432370718623e-06,
+      "loss": 18.1906,
+      "step": 13920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.858755039496092e-06,
+      "loss": 0.6696,
+      "step": 13921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.853187772162414e-06,
+      "loss": 18.7278,
+      "step": 13922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.847621905379417e-06,
+      "loss": 0.6448,
+      "step": 13923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.842057439341245e-06,
+      "loss": 19.0558,
+      "step": 13924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.836494374242062e-06,
+      "loss": 17.8253,
+      "step": 13925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.830932710275914e-06,
+      "loss": 17.0099,
+      "step": 13926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.825372447636827e-06,
+      "loss": 18.24,
+      "step": 13927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.819813586518778e-06,
+      "loss": 0.6686,
+      "step": 13928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.81425612711569e-06,
+      "loss": 0.7022,
+      "step": 13929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.808700069621434e-06,
+      "loss": 17.3339,
+      "step": 13930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.803145414229842e-06,
+      "loss": 0.6382,
+      "step": 13931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.797592161134695e-06,
+      "loss": 0.6748,
+      "step": 13932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.792040310529705e-06,
+      "loss": 16.1699,
+      "step": 13933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.786489862608572e-06,
+      "loss": 17.6727,
+      "step": 13934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.780940817564926e-06,
+      "loss": 17.8754,
+      "step": 13935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.775393175592351e-06,
+      "loss": 18.4973,
+      "step": 13936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.76984693688437e-06,
+      "loss": 17.6846,
+      "step": 13937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.764302101634455e-06,
+      "loss": 0.6819,
+      "step": 13938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.758758670036078e-06,
+      "loss": 18.0437,
+      "step": 13939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.753216642282602e-06,
+      "loss": 17.0185,
+      "step": 13940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.74767601856737e-06,
+      "loss": 0.6893,
+      "step": 13941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.742136799083667e-06,
+      "loss": 17.2201,
+      "step": 13942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.736598984024731e-06,
+      "loss": 0.6695,
+      "step": 13943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.73106257358376e-06,
+      "loss": 0.6975,
+      "step": 13944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.725527567953896e-06,
+      "loss": 17.6294,
+      "step": 13945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.719993967328234e-06,
+      "loss": 18.8061,
+      "step": 13946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.714461771899808e-06,
+      "loss": 0.6629,
+      "step": 13947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.708930981861602e-06,
+      "loss": 16.9566,
+      "step": 13948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.703401597406587e-06,
+      "loss": 16.4748,
+      "step": 13949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.697873618727654e-06,
+      "loss": 0.6441,
+      "step": 13950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.692347046017647e-06,
+      "loss": 17.1448,
+      "step": 13951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.68682187946936e-06,
+      "loss": 17.1996,
+      "step": 13952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.681298119275528e-06,
+      "loss": 19.5978,
+      "step": 13953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.675775765628886e-06,
+      "loss": 17.7528,
+      "step": 13954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.670254818722069e-06,
+      "loss": 17.7923,
+      "step": 13955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.664735278747677e-06,
+      "loss": 18.7383,
+      "step": 13956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.659217145898264e-06,
+      "loss": 17.694,
+      "step": 13957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.653700420366325e-06,
+      "loss": 17.2748,
+      "step": 13958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.648185102344337e-06,
+      "loss": 17.8675,
+      "step": 13959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.642671192024687e-06,
+      "loss": 17.6505,
+      "step": 13960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.637158689599746e-06,
+      "loss": 17.6687,
+      "step": 13961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63164759526181e-06,
+      "loss": 17.0865,
+      "step": 13962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.626137909203131e-06,
+      "loss": 16.1262,
+      "step": 13963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.62062963161594e-06,
+      "loss": 0.665,
+      "step": 13964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.615122762692385e-06,
+      "loss": 20.7629,
+      "step": 13965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.609617302624579e-06,
+      "loss": 18.2202,
+      "step": 13966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.604113251604585e-06,
+      "loss": 18.9105,
+      "step": 13967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.598610609824398e-06,
+      "loss": 17.2553,
+      "step": 13968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.593109377476012e-06,
+      "loss": 17.3345,
+      "step": 13969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.587609554751326e-06,
+      "loss": 0.662,
+      "step": 13970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.582111141842204e-06,
+      "loss": 0.7166,
+      "step": 13971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.57661413894047e-06,
+      "loss": 0.7119,
+      "step": 13972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.571118546237868e-06,
+      "loss": 17.3338,
+      "step": 13973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.565624363926152e-06,
+      "loss": 19.9734,
+      "step": 13974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.560131592196964e-06,
+      "loss": 18.1041,
+      "step": 13975
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.554640231241935e-06,
+      "loss": 15.6018,
+      "step": 13976
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.549150281252633e-06,
+      "loss": 16.767,
+      "step": 13977
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.543661742420561e-06,
+      "loss": 0.7139,
+      "step": 13978
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.538174614937223e-06,
+      "loss": 17.0281,
+      "step": 13979
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.532688898994019e-06,
+      "loss": 0.6674,
+      "step": 13980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.527204594782334e-06,
+      "loss": 17.8813,
+      "step": 13981
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.521721702493486e-06,
+      "loss": 16.6657,
+      "step": 13982
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.516240222318745e-06,
+      "loss": 17.3841,
+      "step": 13983
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.510760154449344e-06,
+      "loss": 19.2918,
+      "step": 13984
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.505281499076458e-06,
+      "loss": 16.4557,
+      "step": 13985
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.499804256391214e-06,
+      "loss": 19.4116,
+      "step": 13986
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.49432842658467e-06,
+      "loss": 17.0613,
+      "step": 13987
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.488854009847887e-06,
+      "loss": 16.5989,
+      "step": 13988
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.483381006371834e-06,
+      "loss": 16.9907,
+      "step": 13989
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.47790941634743e-06,
+      "loss": 16.7157,
+      "step": 13990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.472439239965563e-06,
+      "loss": 17.916,
+      "step": 13991
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.466970477417048e-06,
+      "loss": 0.7513,
+      "step": 13992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.461503128892696e-06,
+      "loss": 16.4411,
+      "step": 13993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.456037194583223e-06,
+      "loss": 0.6258,
+      "step": 13994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.450572674679314e-06,
+      "loss": 18.5854,
+      "step": 13995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.445109569371602e-06,
+      "loss": 19.4406,
+      "step": 13996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.43964787885066e-06,
+      "loss": 15.4691,
+      "step": 13997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.434187603307048e-06,
+      "loss": 0.6775,
+      "step": 13998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.428728742931236e-06,
+      "loss": 18.2093,
+      "step": 13999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.423271297913667e-06,
+      "loss": 17.904,
+      "step": 14000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.417815268444718e-06,
+      "loss": 16.2378,
+      "step": 14001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.412360654714742e-06,
+      "loss": 19.3708,
+      "step": 14002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.406907456913994e-06,
+      "loss": 16.9959,
+      "step": 14003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.401455675232756e-06,
+      "loss": 18.1131,
+      "step": 14004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.396005309861194e-06,
+      "loss": 0.6977,
+      "step": 14005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.39055636098945e-06,
+      "loss": 0.7262,
+      "step": 14006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.38510882880762e-06,
+      "loss": 16.2368,
+      "step": 14007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.379662713505727e-06,
+      "loss": 0.6789,
+      "step": 14008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.37421801527379e-06,
+      "loss": 0.6449,
+      "step": 14009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.368774734301733e-06,
+      "loss": 18.0887,
+      "step": 14010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.363332870779462e-06,
+      "loss": 18.2364,
+      "step": 14011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.357892424896807e-06,
+      "loss": 20.1242,
+      "step": 14012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.352453396843557e-06,
+      "loss": 0.6677,
+      "step": 14013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.347015786809477e-06,
+      "loss": 17.2884,
+      "step": 14014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.341579594984251e-06,
+      "loss": 17.3939,
+      "step": 14015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.336144821557529e-06,
+      "loss": 0.7124,
+      "step": 14016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.3307114667189e-06,
+      "loss": 18.6682,
+      "step": 14017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.3252795306579e-06,
+      "loss": 16.9606,
+      "step": 14018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.319849013564058e-06,
+      "loss": 18.0847,
+      "step": 14019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.314419915626799e-06,
+      "loss": 16.3097,
+      "step": 14020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.308992237035524e-06,
+      "loss": 19.1331,
+      "step": 14021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.303565977979583e-06,
+      "loss": 17.1664,
+      "step": 14022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.298141138648264e-06,
+      "loss": 17.9226,
+      "step": 14023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.292717719230837e-06,
+      "loss": 0.6301,
+      "step": 14024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.287295719916495e-06,
+      "loss": 0.6275,
+      "step": 14025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.281875140894381e-06,
+      "loss": 17.3073,
+      "step": 14026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.276455982353605e-06,
+      "loss": 0.7554,
+      "step": 14027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.2710382444832e-06,
+      "loss": 0.6247,
+      "step": 14028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.265621927472196e-06,
+      "loss": 0.6768,
+      "step": 14029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.260207031509526e-06,
+      "loss": 18.7545,
+      "step": 14030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.254793556784097e-06,
+      "loss": 16.7339,
+      "step": 14031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.249381503484766e-06,
+      "loss": 0.6351,
+      "step": 14032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.243970871800317e-06,
+      "loss": 16.7004,
+      "step": 14033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.238561661919531e-06,
+      "loss": 17.2297,
+      "step": 14034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.233153874031102e-06,
+      "loss": 19.3531,
+      "step": 14035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.22774750832368e-06,
+      "loss": 17.9107,
+      "step": 14036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.222342564985875e-06,
+      "loss": 0.7029,
+      "step": 14037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.216939044206241e-06,
+      "loss": 15.9218,
+      "step": 14038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.211536946173283e-06,
+      "loss": 16.9271,
+      "step": 14039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.206136271075461e-06,
+      "loss": 0.6191,
+      "step": 14040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.200737019101168e-06,
+      "loss": 0.6493,
+      "step": 14041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.195339190438768e-06,
+      "loss": 17.2471,
+      "step": 14042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.189942785276578e-06,
+      "loss": 0.7246,
+      "step": 14043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.18454780380285e-06,
+      "loss": 17.8691,
+      "step": 14044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.17915424620579e-06,
+      "loss": 15.8606,
+      "step": 14045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.173762112673555e-06,
+      "loss": 16.4456,
+      "step": 14046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.168371403394243e-06,
+      "loss": 0.6602,
+      "step": 14047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.162982118555935e-06,
+      "loss": 17.4161,
+      "step": 14048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.15759425834663e-06,
+      "loss": 17.4244,
+      "step": 14049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.15220782295429e-06,
+      "loss": 15.7946,
+      "step": 14050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.146822812566818e-06,
+      "loss": 17.5564,
+      "step": 14051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.141439227372068e-06,
+      "loss": 0.6266,
+      "step": 14052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.136057067557874e-06,
+      "loss": 17.9374,
+      "step": 14053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.130676333311978e-06,
+      "loss": 19.2728,
+      "step": 14054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.1252970248221e-06,
+      "loss": 0.6962,
+      "step": 14055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.11991914227589e-06,
+      "loss": 0.6469,
+      "step": 14056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.11454268586096e-06,
+      "loss": 17.724,
+      "step": 14057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.109167655764889e-06,
+      "loss": 19.6405,
+      "step": 14058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.103794052175174e-06,
+      "loss": 16.8505,
+      "step": 14059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.098421875279278e-06,
+      "loss": 0.6595,
+      "step": 14060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.093051125264623e-06,
+      "loss": 18.0288,
+      "step": 14061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.087681802318548e-06,
+      "loss": 17.2955,
+      "step": 14062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.082313906628393e-06,
+      "loss": 15.7831,
+      "step": 14063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.076947438381412e-06,
+      "loss": 18.7847,
+      "step": 14064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.071582397764816e-06,
+      "loss": 16.1113,
+      "step": 14065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.066218784965763e-06,
+      "loss": 0.6446,
+      "step": 14066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.060856600171364e-06,
+      "loss": 16.8023,
+      "step": 14067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.0554958435687e-06,
+      "loss": 19.1753,
+      "step": 14068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.050136515344781e-06,
+      "loss": 17.5696,
+      "step": 14069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.04477861568656e-06,
+      "loss": 17.4582,
+      "step": 14070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.039422144780957e-06,
+      "loss": 0.7068,
+      "step": 14071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.034067102814824e-06,
+      "loss": 18.4563,
+      "step": 14072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.028713489975005e-06,
+      "loss": 0.6938,
+      "step": 14073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.023361306448242e-06,
+      "loss": 17.6974,
+      "step": 14074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.018010552421252e-06,
+      "loss": 17.1849,
+      "step": 14075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.01266122808071e-06,
+      "loss": 0.6482,
+      "step": 14076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.00731333361321e-06,
+      "loss": 17.315,
+      "step": 14077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.001966869205342e-06,
+      "loss": 17.4377,
+      "step": 14078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.996621835043606e-06,
+      "loss": 18.9017,
+      "step": 14079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.991278231314476e-06,
+      "loss": 0.6898,
+      "step": 14080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.985936058204364e-06,
+      "loss": 16.4973,
+      "step": 14081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.980595315899621e-06,
+      "loss": 15.9433,
+      "step": 14082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.97525600458659e-06,
+      "loss": 17.5699,
+      "step": 14083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.969918124451527e-06,
+      "loss": 18.3711,
+      "step": 14084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.964581675680639e-06,
+      "loss": 17.648,
+      "step": 14085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.959246658460096e-06,
+      "loss": 0.6641,
+      "step": 14086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.953913072976006e-06,
+      "loss": 0.7174,
+      "step": 14087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.948580919414457e-06,
+      "loss": 17.1151,
+      "step": 14088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.94325019796145e-06,
+      "loss": 17.3732,
+      "step": 14089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.937920908802955e-06,
+      "loss": 0.6717,
+      "step": 14090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.932593052124882e-06,
+      "loss": 17.4645,
+      "step": 14091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.927266628113107e-06,
+      "loss": 16.253,
+      "step": 14092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.921941636953435e-06,
+      "loss": 0.6621,
+      "step": 14093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.91661807883164e-06,
+      "loss": 18.6065,
+      "step": 14094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.911295953933435e-06,
+      "loss": 0.7459,
+      "step": 14095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.905975262444477e-06,
+      "loss": 18.0046,
+      "step": 14096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.900656004550406e-06,
+      "loss": 17.5485,
+      "step": 14097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.89533818043677e-06,
+      "loss": 16.391,
+      "step": 14098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.890021790289094e-06,
+      "loss": 0.6571,
+      "step": 14099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.884706834292833e-06,
+      "loss": 0.6238,
+      "step": 14100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.879393312633406e-06,
+      "loss": 0.6227,
+      "step": 14101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.87408122549619e-06,
+      "loss": 21.2752,
+      "step": 14102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.868770573066493e-06,
+      "loss": 0.6736,
+      "step": 14103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.863461355529585e-06,
+      "loss": 19.1304,
+      "step": 14104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.858153573070677e-06,
+      "loss": 17.3873,
+      "step": 14105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.852847225874928e-06,
+      "loss": 18.3189,
+      "step": 14106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.847542314127472e-06,
+      "loss": 17.7751,
+      "step": 14107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.842238838013361e-06,
+      "loss": 0.7057,
+      "step": 14108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.83693679771762e-06,
+      "loss": 0.584,
+      "step": 14109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.831636193425209e-06,
+      "loss": 0.676,
+      "step": 14110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.82633702532103e-06,
+      "loss": 16.6337,
+      "step": 14111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.821039293589972e-06,
+      "loss": 0.6402,
+      "step": 14112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.815742998416843e-06,
+      "loss": 17.2506,
+      "step": 14113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.810448139986405e-06,
+      "loss": 0.6533,
+      "step": 14114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.805154718483372e-06,
+      "loss": 17.6609,
+      "step": 14115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.799862734092401e-06,
+      "loss": 18.0279,
+      "step": 14116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.794572186998124e-06,
+      "loss": 17.7012,
+      "step": 14117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.789283077385102e-06,
+      "loss": 18.4448,
+      "step": 14118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.78399540543784e-06,
+      "loss": 18.7849,
+      "step": 14119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.77870917134081e-06,
+      "loss": 0.6617,
+      "step": 14120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.77342437527841e-06,
+      "loss": 0.6104,
+      "step": 14121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.768141017435034e-06,
+      "loss": 16.7524,
+      "step": 14122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.762859097994975e-06,
+      "loss": 17.5553,
+      "step": 14123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.757578617142499e-06,
+      "loss": 16.0043,
+      "step": 14124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.752299575061823e-06,
+      "loss": 0.6756,
+      "step": 14125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.747021971937092e-06,
+      "loss": 15.4844,
+      "step": 14126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.74174580795245e-06,
+      "loss": 18.4611,
+      "step": 14127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.736471083291947e-06,
+      "loss": 0.7142,
+      "step": 14128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.731197798139595e-06,
+      "loss": 16.6917,
+      "step": 14129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.725925952679347e-06,
+      "loss": 17.7298,
+      "step": 14130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.72065554709512e-06,
+      "loss": 18.2893,
+      "step": 14131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.715386581570783e-06,
+      "loss": 0.6403,
+      "step": 14132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.71011905629015e-06,
+      "loss": 0.5941,
+      "step": 14133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.704852971436977e-06,
+      "loss": 0.6187,
+      "step": 14134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.699588327194968e-06,
+      "loss": 18.5389,
+      "step": 14135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.694325123747782e-06,
+      "loss": 16.1752,
+      "step": 14136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.689063361279049e-06,
+      "loss": 17.3125,
+      "step": 14137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.683803039972321e-06,
+      "loss": 16.4887,
+      "step": 14138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.678544160011104e-06,
+      "loss": 0.6323,
+      "step": 14139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.67328672157886e-06,
+      "loss": 0.671,
+      "step": 14140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.668030724858983e-06,
+      "loss": 17.034,
+      "step": 14141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.662776170034858e-06,
+      "loss": 19.1318,
+      "step": 14142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.657523057289791e-06,
+      "loss": 17.1685,
+      "step": 14143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.652271386807025e-06,
+      "loss": 17.2155,
+      "step": 14144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.64702115876978e-06,
+      "loss": 0.625,
+      "step": 14145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.641772373361196e-06,
+      "loss": 0.6748,
+      "step": 14146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.636525030764419e-06,
+      "loss": 17.2189,
+      "step": 14147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.631279131162473e-06,
+      "loss": 18.1258,
+      "step": 14148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.62603467473837e-06,
+      "loss": 0.6291,
+      "step": 14149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.620791661675065e-06,
+      "loss": 16.6294,
+      "step": 14150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.615550092155478e-06,
+      "loss": 0.662,
+      "step": 14151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.610309966362457e-06,
+      "loss": 16.5462,
+      "step": 14152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.605071284478804e-06,
+      "loss": 16.6345,
+      "step": 14153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.599834046687284e-06,
+      "loss": 0.6072,
+      "step": 14154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.594598253170582e-06,
+      "loss": 16.879,
+      "step": 14155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.589363904111375e-06,
+      "loss": 0.6686,
+      "step": 14156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.584130999692264e-06,
+      "loss": 17.2694,
+      "step": 14157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.578899540095792e-06,
+      "loss": 0.6651,
+      "step": 14158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.57366952550447e-06,
+      "loss": 18.8637,
+      "step": 14159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.568440956100737e-06,
+      "loss": 17.8391,
+      "step": 14160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.563213832067014e-06,
+      "loss": 0.6405,
+      "step": 14161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.55798815358565e-06,
+      "loss": 0.6574,
+      "step": 14162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.55276392083894e-06,
+      "loss": 18.0134,
+      "step": 14163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.54754113400914e-06,
+      "loss": 18.1813,
+      "step": 14164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.54231979327843e-06,
+      "loss": 0.7066,
+      "step": 14165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.537099898828992e-06,
+      "loss": 0.6732,
+      "step": 14166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.531881450842915e-06,
+      "loss": 19.3849,
+      "step": 14167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.526664449502247e-06,
+      "loss": 0.7057,
+      "step": 14168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.521448894988982e-06,
+      "loss": 18.7306,
+      "step": 14169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.516234787485062e-06,
+      "loss": 0.7205,
+      "step": 14170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.511022127172402e-06,
+      "loss": 17.6346,
+      "step": 14171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.505810914232843e-06,
+      "loss": 18.6483,
+      "step": 14172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.500601148848186e-06,
+      "loss": 0.7169,
+      "step": 14173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.495392831200167e-06,
+      "loss": 0.629,
+      "step": 14174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.490185961470476e-06,
+      "loss": 17.3492,
+      "step": 14175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.484980539840787e-06,
+      "loss": 18.5459,
+      "step": 14176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.47977656649267e-06,
+      "loss": 16.8942,
+      "step": 14177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.47457404160768e-06,
+      "loss": 17.3108,
+      "step": 14178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.469372965367311e-06,
+      "loss": 0.7036,
+      "step": 14179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.46417333795299e-06,
+      "loss": 0.646,
+      "step": 14180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.458975159546135e-06,
+      "loss": 0.6622,
+      "step": 14181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.453778430328074e-06,
+      "loss": 18.1686,
+      "step": 14182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.4485831504801e-06,
+      "loss": 18.2941,
+      "step": 14183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.44338932018346e-06,
+      "loss": 17.3241,
+      "step": 14184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.438196939619336e-06,
+      "loss": 19.2881,
+      "step": 14185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.433006008968858e-06,
+      "loss": 0.7081,
+      "step": 14186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.427816528413145e-06,
+      "loss": 0.6699,
+      "step": 14187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.422628498133217e-06,
+      "loss": 19.4488,
+      "step": 14188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.417441918310066e-06,
+      "loss": 15.1747,
+      "step": 14189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.412256789124629e-06,
+      "loss": 16.4129,
+      "step": 14190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.40707311075778e-06,
+      "loss": 18.1148,
+      "step": 14191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.401890883390379e-06,
+      "loss": 18.0855,
+      "step": 14192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.396710107203203e-06,
+      "loss": 17.3164,
+      "step": 14193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.391530782376982e-06,
+      "loss": 16.0352,
+      "step": 14194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.386352909092404e-06,
+      "loss": 0.7031,
+      "step": 14195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.38117648753009e-06,
+      "loss": 0.6987,
+      "step": 14196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.376001517870652e-06,
+      "loss": 18.0403,
+      "step": 14197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.370828000294601e-06,
+      "loss": 16.9493,
+      "step": 14198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.365655934982425e-06,
+      "loss": 17.2408,
+      "step": 14199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.360485322114552e-06,
+      "loss": 18.0556,
+      "step": 14200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.355316161871369e-06,
+      "loss": 17.7035,
+      "step": 14201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.3501484544332e-06,
+      "loss": 17.8526,
+      "step": 14202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.344982199980323e-06,
+      "loss": 0.6698,
+      "step": 14203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.339817398692973e-06,
+      "loss": 18.2837,
+      "step": 14204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.334654050751311e-06,
+      "loss": 16.8383,
+      "step": 14205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.329492156335489e-06,
+      "loss": 0.6023,
+      "step": 14206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.324331715625571e-06,
+      "loss": 0.6928,
+      "step": 14207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.319172728801583e-06,
+      "loss": 17.4625,
+      "step": 14208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.3140151960435e-06,
+      "loss": 0.6009,
+      "step": 14209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.308859117531236e-06,
+      "loss": 0.6843,
+      "step": 14210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.303704493444692e-06,
+      "loss": 0.6785,
+      "step": 14211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.298551323963666e-06,
+      "loss": 0.6692,
+      "step": 14212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.293399609267943e-06,
+      "loss": 17.9479,
+      "step": 14213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.288249349537242e-06,
+      "loss": 0.7117,
+      "step": 14214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.283100544951212e-06,
+      "loss": 16.8002,
+      "step": 14215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.277953195689508e-06,
+      "loss": 0.6595,
+      "step": 14216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.272807301931684e-06,
+      "loss": 0.6905,
+      "step": 14217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.267662863857256e-06,
+      "loss": 18.3544,
+      "step": 14218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.262519881645691e-06,
+      "loss": 0.6556,
+      "step": 14219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.257378355476392e-06,
+      "loss": 17.135,
+      "step": 14220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.252238285528758e-06,
+      "loss": 17.9112,
+      "step": 14221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.247099671982084e-06,
+      "loss": 0.6123,
+      "step": 14222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.241962515015638e-06,
+      "loss": 0.6264,
+      "step": 14223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.236826814808629e-06,
+      "loss": 20.7446,
+      "step": 14224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.231692571540212e-06,
+      "loss": 0.6859,
+      "step": 14225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.226559785389521e-06,
+      "loss": 17.5308,
+      "step": 14226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.221428456535607e-06,
+      "loss": 0.7031,
+      "step": 14227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.216298585157473e-06,
+      "loss": 19.6497,
+      "step": 14228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.21117017143409e-06,
+      "loss": 0.645,
+      "step": 14229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.206043215544345e-06,
+      "loss": 16.0162,
+      "step": 14230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.200917717667123e-06,
+      "loss": 0.6842,
+      "step": 14231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.195793677981217e-06,
+      "loss": 17.3733,
+      "step": 14232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.190671096665387e-06,
+      "loss": 17.1584,
+      "step": 14233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.185549973898332e-06,
+      "loss": 17.5722,
+      "step": 14234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.1804303098587e-06,
+      "loss": 0.7043,
+      "step": 14235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.175312104725113e-06,
+      "loss": 16.4583,
+      "step": 14236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.170195358676113e-06,
+      "loss": 17.2211,
+      "step": 14237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.165080071890208e-06,
+      "loss": 19.6972,
+      "step": 14238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.159966244545835e-06,
+      "loss": 0.6997,
+      "step": 14239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.154853876821394e-06,
+      "loss": 18.4186,
+      "step": 14240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.149742968895251e-06,
+      "loss": 15.7855,
+      "step": 14241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.1446335209457e-06,
+      "loss": 18.1492,
+      "step": 14242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.139525533150976e-06,
+      "loss": 0.6393,
+      "step": 14243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.134419005689281e-06,
+      "loss": 0.5957,
+      "step": 14244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.129313938738742e-06,
+      "loss": 17.0803,
+      "step": 14245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.12421033247749e-06,
+      "loss": 0.6714,
+      "step": 14246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.119108187083545e-06,
+      "loss": 18.3866,
+      "step": 14247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.1140075027349e-06,
+      "loss": 18.3637,
+      "step": 14248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.1089082796095e-06,
+      "loss": 0.6156,
+      "step": 14249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.103810517885225e-06,
+      "loss": 17.5767,
+      "step": 14250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.098714217739928e-06,
+      "loss": 15.9888,
+      "step": 14251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.093619379351397e-06,
+      "loss": 0.6641,
+      "step": 14252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.088526002897363e-06,
+      "loss": 0.6677,
+      "step": 14253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.083434088555508e-06,
+      "loss": 0.6629,
+      "step": 14254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.078343636503467e-06,
+      "loss": 16.3751,
+      "step": 14255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.073254646918849e-06,
+      "loss": 0.723,
+      "step": 14256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.068167119979154e-06,
+      "loss": 17.6295,
+      "step": 14257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.063081055861883e-06,
+      "loss": 16.9653,
+      "step": 14258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.057996454744443e-06,
+      "loss": 0.6604,
+      "step": 14259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.052913316804245e-06,
+      "loss": 0.5934,
+      "step": 14260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.047831642218611e-06,
+      "loss": 0.638,
+      "step": 14261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.042751431164808e-06,
+      "loss": 17.9128,
+      "step": 14262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.037672683820069e-06,
+      "loss": 17.6336,
+      "step": 14263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.032595400361559e-06,
+      "loss": 19.0254,
+      "step": 14264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.02751958096642e-06,
+      "loss": 0.646,
+      "step": 14265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.02244522581172e-06,
+      "loss": 16.9663,
+      "step": 14266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.017372335074486e-06,
+      "loss": 17.6653,
+      "step": 14267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.012300908931675e-06,
+      "loss": 17.0208,
+      "step": 14268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.007230947560207e-06,
+      "loss": 0.6445,
+      "step": 14269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.002162451136973e-06,
+      "loss": 0.5724,
+      "step": 14270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.997095419838773e-06,
+      "loss": 18.0935,
+      "step": 14271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.992029853842386e-06,
+      "loss": 17.8683,
+      "step": 14272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.986965753324516e-06,
+      "loss": 17.593,
+      "step": 14273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.981903118461826e-06,
+      "loss": 17.5976,
+      "step": 14274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.976841949430942e-06,
+      "loss": 18.1767,
+      "step": 14275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.971782246408426e-06,
+      "loss": 18.7065,
+      "step": 14276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.966724009570781e-06,
+      "loss": 17.5325,
+      "step": 14277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.961667239094472e-06,
+      "loss": 16.6756,
+      "step": 14278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.9566119351559e-06,
+      "loss": 0.6869,
+      "step": 14279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.951558097931439e-06,
+      "loss": 0.7305,
+      "step": 14280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.946505727597381e-06,
+      "loss": 16.8379,
+      "step": 14281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.941454824329992e-06,
+      "loss": 0.642,
+      "step": 14282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.936405388305473e-06,
+      "loss": 17.2084,
+      "step": 14283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.931357419699959e-06,
+      "loss": 0.6688,
+      "step": 14284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.92631091868959e-06,
+      "loss": 15.714,
+      "step": 14285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.921265885450385e-06,
+      "loss": 17.6854,
+      "step": 14286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.916222320158361e-06,
+      "loss": 18.4939,
+      "step": 14287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.911180222989457e-06,
+      "loss": 17.2425,
+      "step": 14288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.906139594119565e-06,
+      "loss": 16.5963,
+      "step": 14289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.901100433724545e-06,
+      "loss": 0.5583,
+      "step": 14290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.896062741980192e-06,
+      "loss": 0.6624,
+      "step": 14291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.89102651906224e-06,
+      "loss": 0.7092,
+      "step": 14292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.885991765146389e-06,
+      "loss": 17.1737,
+      "step": 14293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.880958480408258e-06,
+      "loss": 18.6323,
+      "step": 14294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.87592666502347e-06,
+      "loss": 0.6595,
+      "step": 14295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.870896319167548e-06,
+      "loss": 15.9643,
+      "step": 14296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.865867443015979e-06,
+      "loss": 17.4951,
+      "step": 14297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.860840036744193e-06,
+      "loss": 19.0159,
+      "step": 14298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.855814100527575e-06,
+      "loss": 16.1176,
+      "step": 14299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.850789634541477e-06,
+      "loss": 17.4084,
+      "step": 14300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.845766638961171e-06,
+      "loss": 17.3651,
+      "step": 14301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.840745113961879e-06,
+      "loss": 16.1084,
+      "step": 14302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.835725059718791e-06,
+      "loss": 0.6566,
+      "step": 14303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.830706476407018e-06,
+      "loss": 17.5868,
+      "step": 14304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.825689364201665e-06,
+      "loss": 18.6689,
+      "step": 14305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.820673723277738e-06,
+      "loss": 16.7629,
+      "step": 14306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.815659553810217e-06,
+      "loss": 19.4343,
+      "step": 14307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.810646855974029e-06,
+      "loss": 17.4054,
+      "step": 14308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.805635629944025e-06,
+      "loss": 20.9549,
+      "step": 14309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.80062587589505e-06,
+      "loss": 16.785,
+      "step": 14310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.79561759400188e-06,
+      "loss": 0.7028,
+      "step": 14311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.790610784439207e-06,
+      "loss": 16.4129,
+      "step": 14312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.785605447381689e-06,
+      "loss": 18.4107,
+      "step": 14313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.780601583003971e-06,
+      "loss": 17.9737,
+      "step": 14314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.775599191480603e-06,
+      "loss": 18.191,
+      "step": 14315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.7705982729861e-06,
+      "loss": 19.283,
+      "step": 14316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.76559882769492e-06,
+      "loss": 18.162,
+      "step": 14317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.760600855781457e-06,
+      "loss": 0.6484,
+      "step": 14318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.755604357420104e-06,
+      "loss": 17.2954,
+      "step": 14319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.750609332785141e-06,
+      "loss": 17.6028,
+      "step": 14320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.745615782050836e-06,
+      "loss": 16.1706,
+      "step": 14321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.740623705391381e-06,
+      "loss": 0.663,
+      "step": 14322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.735633102980927e-06,
+      "loss": 0.6603,
+      "step": 14323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.73064397499359e-06,
+      "loss": 0.6976,
+      "step": 14324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.725656321603413e-06,
+      "loss": 0.6566,
+      "step": 14325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.720670142984393e-06,
+      "loss": 0.6338,
+      "step": 14326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.715685439310477e-06,
+      "loss": 0.6655,
+      "step": 14327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.710702210755544e-06,
+      "loss": 0.6141,
+      "step": 14328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.705720457493465e-06,
+      "loss": 18.3689,
+      "step": 14329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.700740179698018e-06,
+      "loss": 15.9945,
+      "step": 14330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.695761377542942e-06,
+      "loss": 0.6697,
+      "step": 14331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.690784051201933e-06,
+      "loss": 16.8438,
+      "step": 14332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.68580820084861e-06,
+      "loss": 17.9512,
+      "step": 14333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.680833826656587e-06,
+      "loss": 0.6271,
+      "step": 14334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.675860928799388e-06,
+      "loss": 17.0114,
+      "step": 14335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.670889507450491e-06,
+      "loss": 17.7547,
+      "step": 14336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.66591956278333e-06,
+      "loss": 18.0301,
+      "step": 14337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.66095109497127e-06,
+      "loss": 16.0857,
+      "step": 14338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.655984104187674e-06,
+      "loss": 16.1171,
+      "step": 14339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.651018590605797e-06,
+      "loss": 0.6801,
+      "step": 14340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.646054554398863e-06,
+      "loss": 17.4457,
+      "step": 14341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.641091995740051e-06,
+      "loss": 17.6199,
+      "step": 14342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.636130914802475e-06,
+      "loss": 17.1435,
+      "step": 14343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.631171311759227e-06,
+      "loss": 0.6337,
+      "step": 14344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.62621318678331e-06,
+      "loss": 0.6755,
+      "step": 14345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.621256540047694e-06,
+      "loss": 0.6659,
+      "step": 14346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.6163013717252985e-06,
+      "loss": 17.7709,
+      "step": 14347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.611347681988973e-06,
+      "loss": 17.8904,
+      "step": 14348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.606395471011557e-06,
+      "loss": 0.6549,
+      "step": 14349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.601444738965796e-06,
+      "loss": 16.4758,
+      "step": 14350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.596495486024402e-06,
+      "loss": 18.0343,
+      "step": 14351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.591547712360037e-06,
+      "loss": 17.6284,
+      "step": 14352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.586601418145289e-06,
+      "loss": 0.6686,
+      "step": 14353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.581656603552745e-06,
+      "loss": 18.2261,
+      "step": 14354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.576713268754887e-06,
+      "loss": 15.7531,
+      "step": 14355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.571771413924173e-06,
+      "loss": 17.0828,
+      "step": 14356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.5668310392329965e-06,
+      "loss": 18.006,
+      "step": 14357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.561892144853705e-06,
+      "loss": 17.974,
+      "step": 14358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.556954730958616e-06,
+      "loss": 18.0517,
+      "step": 14359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.552018797719956e-06,
+      "loss": 0.6247,
+      "step": 14360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.547084345309924e-06,
+      "loss": 0.6546,
+      "step": 14361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.542151373900658e-06,
+      "loss": 0.6589,
+      "step": 14362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.537219883664243e-06,
+      "loss": 18.0377,
+      "step": 14363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.532289874772741e-06,
+      "loss": 0.6124,
+      "step": 14364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.5273613473981315e-06,
+      "loss": 19.2852,
+      "step": 14365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.522434301712328e-06,
+      "loss": 17.8316,
+      "step": 14366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.517508737887219e-06,
+      "loss": 16.9681,
+      "step": 14367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.512584656094662e-06,
+      "loss": 17.8435,
+      "step": 14368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.5076620565064145e-06,
+      "loss": 19.7707,
+      "step": 14369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.5027409392942115e-06,
+      "loss": 16.2437,
+      "step": 14370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4978213046297315e-06,
+      "loss": 0.6853,
+      "step": 14371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4929031526845825e-06,
+      "loss": 17.6031,
+      "step": 14372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4879864836303716e-06,
+      "loss": 18.0239,
+      "step": 14373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.483071297638594e-06,
+      "loss": 18.4823,
+      "step": 14374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.47815759488073e-06,
+      "loss": 17.4506,
+      "step": 14375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4732453755281986e-06,
+      "loss": 17.2338,
+      "step": 14376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.468334639752345e-06,
+      "loss": 0.699,
+      "step": 14377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.463425387724515e-06,
+      "loss": 16.3422,
+      "step": 14378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.458517619615957e-06,
+      "loss": 17.3793,
+      "step": 14379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.453611335597882e-06,
+      "loss": 0.6673,
+      "step": 14380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4487065358414485e-06,
+      "loss": 17.1482,
+      "step": 14381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.44380322051777e-06,
+      "loss": 16.1735,
+      "step": 14382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.438901389797881e-06,
+      "loss": 17.1795,
+      "step": 14383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.434001043852817e-06,
+      "loss": 0.7257,
+      "step": 14384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.429102182853515e-06,
+      "loss": 15.6154,
+      "step": 14385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.424204806970869e-06,
+      "loss": 18.2042,
+      "step": 14386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.419308916375739e-06,
+      "loss": 0.6797,
+      "step": 14387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.414414511238909e-06,
+      "loss": 0.7253,
+      "step": 14388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4095215917311385e-06,
+      "loss": 14.9972,
+      "step": 14389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.404630158023113e-06,
+      "loss": 16.875,
+      "step": 14390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.399740210285472e-06,
+      "loss": 16.9734,
+      "step": 14391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.394851748688814e-06,
+      "loss": 0.6591,
+      "step": 14392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.389964773403651e-06,
+      "loss": 0.6363,
+      "step": 14393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3850792846005045e-06,
+      "loss": 18.1903,
+      "step": 14394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.380195282449787e-06,
+      "loss": 0.6521,
+      "step": 14395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3753127671218895e-06,
+      "loss": 0.6975,
+      "step": 14396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.370431738787131e-06,
+      "loss": 0.6752,
+      "step": 14397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.365552197615788e-06,
+      "loss": 0.641,
+      "step": 14398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.360674143778107e-06,
+      "loss": 18.3055,
+      "step": 14399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.355797577444246e-06,
+      "loss": 0.672,
+      "step": 14400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.350922498784335e-06,
+      "loss": 16.994,
+      "step": 14401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.346048907968445e-06,
+      "loss": 18.4971,
+      "step": 14402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.341176805166572e-06,
+      "loss": 15.9824,
+      "step": 14403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.33630619054872e-06,
+      "loss": 0.6609,
+      "step": 14404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.33143706428478e-06,
+      "loss": 17.805,
+      "step": 14405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.326569426544627e-06,
+      "loss": 0.6672,
+      "step": 14406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.32170327749806e-06,
+      "loss": 19.1779,
+      "step": 14407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.316838617314836e-06,
+      "loss": 17.614,
+      "step": 14408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3119754461646815e-06,
+      "loss": 16.5443,
+      "step": 14409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.307113764217238e-06,
+      "loss": 18.7571,
+      "step": 14410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.302253571642115e-06,
+      "loss": 16.3286,
+      "step": 14411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.2973948686088585e-06,
+      "loss": 0.6912,
+      "step": 14412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.292537655286957e-06,
+      "loss": 0.6784,
+      "step": 14413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.2876819318458855e-06,
+      "loss": 17.9415,
+      "step": 14414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.282827698455025e-06,
+      "loss": 0.657,
+      "step": 14415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.277974955283717e-06,
+      "loss": 16.5146,
+      "step": 14416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.27312370250125e-06,
+      "loss": 16.8627,
+      "step": 14417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.26827394027686e-06,
+      "loss": 0.6891,
+      "step": 14418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.263425668779755e-06,
+      "loss": 18.5397,
+      "step": 14419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.258578888179069e-06,
+      "loss": 17.472,
+      "step": 14420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.253733598643858e-06,
+      "loss": 0.646,
+      "step": 14421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.2488898003431614e-06,
+      "loss": 0.6634,
+      "step": 14422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.244047493445982e-06,
+      "loss": 0.7,
+      "step": 14423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.23920667812123e-06,
+      "loss": 0.6829,
+      "step": 14424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.23436735453778e-06,
+      "loss": 18.43,
+      "step": 14425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.229529522864459e-06,
+      "loss": 0.7126,
+      "step": 14426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.224693183270026e-06,
+      "loss": 17.2226,
+      "step": 14427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.219858335923224e-06,
+      "loss": 0.6625,
+      "step": 14428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.215024980992707e-06,
+      "loss": 0.6213,
+      "step": 14429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.210193118647096e-06,
+      "loss": 16.734,
+      "step": 14430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.2053627490549444e-06,
+      "loss": 0.7038,
+      "step": 14431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.200533872384757e-06,
+      "loss": 16.6916,
+      "step": 14432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.195706488805009e-06,
+      "loss": 0.6137,
+      "step": 14433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.190880598484112e-06,
+      "loss": 17.8687,
+      "step": 14434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.186056201590402e-06,
+      "loss": 17.2724,
+      "step": 14435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1812332982921944e-06,
+      "loss": 0.64,
+      "step": 14436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.176411888757722e-06,
+      "loss": 0.6422,
+      "step": 14437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.171591973155206e-06,
+      "loss": 0.5944,
+      "step": 14438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.166773551652784e-06,
+      "loss": 18.6989,
+      "step": 14439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.161956624418553e-06,
+      "loss": 18.7044,
+      "step": 14440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.157141191620548e-06,
+      "loss": 19.3641,
+      "step": 14441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.152327253426749e-06,
+      "loss": 17.4823,
+      "step": 14442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1475148100051185e-06,
+      "loss": 17.8562,
+      "step": 14443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1427038615235305e-06,
+      "loss": 16.9165,
+      "step": 14444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.13789440814982e-06,
+      "loss": 0.6288,
+      "step": 14445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.133086450051762e-06,
+      "loss": 17.7951,
+      "step": 14446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.128279987397085e-06,
+      "loss": 0.6605,
+      "step": 14447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.123475020353476e-06,
+      "loss": 16.6545,
+      "step": 14448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1186715490885625e-06,
+      "loss": 0.6325,
+      "step": 14449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.113869573769905e-06,
+      "loss": 0.6879,
+      "step": 14450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.109069094565024e-06,
+      "loss": 18.6398,
+      "step": 14451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1042701116413865e-06,
+      "loss": 0.666,
+      "step": 14452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0994726251664205e-06,
+      "loss": 18.1722,
+      "step": 14453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.09467663530749e-06,
+      "loss": 0.6452,
+      "step": 14454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.089882142231891e-06,
+      "loss": 17.4906,
+      "step": 14455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.085089146106899e-06,
+      "loss": 0.638,
+      "step": 14456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0802976470996986e-06,
+      "loss": 17.0385,
+      "step": 14457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.075507645377472e-06,
+      "loss": 0.6359,
+      "step": 14458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.070719141107313e-06,
+      "loss": 16.7686,
+      "step": 14459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.065932134456266e-06,
+      "loss": 0.7258,
+      "step": 14460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.061146625591331e-06,
+      "loss": 0.6603,
+      "step": 14461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.056362614679446e-06,
+      "loss": 17.4583,
+      "step": 14462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.051580101887523e-06,
+      "loss": 0.6799,
+      "step": 14463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0467990873823945e-06,
+      "loss": 18.167,
+      "step": 14464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.042019571330849e-06,
+      "loss": 18.9702,
+      "step": 14465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0372415538996265e-06,
+      "loss": 17.1975,
+      "step": 14466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.032465035255393e-06,
+      "loss": 0.6724,
+      "step": 14467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.027690015564809e-06,
+      "loss": 16.6893,
+      "step": 14468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.022916494994441e-06,
+      "loss": 17.1511,
+      "step": 14469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.018144473710825e-06,
+      "loss": 18.4191,
+      "step": 14470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.013373951880425e-06,
+      "loss": 0.6575,
+      "step": 14471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.008604929669655e-06,
+      "loss": 19.0077,
+      "step": 14472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.003837407244912e-06,
+      "loss": 15.9045,
+      "step": 14473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.99907138477251e-06,
+      "loss": 18.3486,
+      "step": 14474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.9943068624187034e-06,
+      "loss": 18.9947,
+      "step": 14475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.989543840349689e-06,
+      "loss": 16.5685,
+      "step": 14476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.984782318731664e-06,
+      "loss": 15.7998,
+      "step": 14477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.980022297730726e-06,
+      "loss": 0.6371,
+      "step": 14478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.975263777512925e-06,
+      "loss": 0.6229,
+      "step": 14479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.970506758244266e-06,
+      "loss": 0.6867,
+      "step": 14480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.965751240090696e-06,
+      "loss": 17.303,
+      "step": 14481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.96099722321813e-06,
+      "loss": 0.6398,
+      "step": 14482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.956244707792409e-06,
+      "loss": 18.844,
+      "step": 14483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.951493693979327e-06,
+      "loss": 16.1755,
+      "step": 14484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.9467441819446255e-06,
+      "loss": 0.636,
+      "step": 14485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.9419961718539875e-06,
+      "loss": 0.6065,
+      "step": 14486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.937249663873064e-06,
+      "loss": 17.9733,
+      "step": 14487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.932504658167438e-06,
+      "loss": 0.7065,
+      "step": 14488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.92776115490264e-06,
+      "loss": 16.6869,
+      "step": 14489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.923019154244148e-06,
+      "loss": 0.6945,
+      "step": 14490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.918278656357385e-06,
+      "loss": 17.2843,
+      "step": 14491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.913539661407742e-06,
+      "loss": 19.8896,
+      "step": 14492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.9088021695605295e-06,
+      "loss": 17.5228,
+      "step": 14493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.904066180981028e-06,
+      "loss": 15.3761,
+      "step": 14494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.899331695834454e-06,
+      "loss": 17.4457,
+      "step": 14495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.894598714285954e-06,
+      "loss": 17.2957,
+      "step": 14496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8898672365006685e-06,
+      "loss": 0.6827,
+      "step": 14497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.885137262643649e-06,
+      "loss": 18.3677,
+      "step": 14498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.880408792879905e-06,
+      "loss": 18.7616,
+      "step": 14499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8756818273743914e-06,
+      "loss": 17.1302,
+      "step": 14500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.870956366291997e-06,
+      "loss": 15.8177,
+      "step": 14501
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8662324097975964e-06,
+      "loss": 0.6557,
+      "step": 14502
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.861509958055978e-06,
+      "loss": 0.6394,
+      "step": 14503
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.856789011231896e-06,
+      "loss": 0.6816,
+      "step": 14504
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.852069569490027e-06,
+      "loss": 16.3762,
+      "step": 14505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8473516329950195e-06,
+      "loss": 0.6664,
+      "step": 14506
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.842635201911468e-06,
+      "loss": 17.5696,
+      "step": 14507
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.837920276403908e-06,
+      "loss": 17.7702,
+      "step": 14508
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.83320685663682e-06,
+      "loss": 14.6063,
+      "step": 14509
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.828494942774638e-06,
+      "loss": 0.6319,
+      "step": 14510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.823784534981725e-06,
+      "loss": 0.6724,
+      "step": 14511
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8190756334224285e-06,
+      "loss": 18.4935,
+      "step": 14512
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8143682382610154e-06,
+      "loss": 0.6638,
+      "step": 14513
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.809662349661705e-06,
+      "loss": 0.6758,
+      "step": 14514
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.804957967788666e-06,
+      "loss": 17.0109,
+      "step": 14515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.800255092806002e-06,
+      "loss": 19.434,
+      "step": 14516
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.7955537248777955e-06,
+      "loss": 18.4493,
+      "step": 14517
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.7908538641680505e-06,
+      "loss": 0.7076,
+      "step": 14518
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.786155510840725e-06,
+      "loss": 18.8436,
+      "step": 14519
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78145866505972e-06,
+      "loss": 17.5083,
+      "step": 14520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.776763326988883e-06,
+      "loss": 0.6687,
+      "step": 14521
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.772069496792033e-06,
+      "loss": 18.7417,
+      "step": 14522
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.767377174632905e-06,
+      "loss": 0.6188,
+      "step": 14523
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.762686360675202e-06,
+      "loss": 0.6977,
+      "step": 14524
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.757997055082555e-06,
+      "loss": 18.2451,
+      "step": 14525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.753309258018553e-06,
+      "loss": 0.7162,
+      "step": 14526
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.748622969646751e-06,
+      "loss": 16.7535,
+      "step": 14527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.7439381901306165e-06,
+      "loss": 17.9569,
+      "step": 14528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.739254919633603e-06,
+      "loss": 18.5506,
+      "step": 14529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.734573158319052e-06,
+      "loss": 16.683,
+      "step": 14530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.729892906350327e-06,
+      "loss": 16.2315,
+      "step": 14531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.725214163890681e-06,
+      "loss": 0.6208,
+      "step": 14532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.720536931103344e-06,
+      "loss": 17.3214,
+      "step": 14533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.715861208151486e-06,
+      "loss": 20.1599,
+      "step": 14534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.711186995198205e-06,
+      "loss": 18.4044,
+      "step": 14535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.70651429240659e-06,
+      "loss": 17.6866,
+      "step": 14536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.701843099939642e-06,
+      "loss": 17.3184,
+      "step": 14537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.697173417960312e-06,
+      "loss": 0.6889,
+      "step": 14538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.692505246631514e-06,
+      "loss": 17.0124,
+      "step": 14539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.687838586116086e-06,
+      "loss": 0.6814,
+      "step": 14540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.68317343657685e-06,
+      "loss": 17.0887,
+      "step": 14541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.678509798176541e-06,
+      "loss": 0.7316,
+      "step": 14542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.6738476710778595e-06,
+      "loss": 19.5062,
+      "step": 14543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.669187055443438e-06,
+      "loss": 0.7008,
+      "step": 14544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.664527951435856e-06,
+      "loss": 0.6829,
+      "step": 14545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.659870359217679e-06,
+      "loss": 0.632,
+      "step": 14546
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.655214278951372e-06,
+      "loss": 16.053,
+      "step": 14547
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.65055971079937e-06,
+      "loss": 16.3692,
+      "step": 14548
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.645906654924044e-06,
+      "loss": 17.0968,
+      "step": 14549
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.6412551114877205e-06,
+      "loss": 0.701,
+      "step": 14550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.636605080652686e-06,
+      "loss": 17.4591,
+      "step": 14551
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.631956562581148e-06,
+      "loss": 0.7374,
+      "step": 14552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.627309557435274e-06,
+      "loss": 17.835,
+      "step": 14553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.622664065377188e-06,
+      "loss": 19.3732,
+      "step": 14554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.618020086568927e-06,
+      "loss": 0.6456,
+      "step": 14555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.613377621172523e-06,
+      "loss": 17.5594,
+      "step": 14556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.6087366693499295e-06,
+      "loss": 18.7545,
+      "step": 14557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.604097231263046e-06,
+      "loss": 0.7009,
+      "step": 14558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.599459307073719e-06,
+      "loss": 16.6276,
+      "step": 14559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.594822896943736e-06,
+      "loss": 0.6714,
+      "step": 14560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.590188001034864e-06,
+      "loss": 18.2086,
+      "step": 14561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.585554619508788e-06,
+      "loss": 17.2607,
+      "step": 14562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.580922752527141e-06,
+      "loss": 17.0073,
+      "step": 14563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.57629240025151e-06,
+      "loss": 0.6796,
+      "step": 14564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.571663562843422e-06,
+      "loss": 0.622,
+      "step": 14565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.567036240464375e-06,
+      "loss": 0.6435,
+      "step": 14566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.562410433275784e-06,
+      "loss": 17.7984,
+      "step": 14567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.557786141439026e-06,
+      "loss": 15.7572,
+      "step": 14568
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.553163365115428e-06,
+      "loss": 16.3442,
+      "step": 14569
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.548542104466249e-06,
+      "loss": 17.2216,
+      "step": 14570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.543922359652699e-06,
+      "loss": 0.7197,
+      "step": 14571
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.539304130835966e-06,
+      "loss": 0.6492,
+      "step": 14572
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.5346874181771436e-06,
+      "loss": 0.66,
+      "step": 14573
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.5300722218372915e-06,
+      "loss": 17.2864,
+      "step": 14574
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.525458541977414e-06,
+      "loss": 17.4074,
+      "step": 14575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.520846378758449e-06,
+      "loss": 0.6356,
+      "step": 14576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.516235732341325e-06,
+      "loss": 0.7123,
+      "step": 14577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.5116266028868725e-06,
+      "loss": 17.3097,
+      "step": 14578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.50701899055588e-06,
+      "loss": 17.2241,
+      "step": 14579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.502412895509097e-06,
+      "loss": 16.6197,
+      "step": 14580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.4978083179071895e-06,
+      "loss": 16.8132,
+      "step": 14581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.493205257910817e-06,
+      "loss": 0.6324,
+      "step": 14582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.488603715680547e-06,
+      "loss": 17.6782,
+      "step": 14583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.484003691376927e-06,
+      "loss": 0.6737,
+      "step": 14584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.479405185160392e-06,
+      "loss": 0.7015,
+      "step": 14585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.474808197191401e-06,
+      "loss": 17.7174,
+      "step": 14586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.470212727630309e-06,
+      "loss": 16.69,
+      "step": 14587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.465618776637439e-06,
+      "loss": 18.184,
+      "step": 14588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.461026344373045e-06,
+      "loss": 0.6981,
+      "step": 14589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.456435430997326e-06,
+      "loss": 18.1861,
+      "step": 14590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.451846036670473e-06,
+      "loss": 17.3217,
+      "step": 14591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.4472581615525715e-06,
+      "loss": 17.6345,
+      "step": 14592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.4426718058036685e-06,
+      "loss": 18.2283,
+      "step": 14593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.438086969583773e-06,
+      "loss": 18.2921,
+      "step": 14594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.433503653052808e-06,
+      "loss": 17.419,
+      "step": 14595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.428921856370701e-06,
+      "loss": 17.8349,
+      "step": 14596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.424341579697268e-06,
+      "loss": 16.8202,
+      "step": 14597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.419762823192299e-06,
+      "loss": 0.6897,
+      "step": 14598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.4151855870155315e-06,
+      "loss": 0.6722,
+      "step": 14599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.410609871326629e-06,
+      "loss": 17.6642,
+      "step": 14600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.406035676285244e-06,
+      "loss": 16.2118,
+      "step": 14601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.40146300205094e-06,
+      "loss": 18.1499,
+      "step": 14602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.396891848783238e-06,
+      "loss": 19.3361,
+      "step": 14603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.392322216641599e-06,
+      "loss": 17.4022,
+      "step": 14604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.3877541057854375e-06,
+      "loss": 16.75,
+      "step": 14605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.383187516374134e-06,
+      "loss": 16.7795,
+      "step": 14606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.378622448566985e-06,
+      "loss": 17.6571,
+      "step": 14607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.374058902523239e-06,
+      "loss": 16.9723,
+      "step": 14608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.369496878402115e-06,
+      "loss": 18.8375,
+      "step": 14609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.36493637636274e-06,
+      "loss": 18.5142,
+      "step": 14610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.360377396564232e-06,
+      "loss": 17.8472,
+      "step": 14611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.355819939165624e-06,
+      "loss": 16.8581,
+      "step": 14612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.351264004325913e-06,
+      "loss": 16.5963,
+      "step": 14613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.3467095922040306e-06,
+      "loss": 18.1968,
+      "step": 14614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.342156702958851e-06,
+      "loss": 0.7022,
+      "step": 14615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.337605336749225e-06,
+      "loss": 17.9668,
+      "step": 14616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.333055493733919e-06,
+      "loss": 17.4977,
+      "step": 14617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.328507174071663e-06,
+      "loss": 0.6145,
+      "step": 14618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.3239603779211234e-06,
+      "loss": 17.0193,
+      "step": 14619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.319415105440912e-06,
+      "loss": 16.4975,
+      "step": 14620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.314871356789609e-06,
+      "loss": 17.982,
+      "step": 14621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.31032913212572e-06,
+      "loss": 17.313,
+      "step": 14622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.305788431607706e-06,
+      "loss": 0.6801,
+      "step": 14623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.301249255393965e-06,
+      "loss": 18.6267,
+      "step": 14624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.296711603642846e-06,
+      "loss": 17.0463,
+      "step": 14625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.2921754765126684e-06,
+      "loss": 16.4459,
+      "step": 14626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.287640874161665e-06,
+      "loss": 16.9235,
+      "step": 14627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.2831077967480274e-06,
+      "loss": 17.035,
+      "step": 14628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.2785762444299015e-06,
+      "loss": 0.5966,
+      "step": 14629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.274046217365359e-06,
+      "loss": 0.666,
+      "step": 14630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.269517715712459e-06,
+      "loss": 18.0104,
+      "step": 14631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.264990739629162e-06,
+      "loss": 16.0731,
+      "step": 14632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.260465289273399e-06,
+      "loss": 16.8709,
+      "step": 14633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.255941364803048e-06,
+      "loss": 17.9275,
+      "step": 14634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.25141896637591e-06,
+      "loss": 16.2201,
+      "step": 14635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.246898094149784e-06,
+      "loss": 17.9642,
+      "step": 14636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.242378748282363e-06,
+      "loss": 0.7042,
+      "step": 14637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.237860928931327e-06,
+      "loss": 17.5513,
+      "step": 14638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.233344636254246e-06,
+      "loss": 18.9013,
+      "step": 14639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.228829870408709e-06,
+      "loss": 18.6465,
+      "step": 14640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.224316631552208e-06,
+      "loss": 0.6355,
+      "step": 14641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.219804919842182e-06,
+      "loss": 0.6931,
+      "step": 14642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.215294735436034e-06,
+      "loss": 0.6365,
+      "step": 14643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.210786078491087e-06,
+      "loss": 0.644,
+      "step": 14644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.206278949164656e-06,
+      "loss": 16.93,
+      "step": 14645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.201773347613965e-06,
+      "loss": 17.6058,
+      "step": 14646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.197269273996198e-06,
+      "loss": 0.6108,
+      "step": 14647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.192766728468468e-06,
+      "loss": 15.4973,
+      "step": 14648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.188265711187857e-06,
+      "loss": 0.6563,
+      "step": 14649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.183766222311399e-06,
+      "loss": 17.3674,
+      "step": 14650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.179268261996052e-06,
+      "loss": 0.686,
+      "step": 14651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.174771830398729e-06,
+      "loss": 0.5923,
+      "step": 14652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.170276927676294e-06,
+      "loss": 0.6379,
+      "step": 14653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.165783553985538e-06,
+      "loss": 16.409,
+      "step": 14654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.161291709483247e-06,
+      "loss": 17.96,
+      "step": 14655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.156801394326106e-06,
+      "loss": 16.5892,
+      "step": 14656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.152312608670768e-06,
+      "loss": 0.6584,
+      "step": 14657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.147825352673819e-06,
+      "loss": 16.9319,
+      "step": 14658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.1433396264917895e-06,
+      "loss": 17.6335,
+      "step": 14659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.138855430281198e-06,
+      "loss": 18.1912,
+      "step": 14660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.134372764198465e-06,
+      "loss": 17.0281,
+      "step": 14661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.129891628399964e-06,
+      "loss": 0.6222,
+      "step": 14662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.125412023042032e-06,
+      "loss": 17.6838,
+      "step": 14663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.120933948280927e-06,
+      "loss": 16.2713,
+      "step": 14664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.116457404272896e-06,
+      "loss": 0.6873,
+      "step": 14665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.111982391174093e-06,
+      "loss": 17.8867,
+      "step": 14666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.10750890914063e-06,
+      "loss": 0.7092,
+      "step": 14667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.103036958328573e-06,
+      "loss": 0.6552,
+      "step": 14668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.098566538893908e-06,
+      "loss": 17.4149,
+      "step": 14669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.094097650992625e-06,
+      "loss": 0.6279,
+      "step": 14670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.089630294780602e-06,
+      "loss": 17.2372,
+      "step": 14671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.08516447041369e-06,
+      "loss": 0.6607,
+      "step": 14672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.080700178047688e-06,
+      "loss": 0.6842,
+      "step": 14673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.076237417838315e-06,
+      "loss": 0.6757,
+      "step": 14674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0717761899412865e-06,
+      "loss": 15.8147,
+      "step": 14675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0673164945122205e-06,
+      "loss": 17.5122,
+      "step": 14676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0628583317067e-06,
+      "loss": 17.7143,
+      "step": 14677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.05840170168025e-06,
+      "loss": 20.1515,
+      "step": 14678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0539466045883345e-06,
+      "loss": 17.9009,
+      "step": 14679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.04949304058639e-06,
+      "loss": 0.6636,
+      "step": 14680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.045041009829772e-06,
+      "loss": 16.7732,
+      "step": 14681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.040590512473799e-06,
+      "loss": 0.6913,
+      "step": 14682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.036141548673724e-06,
+      "loss": 0.6411,
+      "step": 14683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.031694118584741e-06,
+      "loss": 0.6754,
+      "step": 14684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.027248222362031e-06,
+      "loss": 0.6668,
+      "step": 14685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.02280386016067e-06,
+      "loss": 0.6199,
+      "step": 14686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.018361032135711e-06,
+      "loss": 18.6217,
+      "step": 14687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.013919738442142e-06,
+      "loss": 0.6269,
+      "step": 14688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.009479979234894e-06,
+      "loss": 0.6891,
+      "step": 14689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.005041754668872e-06,
+      "loss": 16.9241,
+      "step": 14690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0006050648988895e-06,
+      "loss": 0.6703,
+      "step": 14691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.99616991007973e-06,
+      "loss": 0.6867,
+      "step": 14692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.991736290366112e-06,
+      "loss": 17.66,
+      "step": 14693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.987304205912709e-06,
+      "loss": 0.7134,
+      "step": 14694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.982873656874138e-06,
+      "loss": 16.5245,
+      "step": 14695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.978444643404963e-06,
+      "loss": 0.6751,
+      "step": 14696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.97401716565969e-06,
+      "loss": 18.9415,
+      "step": 14697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.969591223792764e-06,
+      "loss": 17.9106,
+      "step": 14698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.965166817958612e-06,
+      "loss": 17.6271,
+      "step": 14699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.96074394831157e-06,
+      "loss": 18.4192,
+      "step": 14700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.956322615005927e-06,
+      "loss": 16.9607,
+      "step": 14701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.951902818195937e-06,
+      "loss": 18.3894,
+      "step": 14702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9474845580357705e-06,
+      "loss": 17.8087,
+      "step": 14703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.943067834679583e-06,
+      "loss": 0.7137,
+      "step": 14704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.93865264828144e-06,
+      "loss": 17.0386,
+      "step": 14705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.934238998995378e-06,
+      "loss": 0.6624,
+      "step": 14706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.929826886975359e-06,
+      "loss": 0.6518,
+      "step": 14707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.925416312375303e-06,
+      "loss": 0.6512,
+      "step": 14708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9210072753490944e-06,
+      "loss": 18.0246,
+      "step": 14709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.91659977605053e-06,
+      "loss": 0.6794,
+      "step": 14710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.912193814633371e-06,
+      "loss": 17.4889,
+      "step": 14711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.907789391251328e-06,
+      "loss": 18.4409,
+      "step": 14712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.903386506058034e-06,
+      "loss": 18.0119,
+      "step": 14713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8989851592071155e-06,
+      "loss": 0.6736,
+      "step": 14714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8945853508521e-06,
+      "loss": 0.632,
+      "step": 14715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.890187081146476e-06,
+      "loss": 0.627,
+      "step": 14716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.885790350243692e-06,
+      "loss": 18.6683,
+      "step": 14717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.881395158297104e-06,
+      "loss": 16.6227,
+      "step": 14718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.877001505460078e-06,
+      "loss": 18.3577,
+      "step": 14719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.87260939188587e-06,
+      "loss": 0.6705,
+      "step": 14720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8682188177277054e-06,
+      "loss": 0.6463,
+      "step": 14721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.863829783138752e-06,
+      "loss": 17.0866,
+      "step": 14722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8594422882721155e-06,
+      "loss": 0.6557,
+      "step": 14723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.855056333280879e-06,
+      "loss": 0.6322,
+      "step": 14724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.850671918318034e-06,
+      "loss": 17.6165,
+      "step": 14725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.846289043536535e-06,
+      "loss": 0.7094,
+      "step": 14726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.841907709089284e-06,
+      "loss": 0.6345,
+      "step": 14727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.837527915129115e-06,
+      "loss": 0.6485,
+      "step": 14728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.833149661808846e-06,
+      "loss": 16.9921,
+      "step": 14729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.828772949281197e-06,
+      "loss": 16.3229,
+      "step": 14730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.824397777698859e-06,
+      "loss": 16.8057,
+      "step": 14731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.820024147214459e-06,
+      "loss": 17.8429,
+      "step": 14732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.815652057980564e-06,
+      "loss": 0.5869,
+      "step": 14733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.811281510149724e-06,
+      "loss": 17.9159,
+      "step": 14734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.806912503874395e-06,
+      "loss": 17.7618,
+      "step": 14735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.802545039306989e-06,
+      "loss": 16.9743,
+      "step": 14736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.798179116599878e-06,
+      "loss": 0.7062,
+      "step": 14737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.793814735905345e-06,
+      "loss": 16.9316,
+      "step": 14738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.789451897375681e-06,
+      "loss": 17.0242,
+      "step": 14739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.785090601163068e-06,
+      "loss": 17.8742,
+      "step": 14740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.780730847419652e-06,
+      "loss": 0.681,
+      "step": 14741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.776372636297534e-06,
+      "loss": 17.2093,
+      "step": 14742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.772015967948729e-06,
+      "loss": 0.6264,
+      "step": 14743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.767660842525258e-06,
+      "loss": 16.4854,
+      "step": 14744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.763307260179035e-06,
+      "loss": 0.6525,
+      "step": 14745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7589552210619405e-06,
+      "loss": 16.8335,
+      "step": 14746
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.754604725325791e-06,
+      "loss": 0.7013,
+      "step": 14747
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.75025577312237e-06,
+      "loss": 0.6367,
+      "step": 14748
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.74590836460338e-06,
+      "loss": 0.6214,
+      "step": 14749
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.741562499920494e-06,
+      "loss": 16.4496,
+      "step": 14750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.737218179225317e-06,
+      "loss": 17.2358,
+      "step": 14751
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.732875402669397e-06,
+      "loss": 0.6665,
+      "step": 14752
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7285341704042485e-06,
+      "loss": 18.9525,
+      "step": 14753
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.724194482581308e-06,
+      "loss": 16.1151,
+      "step": 14754
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.719856339351975e-06,
+      "loss": 0.6567,
+      "step": 14755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.715519740867587e-06,
+      "loss": 16.8312,
+      "step": 14756
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.711184687279414e-06,
+      "loss": 0.7449,
+      "step": 14757
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7068511787387145e-06,
+      "loss": 0.6565,
+      "step": 14758
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.702519215396657e-06,
+      "loss": 0.5768,
+      "step": 14759
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.698188797404358e-06,
+      "loss": 18.2135,
+      "step": 14760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.693859924912892e-06,
+      "loss": 17.7757,
+      "step": 14761
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.689532598073261e-06,
+      "loss": 0.7082,
+      "step": 14762
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.685206817036454e-06,
+      "loss": 17.9504,
+      "step": 14763
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.680882581953362e-06,
+      "loss": 0.6197,
+      "step": 14764
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.676559892974842e-06,
+      "loss": 16.8839,
+      "step": 14765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.672238750251696e-06,
+      "loss": 0.6589,
+      "step": 14766
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.667919153934664e-06,
+      "loss": 17.3569,
+      "step": 14767
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.663601104174437e-06,
+      "loss": 0.6919,
+      "step": 14768
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6592846011216675e-06,
+      "loss": 16.5448,
+      "step": 14769
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.654969644926927e-06,
+      "loss": 18.8018,
+      "step": 14770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.650656235740754e-06,
+      "loss": 16.2968,
+      "step": 14771
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.646344373713619e-06,
+      "loss": 17.1635,
+      "step": 14772
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.642034058995937e-06,
+      "loss": 0.7118,
+      "step": 14773
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.637725291738094e-06,
+      "loss": 0.5947,
+      "step": 14774
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.633418072090396e-06,
+      "loss": 16.7909,
+      "step": 14775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.629112400203107e-06,
+      "loss": 19.2047,
+      "step": 14776
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.624808276226429e-06,
+      "loss": 17.9313,
+      "step": 14777
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.620505700310508e-06,
+      "loss": 0.6656,
+      "step": 14778
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.616204672605452e-06,
+      "loss": 18.4888,
+      "step": 14779
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.611905193261313e-06,
+      "loss": 17.177,
+      "step": 14780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6076072624280665e-06,
+      "loss": 19.1467,
+      "step": 14781
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.603310880255652e-06,
+      "loss": 0.7049,
+      "step": 14782
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.599016046893951e-06,
+      "loss": 17.5494,
+      "step": 14783
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.594722762492805e-06,
+      "loss": 0.644,
+      "step": 14784
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.59043102720197e-06,
+      "loss": 19.0213,
+      "step": 14785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.586140841171184e-06,
+      "loss": 0.6628,
+      "step": 14786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.581852204550098e-06,
+      "loss": 15.9573,
+      "step": 14787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.577565117488326e-06,
+      "loss": 17.4674,
+      "step": 14788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.573279580135438e-06,
+      "loss": 0.6743,
+      "step": 14789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.56899559264093e-06,
+      "loss": 0.638,
+      "step": 14790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.56471315515425e-06,
+      "loss": 16.7366,
+      "step": 14791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5604322678248e-06,
+      "loss": 18.4765,
+      "step": 14792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.556152930801911e-06,
+      "loss": 16.791,
+      "step": 14793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.551875144234886e-06,
+      "loss": 0.6295,
+      "step": 14794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5475989082729495e-06,
+      "loss": 17.7772,
+      "step": 14795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.543324223065283e-06,
+      "loss": 19.5206,
+      "step": 14796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5390510887610105e-06,
+      "loss": 0.6497,
+      "step": 14797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.534779505509191e-06,
+      "loss": 17.0416,
+      "step": 14798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.530509473458872e-06,
+      "loss": 17.9885,
+      "step": 14799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.526240992758997e-06,
+      "loss": 17.6376,
+      "step": 14800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.521974063558477e-06,
+      "loss": 0.6491,
+      "step": 14801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.517708686006167e-06,
+      "loss": 0.6762,
+      "step": 14802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.513444860250872e-06,
+      "loss": 17.3549,
+      "step": 14803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.50918258644133e-06,
+      "loss": 0.7583,
+      "step": 14804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.504921864726243e-06,
+      "loss": 0.651,
+      "step": 14805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.500662695254244e-06,
+      "loss": 0.6889,
+      "step": 14806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.496405078173911e-06,
+      "loss": 19.7551,
+      "step": 14807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.492149013633785e-06,
+      "loss": 16.628,
+      "step": 14808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.487894501782342e-06,
+      "loss": 0.6911,
+      "step": 14809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4836415427680035e-06,
+      "loss": 19.4946,
+      "step": 14810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.479390136739126e-06,
+      "loss": 0.6432,
+      "step": 14811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.475140283844027e-06,
+      "loss": 16.8174,
+      "step": 14812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.470891984230975e-06,
+      "loss": 16.8218,
+      "step": 14813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.466645238048174e-06,
+      "loss": 18.1138,
+      "step": 14814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.462400045443766e-06,
+      "loss": 16.2631,
+      "step": 14815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.458156406565851e-06,
+      "loss": 18.1714,
+      "step": 14816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.453914321562465e-06,
+      "loss": 16.1751,
+      "step": 14817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.449673790581611e-06,
+      "loss": 0.6818,
+      "step": 14818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.445434813771216e-06,
+      "loss": 0.6669,
+      "step": 14819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.441197391279157e-06,
+      "loss": 0.609,
+      "step": 14820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.436961523253265e-06,
+      "loss": 0.6488,
+      "step": 14821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.432727209841293e-06,
+      "loss": 0.764,
+      "step": 14822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.428494451190985e-06,
+      "loss": 16.112,
+      "step": 14823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.424263247449996e-06,
+      "loss": 19.3087,
+      "step": 14824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.420033598765928e-06,
+      "loss": 0.6669,
+      "step": 14825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.415805505286337e-06,
+      "loss": 16.6555,
+      "step": 14826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.411578967158715e-06,
+      "loss": 18.1185,
+      "step": 14827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.407353984530527e-06,
+      "loss": 18.3036,
+      "step": 14828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.403130557549152e-06,
+      "loss": 0.6267,
+      "step": 14829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.398908686361931e-06,
+      "loss": 17.8983,
+      "step": 14830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3946883711161465e-06,
+      "loss": 19.7977,
+      "step": 14831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.390469611959014e-06,
+      "loss": 0.6204,
+      "step": 14832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3862524090377345e-06,
+      "loss": 17.0213,
+      "step": 14833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.382036762499409e-06,
+      "loss": 18.2442,
+      "step": 14834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.377822672491112e-06,
+      "loss": 17.591,
+      "step": 14835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.373610139159851e-06,
+      "loss": 17.9577,
+      "step": 14836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.369399162652578e-06,
+      "loss": 18.5256,
+      "step": 14837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.365189743116206e-06,
+      "loss": 17.6082,
+      "step": 14838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.360981880697579e-06,
+      "loss": 0.612,
+      "step": 14839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.356775575543499e-06,
+      "loss": 17.5942,
+      "step": 14840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3525708278006905e-06,
+      "loss": 0.675,
+      "step": 14841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.348367637615842e-06,
+      "loss": 15.9822,
+      "step": 14842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.344166005135598e-06,
+      "loss": 0.6394,
+      "step": 14843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3399659305065295e-06,
+      "loss": 17.8595,
+      "step": 14844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.335767413875159e-06,
+      "loss": 0.652,
+      "step": 14845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.331570455387952e-06,
+      "loss": 16.7779,
+      "step": 14846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.327375055191314e-06,
+      "loss": 17.4923,
+      "step": 14847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.323181213431627e-06,
+      "loss": 17.8282,
+      "step": 14848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.318988930255181e-06,
+      "loss": 18.1767,
+      "step": 14849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.314798205808225e-06,
+      "loss": 15.7928,
+      "step": 14850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.310609040236964e-06,
+      "loss": 0.6605,
+      "step": 14851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3064214336875306e-06,
+      "loss": 0.6466,
+      "step": 14852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.30223538630602e-06,
+      "loss": 0.6744,
+      "step": 14853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.298050898238466e-06,
+      "loss": 0.6877,
+      "step": 14854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.293867969630844e-06,
+      "loss": 16.2064,
+      "step": 14855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.289686600629085e-06,
+      "loss": 16.1003,
+      "step": 14856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.285506791379047e-06,
+      "loss": 0.7028,
+      "step": 14857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.281328542026554e-06,
+      "loss": 0.6549,
+      "step": 14858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.277151852717366e-06,
+      "loss": 18.6627,
+      "step": 14859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.272976723597189e-06,
+      "loss": 15.7756,
+      "step": 14860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.268803154811669e-06,
+      "loss": 17.0831,
+      "step": 14861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.264631146506416e-06,
+      "loss": 17.6999,
+      "step": 14862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.260460698826969e-06,
+      "loss": 17.4308,
+      "step": 14863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.256291811918817e-06,
+      "loss": 0.5865,
+      "step": 14864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.25212448592739e-06,
+      "loss": 0.6831,
+      "step": 14865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.24795872099807e-06,
+      "loss": 0.6586,
+      "step": 14866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.24379451727619e-06,
+      "loss": 0.6667,
+      "step": 14867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.239631874907014e-06,
+      "loss": 17.4133,
+      "step": 14868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.235470794035762e-06,
+      "loss": 0.6767,
+      "step": 14869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.231311274807599e-06,
+      "loss": 0.642,
+      "step": 14870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.227153317367617e-06,
+      "loss": 0.6912,
+      "step": 14871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2229969218608895e-06,
+      "loss": 17.9601,
+      "step": 14872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.218842088432413e-06,
+      "loss": 0.6883,
+      "step": 14873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.21468881722712e-06,
+      "loss": 0.6863,
+      "step": 14874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.210537108389907e-06,
+      "loss": 18.7108,
+      "step": 14875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.206386962065602e-06,
+      "loss": 18.4796,
+      "step": 14876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.202238378398999e-06,
+      "loss": 0.6982,
+      "step": 14877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1980913575348235e-06,
+      "loss": 17.2143,
+      "step": 14878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.193945899617736e-06,
+      "loss": 0.6332,
+      "step": 14879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.189802004792365e-06,
+      "loss": 18.6846,
+      "step": 14880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.185659673203258e-06,
+      "loss": 18.0263,
+      "step": 14881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.181518904994936e-06,
+      "loss": 17.7933,
+      "step": 14882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1773797003118574e-06,
+      "loss": 18.2635,
+      "step": 14883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.173242059298411e-06,
+      "loss": 18.3018,
+      "step": 14884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.169105982098949e-06,
+      "loss": 0.6855,
+      "step": 14885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.164971468857738e-06,
+      "loss": 17.2016,
+      "step": 14886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1608385197190465e-06,
+      "loss": 0.6354,
+      "step": 14887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.156707134827044e-06,
+      "loss": 0.7035,
+      "step": 14888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.152577314325852e-06,
+      "loss": 18.4947,
+      "step": 14889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.148449058359545e-06,
+      "loss": 0.6844,
+      "step": 14890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1443223670721295e-06,
+      "loss": 0.6715,
+      "step": 14891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.140197240607586e-06,
+      "loss": 0.6387,
+      "step": 14892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.136073679109821e-06,
+      "loss": 0.6953,
+      "step": 14893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.131951682722674e-06,
+      "loss": 0.6503,
+      "step": 14894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1278312515899544e-06,
+      "loss": 18.8626,
+      "step": 14895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.123712385855395e-06,
+      "loss": 17.9613,
+      "step": 14896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.119595085662704e-06,
+      "loss": 0.6775,
+      "step": 14897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.115479351155506e-06,
+      "loss": 17.2365,
+      "step": 14898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.111365182477379e-06,
+      "loss": 0.6724,
+      "step": 14899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.107252579771854e-06,
+      "loss": 0.6819,
+      "step": 14900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.103141543182388e-06,
+      "loss": 0.7082,
+      "step": 14901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.099032072852422e-06,
+      "loss": 17.8765,
+      "step": 14902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.094924168925303e-06,
+      "loss": 17.3552,
+      "step": 14903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.090817831544342e-06,
+      "loss": 0.6426,
+      "step": 14904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.086713060852788e-06,
+      "loss": 18.8884,
+      "step": 14905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.082609856993831e-06,
+      "loss": 15.776,
+      "step": 14906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.078508220110634e-06,
+      "loss": 0.6931,
+      "step": 14907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0744081503462715e-06,
+      "loss": 0.6616,
+      "step": 14908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.070309647843791e-06,
+      "loss": 17.8323,
+      "step": 14909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.066212712746154e-06,
+      "loss": 19.5737,
+      "step": 14910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.062117345196293e-06,
+      "loss": 17.9745,
+      "step": 14911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.058023545337082e-06,
+      "loss": 17.8252,
+      "step": 14912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.053931313311333e-06,
+      "loss": 18.9726,
+      "step": 14913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0498406492618e-06,
+      "loss": 16.2318,
+      "step": 14914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.045751553331191e-06,
+      "loss": 19.0358,
+      "step": 14915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.041664025662169e-06,
+      "loss": 19.194,
+      "step": 14916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.037578066397319e-06,
+      "loss": 17.0105,
+      "step": 14917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.033493675679191e-06,
+      "loss": 18.0731,
+      "step": 14918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.029410853650257e-06,
+      "loss": 17.5337,
+      "step": 14919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.025329600452961e-06,
+      "loss": 17.757,
+      "step": 14920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.021249916229681e-06,
+      "loss": 17.1526,
+      "step": 14921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.017171801122733e-06,
+      "loss": 18.8388,
+      "step": 14922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.013095255274397e-06,
+      "loss": 18.4001,
+      "step": 14923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.009020278826876e-06,
+      "loss": 17.3459,
+      "step": 14924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.004946871922317e-06,
+      "loss": 17.9307,
+      "step": 14925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.000875034702851e-06,
+      "loss": 15.4152,
+      "step": 14926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.996804767310515e-06,
+      "loss": 18.9855,
+      "step": 14927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.992736069887299e-06,
+      "loss": 18.42,
+      "step": 14928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.988668942575148e-06,
+      "loss": 17.5235,
+      "step": 14929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.984603385515929e-06,
+      "loss": 17.2043,
+      "step": 14930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.980539398851503e-06,
+      "loss": 18.473,
+      "step": 14931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.976476982723621e-06,
+      "loss": 0.6863,
+      "step": 14932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9724161372740164e-06,
+      "loss": 0.6943,
+      "step": 14933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9683568626443525e-06,
+      "loss": 18.45,
+      "step": 14934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.964299158976221e-06,
+      "loss": 17.4517,
+      "step": 14935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.960243026411204e-06,
+      "loss": 17.9168,
+      "step": 14936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9561884650907995e-06,
+      "loss": 18.9837,
+      "step": 14937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.952135475156439e-06,
+      "loss": 16.8669,
+      "step": 14938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.94808405674953e-06,
+      "loss": 0.691,
+      "step": 14939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.944034210011383e-06,
+      "loss": 16.4796,
+      "step": 14940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.939985935083313e-06,
+      "loss": 0.6422,
+      "step": 14941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.935939232106529e-06,
+      "loss": 18.0985,
+      "step": 14942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.931894101222206e-06,
+      "loss": 16.9654,
+      "step": 14943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.927850542571466e-06,
+      "loss": 0.6759,
+      "step": 14944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9238085562953565e-06,
+      "loss": 0.6859,
+      "step": 14945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.919768142534903e-06,
+      "loss": 16.0468,
+      "step": 14946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.915729301431049e-06,
+      "loss": 0.6865,
+      "step": 14947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.911692033124693e-06,
+      "loss": 16.9627,
+      "step": 14948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.907656337756683e-06,
+      "loss": 0.6482,
+      "step": 14949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.903622215467802e-06,
+      "loss": 0.6614,
+      "step": 14950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.89958966639878e-06,
+      "loss": 0.6648,
+      "step": 14951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.895558690690305e-06,
+      "loss": 0.6486,
+      "step": 14952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8915292884829935e-06,
+      "loss": 0.681,
+      "step": 14953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.887501459917421e-06,
+      "loss": 0.598,
+      "step": 14954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.883475205134092e-06,
+      "loss": 16.1145,
+      "step": 14955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.87945052427346e-06,
+      "loss": 17.7738,
+      "step": 14956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.875427417475953e-06,
+      "loss": 17.7576,
+      "step": 14957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.871405884881902e-06,
+      "loss": 0.6832,
+      "step": 14958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.867385926631607e-06,
+      "loss": 18.8134,
+      "step": 14959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.863367542865304e-06,
+      "loss": 0.6321,
+      "step": 14960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8593507337231666e-06,
+      "loss": 17.3484,
+      "step": 14961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.855335499345348e-06,
+      "loss": 15.7663,
+      "step": 14962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.851321839871908e-06,
+      "loss": 0.6604,
+      "step": 14963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8473097554428696e-06,
+      "loss": 0.6995,
+      "step": 14964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.843299246198196e-06,
+      "loss": 19.225,
+      "step": 14965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.839290312277795e-06,
+      "loss": 18.7449,
+      "step": 14966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.835282953821524e-06,
+      "loss": 0.6538,
+      "step": 14967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.831277170969184e-06,
+      "loss": 0.625,
+      "step": 14968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.827272963860513e-06,
+      "loss": 0.666,
+      "step": 14969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.823270332635199e-06,
+      "loss": 17.5631,
+      "step": 14970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.819269277432886e-06,
+      "loss": 19.8737,
+      "step": 14971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.815269798393157e-06,
+      "loss": 0.6747,
+      "step": 14972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.811271895655528e-06,
+      "loss": 0.698,
+      "step": 14973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.807275569359465e-06,
+      "loss": 17.5358,
+      "step": 14974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.803280819644385e-06,
+      "loss": 15.3691,
+      "step": 14975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.799287646649658e-06,
+      "loss": 0.6642,
+      "step": 14976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.795296050514586e-06,
+      "loss": 15.929,
+      "step": 14977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.791306031378407e-06,
+      "loss": 19.5347,
+      "step": 14978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7873175893803306e-06,
+      "loss": 18.5972,
+      "step": 14979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.783330724659474e-06,
+      "loss": 17.6986,
+      "step": 14980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.779345437354949e-06,
+      "loss": 18.3493,
+      "step": 14981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.775361727605776e-06,
+      "loss": 0.6597,
+      "step": 14982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.771379595550923e-06,
+      "loss": 16.5083,
+      "step": 14983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7673990413293125e-06,
+      "loss": 16.1571,
+      "step": 14984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.763420065079804e-06,
+      "loss": 16.5971,
+      "step": 14985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.759442666941221e-06,
+      "loss": 0.6569,
+      "step": 14986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.755466847052309e-06,
+      "loss": 17.292,
+      "step": 14987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.751492605551772e-06,
+      "loss": 17.1747,
+      "step": 14988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7475199425782436e-06,
+      "loss": 18.0699,
+      "step": 14989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.743548858270319e-06,
+      "loss": 17.6172,
+      "step": 14990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.739579352766538e-06,
+      "loss": 17.7578,
+      "step": 14991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.735611426205372e-06,
+      "loss": 19.5997,
+      "step": 14992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.73164507872525e-06,
+      "loss": 18.4487,
+      "step": 14993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.727680310464544e-06,
+      "loss": 16.7879,
+      "step": 14994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.72371712156155e-06,
+      "loss": 18.2638,
+      "step": 14995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.719755512154544e-06,
+      "loss": 0.6793,
+      "step": 14996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.71579548238173e-06,
+      "loss": 0.7157,
+      "step": 14997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7118370323812495e-06,
+      "loss": 17.3451,
+      "step": 14998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.707880162291201e-06,
+      "loss": 0.7061,
+      "step": 14999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.70392487224961e-06,
+      "loss": 0.6361,
+      "step": 15000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.699971162394479e-06,
+      "loss": 0.6249,
+      "step": 15001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.696019032863725e-06,
+      "loss": 16.3235,
+      "step": 15002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6920684837952204e-06,
+      "loss": 16.2138,
+      "step": 15003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6881195153267885e-06,
+      "loss": 16.7533,
+      "step": 15004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.684172127596176e-06,
+      "loss": 15.6584,
+      "step": 15005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.68022632074111e-06,
+      "loss": 17.6791,
+      "step": 15006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6762820948992424e-06,
+      "loss": 16.9202,
+      "step": 15007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.672339450208157e-06,
+      "loss": 0.6502,
+      "step": 15008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.668398386805406e-06,
+      "loss": 18.017,
+      "step": 15009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.664458904828467e-06,
+      "loss": 0.6751,
+      "step": 15010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.660521004414781e-06,
+      "loss": 17.1852,
+      "step": 15011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.65658468570172e-06,
+      "loss": 0.6583,
+      "step": 15012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.65264994882661e-06,
+      "loss": 16.5544,
+      "step": 15013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.64871679392671e-06,
+      "loss": 0.624,
+      "step": 15014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.644785221139231e-06,
+      "loss": 16.5649,
+      "step": 15015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.640855230601332e-06,
+      "loss": 0.6932,
+      "step": 15016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6369268224501215e-06,
+      "loss": 18.4808,
+      "step": 15017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.632999996822635e-06,
+      "loss": 17.2847,
+      "step": 15018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.629074753855861e-06,
+      "loss": 0.6449,
+      "step": 15019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.625151093686736e-06,
+      "loss": 16.4895,
+      "step": 15020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.621229016452156e-06,
+      "loss": 17.7576,
+      "step": 15021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.617308522288921e-06,
+      "loss": 17.3566,
+      "step": 15022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6133896113338126e-06,
+      "loss": 17.4704,
+      "step": 15023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.609472283723531e-06,
+      "loss": 17.3468,
+      "step": 15024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.605556539594757e-06,
+      "loss": 0.6433,
+      "step": 15025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6016423790840865e-06,
+      "loss": 16.606,
+      "step": 15026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5977298023280655e-06,
+      "loss": 17.8811,
+      "step": 15027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5938188094631855e-06,
+      "loss": 18.0432,
+      "step": 15028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5899094006258744e-06,
+      "loss": 17.2965,
+      "step": 15029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.586001575952537e-06,
+      "loss": 17.7273,
+      "step": 15030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.582095335579489e-06,
+      "loss": 19.589,
+      "step": 15031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5781906796430075e-06,
+      "loss": 0.6761,
+      "step": 15032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5742876082792984e-06,
+      "loss": 0.7352,
+      "step": 15033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.570386121624526e-06,
+      "loss": 0.6671,
+      "step": 15034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.566486219814809e-06,
+      "loss": 16.4788,
+      "step": 15035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5625879029861874e-06,
+      "loss": 16.8126,
+      "step": 15036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.558691171274659e-06,
+      "loss": 18.0859,
+      "step": 15037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.554796024816166e-06,
+      "loss": 0.6647,
+      "step": 15038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.550902463746582e-06,
+      "loss": 0.7608,
+      "step": 15039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.547010488201753e-06,
+      "loss": 19.3122,
+      "step": 15040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.543120098317449e-06,
+      "loss": 18.3789,
+      "step": 15041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.539231294229385e-06,
+      "loss": 0.6708,
+      "step": 15042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.535344076073228e-06,
+      "loss": 17.7061,
+      "step": 15043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.531458443984571e-06,
+      "loss": 0.6536,
+      "step": 15044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.527574398098994e-06,
+      "loss": 0.6796,
+      "step": 15045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.523691938551982e-06,
+      "loss": 0.6683,
+      "step": 15046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5198110654789795e-06,
+      "loss": 0.62,
+      "step": 15047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.515931779015365e-06,
+      "loss": 0.6212,
+      "step": 15048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.512054079296469e-06,
+      "loss": 16.7971,
+      "step": 15049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.508177966457589e-06,
+      "loss": 17.0577,
+      "step": 15050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.504303440633928e-06,
+      "loss": 18.3723,
+      "step": 15051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.500430501960656e-06,
+      "loss": 0.7201,
+      "step": 15052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.496559150572882e-06,
+      "loss": 0.6118,
+      "step": 15053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.492689386605659e-06,
+      "loss": 0.7014,
+      "step": 15054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.488821210193994e-06,
+      "loss": 18.011,
+      "step": 15055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4849546214728255e-06,
+      "loss": 0.6489,
+      "step": 15056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.481089620577045e-06,
+      "loss": 15.8257,
+      "step": 15057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4772262076414885e-06,
+      "loss": 16.5996,
+      "step": 15058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.473364382800915e-06,
+      "loss": 17.423,
+      "step": 15059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.469504146190073e-06,
+      "loss": 0.6131,
+      "step": 15060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.465645497943621e-06,
+      "loss": 0.6512,
+      "step": 15061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.461788438196163e-06,
+      "loss": 17.3546,
+      "step": 15062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.457932967082268e-06,
+      "loss": 19.2952,
+      "step": 15063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.454079084736412e-06,
+      "loss": 17.2708,
+      "step": 15064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.450226791293071e-06,
+      "loss": 0.7013,
+      "step": 15065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.446376086886623e-06,
+      "loss": 0.6154,
+      "step": 15066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4425269716514005e-06,
+      "loss": 17.5105,
+      "step": 15067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.438679445721683e-06,
+      "loss": 16.0861,
+      "step": 15068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4348335092316895e-06,
+      "loss": 17.7476,
+      "step": 15069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.430989162315602e-06,
+      "loss": 0.6892,
+      "step": 15070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.427146405107524e-06,
+      "loss": 0.6854,
+      "step": 15071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.423305237741515e-06,
+      "loss": 16.684,
+      "step": 15072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.419465660351574e-06,
+      "loss": 0.6933,
+      "step": 15073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.415627673071637e-06,
+      "loss": 0.6831,
+      "step": 15074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.411791276035632e-06,
+      "loss": 0.692,
+      "step": 15075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.407956469377361e-06,
+      "loss": 0.6811,
+      "step": 15076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4041232532306075e-06,
+      "loss": 16.9135,
+      "step": 15077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.400291627729097e-06,
+      "loss": 0.6466,
+      "step": 15078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.396461593006512e-06,
+      "loss": 15.877,
+      "step": 15079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3926331491964565e-06,
+      "loss": 0.7021,
+      "step": 15080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.388806296432485e-06,
+      "loss": 0.667,
+      "step": 15081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.384981034848107e-06,
+      "loss": 0.6959,
+      "step": 15082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.38115736457676e-06,
+      "loss": 16.2359,
+      "step": 15083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.377335285751849e-06,
+      "loss": 17.9072,
+      "step": 15084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3735147985067004e-06,
+      "loss": 17.189,
+      "step": 15085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3696959029746025e-06,
+      "loss": 19.6664,
+      "step": 15086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3658785992887755e-06,
+      "loss": 0.7035,
+      "step": 15087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.362062887582374e-06,
+      "loss": 17.0603,
+      "step": 15088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.358248767988538e-06,
+      "loss": 15.9053,
+      "step": 15089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.354436240640314e-06,
+      "loss": 18.3663,
+      "step": 15090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.350625305670708e-06,
+      "loss": 0.6918,
+      "step": 15091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.346815963212658e-06,
+      "loss": 16.7072,
+      "step": 15092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.343008213399058e-06,
+      "loss": 17.1592,
+      "step": 15093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.339202056362757e-06,
+      "loss": 0.6642,
+      "step": 15094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.33539749223652e-06,
+      "loss": 0.6597,
+      "step": 15095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.331594521153087e-06,
+      "loss": 0.7203,
+      "step": 15096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.327793143245112e-06,
+      "loss": 18.1951,
+      "step": 15097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3239933586452115e-06,
+      "loss": 17.6312,
+      "step": 15098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.320195167485952e-06,
+      "loss": 17.299,
+      "step": 15099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.316398569899838e-06,
+      "loss": 15.495,
+      "step": 15100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.312603566019308e-06,
+      "loss": 0.6651,
+      "step": 15101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.308810155976756e-06,
+      "loss": 0.7083,
+      "step": 15102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.30501833990451e-06,
+      "loss": 17.8446,
+      "step": 15103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.301228117934863e-06,
+      "loss": 17.2908,
+      "step": 15104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.297439490200039e-06,
+      "loss": 15.2025,
+      "step": 15105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2936524568322035e-06,
+      "loss": 0.6474,
+      "step": 15106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.289867017963467e-06,
+      "loss": 16.2758,
+      "step": 15107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.286083173725885e-06,
+      "loss": 0.6755,
+      "step": 15108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.28230092425147e-06,
+      "loss": 0.6476,
+      "step": 15109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.278520269672159e-06,
+      "loss": 18.1344,
+      "step": 15110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.274741210119854e-06,
+      "loss": 18.3214,
+      "step": 15111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.270963745726381e-06,
+      "loss": 16.2102,
+      "step": 15112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.267187876623508e-06,
+      "loss": 15.816,
+      "step": 15113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.263413602942984e-06,
+      "loss": 17.8016,
+      "step": 15114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.259640924816466e-06,
+      "loss": 0.6579,
+      "step": 15115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.255869842375565e-06,
+      "loss": 0.6473,
+      "step": 15116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2521003557518426e-06,
+      "loss": 0.6858,
+      "step": 15117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.248332465076788e-06,
+      "loss": 17.7318,
+      "step": 15118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.244566170481862e-06,
+      "loss": 0.6573,
+      "step": 15119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.240801472098443e-06,
+      "loss": 0.7138,
+      "step": 15120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.237038370057877e-06,
+      "loss": 0.6717,
+      "step": 15121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.233276864491436e-06,
+      "loss": 17.0037,
+      "step": 15122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.229516955530333e-06,
+      "loss": 0.6724,
+      "step": 15123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.225758643305755e-06,
+      "loss": 16.178,
+      "step": 15124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.222001927948799e-06,
+      "loss": 15.7238,
+      "step": 15125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2182468095905255e-06,
+      "loss": 0.6809,
+      "step": 15126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.214493288361937e-06,
+      "loss": 17.8835,
+      "step": 15127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.210741364393967e-06,
+      "loss": 0.6816,
+      "step": 15128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.206991037817515e-06,
+      "loss": 0.6159,
+      "step": 15129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.203242308763428e-06,
+      "loss": 16.5108,
+      "step": 15130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1994951773624504e-06,
+      "loss": 17.0643,
+      "step": 15131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.195749643745311e-06,
+      "loss": 18.2198,
+      "step": 15132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.192005708042695e-06,
+      "loss": 0.6653,
+      "step": 15133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1882633703851985e-06,
+      "loss": 18.1799,
+      "step": 15134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1845226309033814e-06,
+      "loss": 0.6982,
+      "step": 15135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1807834897277355e-06,
+      "loss": 0.6871,
+      "step": 15136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1770459469887005e-06,
+      "loss": 16.4833,
+      "step": 15137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.173310002816677e-06,
+      "loss": 17.6961,
+      "step": 15138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1695756573419895e-06,
+      "loss": 18.1555,
+      "step": 15139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.165842910694912e-06,
+      "loss": 16.5414,
+      "step": 15140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.162111763005661e-06,
+      "loss": 17.268,
+      "step": 15141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1583822144044e-06,
+      "loss": 0.6373,
+      "step": 15142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.154654265021246e-06,
+      "loss": 0.6928,
+      "step": 15143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.150927914986253e-06,
+      "loss": 17.5511,
+      "step": 15144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.147203164429403e-06,
+      "loss": 0.7033,
+      "step": 15145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.143480013480649e-06,
+      "loss": 16.861,
+      "step": 15146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.139758462269866e-06,
+      "loss": 17.6261,
+      "step": 15147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.136038510926887e-06,
+      "loss": 17.0497,
+      "step": 15148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1323201595814875e-06,
+      "loss": 0.692,
+      "step": 15149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.128603408363385e-06,
+      "loss": 0.7065,
+      "step": 15150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1248882574022426e-06,
+      "loss": 18.4598,
+      "step": 15151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.121174706827663e-06,
+      "loss": 0.6857,
+      "step": 15152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1174627567691855e-06,
+      "loss": 15.7507,
+      "step": 15153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.113752407356325e-06,
+      "loss": 17.4337,
+      "step": 15154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1100436587185095e-06,
+      "loss": 16.3194,
+      "step": 15155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.10633651098512e-06,
+      "loss": 18.4111,
+      "step": 15156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.102630964285487e-06,
+      "loss": 0.6436,
+      "step": 15157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.098927018748866e-06,
+      "loss": 0.6522,
+      "step": 15158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.095224674504494e-06,
+      "loss": 0.7039,
+      "step": 15159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.091523931681524e-06,
+      "loss": 0.7535,
+      "step": 15160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.08782479040905e-06,
+      "loss": 16.3596,
+      "step": 15161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.084127250816133e-06,
+      "loss": 17.4025,
+      "step": 15162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.08043131303174e-06,
+      "loss": 0.5925,
+      "step": 15163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.076736977184836e-06,
+      "loss": 0.6203,
+      "step": 15164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.073044243404284e-06,
+      "loss": 17.7934,
+      "step": 15165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.069353111818913e-06,
+      "loss": 0.6466,
+      "step": 15166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.065663582557488e-06,
+      "loss": 16.1469,
+      "step": 15167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0619756557487175e-06,
+      "loss": 16.8241,
+      "step": 15168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.058289331521264e-06,
+      "loss": 17.7508,
+      "step": 15169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.054604610003726e-06,
+      "loss": 15.0737,
+      "step": 15170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.05092149132465e-06,
+      "loss": 17.847,
+      "step": 15171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0472399756125215e-06,
+      "loss": 17.9884,
+      "step": 15172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.043560062995761e-06,
+      "loss": 0.6374,
+      "step": 15173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.039881753602764e-06,
+      "loss": 17.9977,
+      "step": 15174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.036205047561848e-06,
+      "loss": 17.6189,
+      "step": 15175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0325299450012735e-06,
+      "loss": 0.698,
+      "step": 15176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.028856446049245e-06,
+      "loss": 17.1917,
+      "step": 15177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.02518455083391e-06,
+      "loss": 17.0853,
+      "step": 15178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.021514259483384e-06,
+      "loss": 17.883,
+      "step": 15179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.017845572125695e-06,
+      "loss": 17.0109,
+      "step": 15180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.014178488888837e-06,
+      "loss": 0.6595,
+      "step": 15181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.010513009900729e-06,
+      "loss": 18.063,
+      "step": 15182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.006849135289237e-06,
+      "loss": 16.4851,
+      "step": 15183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.003186865182207e-06,
+      "loss": 0.6986,
+      "step": 15184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.999526199707371e-06,
+      "loss": 0.7196,
+      "step": 15185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.995867138992448e-06,
+      "loss": 0.6526,
+      "step": 15186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.992209683165072e-06,
+      "loss": 0.7041,
+      "step": 15187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.988553832352854e-06,
+      "loss": 0.6642,
+      "step": 15188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9848995866833275e-06,
+      "loss": 18.4923,
+      "step": 15189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.981246946283973e-06,
+      "loss": 0.6073,
+      "step": 15190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.977595911282206e-06,
+      "loss": 18.3549,
+      "step": 15191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.973946481805396e-06,
+      "loss": 0.6945,
+      "step": 15192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.970298657980876e-06,
+      "loss": 17.8636,
+      "step": 15193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.966652439935886e-06,
+      "loss": 15.2991,
+      "step": 15194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.963007827797627e-06,
+      "loss": 0.6651,
+      "step": 15195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9593648216932515e-06,
+      "loss": 17.1222,
+      "step": 15196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.955723421749829e-06,
+      "loss": 18.9389,
+      "step": 15197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9520836280944215e-06,
+      "loss": 18.3705,
+      "step": 15198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.948445440853993e-06,
+      "loss": 16.4009,
+      "step": 15199
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.944808860155463e-06,
+      "loss": 16.9669,
+      "step": 15200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.941173886125693e-06,
+      "loss": 0.6722,
+      "step": 15201
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.937540518891486e-06,
+      "loss": 17.2991,
+      "step": 15202
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.933908758579613e-06,
+      "loss": 20.4816,
+      "step": 15203
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.930278605316767e-06,
+      "loss": 18.3286,
+      "step": 15204
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.926650059229575e-06,
+      "loss": 17.9286,
+      "step": 15205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.923023120444635e-06,
+      "loss": 15.5641,
+      "step": 15206
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.919397789088458e-06,
+      "loss": 17.5356,
+      "step": 15207
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9157740652875365e-06,
+      "loss": 17.8399,
+      "step": 15208
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.91215194916828e-06,
+      "loss": 0.6098,
+      "step": 15209
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.908531440857044e-06,
+      "loss": 0.6741,
+      "step": 15210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.904912540480138e-06,
+      "loss": 0.6719,
+      "step": 15211
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9012952481638e-06,
+      "loss": 16.0641,
+      "step": 15212
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8976795640342345e-06,
+      "loss": 17.3181,
+      "step": 15213
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.894065488217574e-06,
+      "loss": 0.6418,
+      "step": 15214
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8904530208398955e-06,
+      "loss": 17.6922,
+      "step": 15215
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.886842162027221e-06,
+      "loss": 0.6577,
+      "step": 15216
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8832329119055155e-06,
+      "loss": 0.6712,
+      "step": 15217
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.879625270600701e-06,
+      "loss": 17.1193,
+      "step": 15218
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8760192382386316e-06,
+      "loss": 17.9018,
+      "step": 15219
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.872414814945097e-06,
+      "loss": 0.7229,
+      "step": 15220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.868812000845845e-06,
+      "loss": 17.33,
+      "step": 15221
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.865210796066554e-06,
+      "loss": 0.6741,
+      "step": 15222
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.861611200732868e-06,
+      "loss": 0.6802,
+      "step": 15223
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.858013214970363e-06,
+      "loss": 0.7048,
+      "step": 15224
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.854416838904546e-06,
+      "loss": 17.1453,
+      "step": 15225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8508220726608814e-06,
+      "loss": 0.5999,
+      "step": 15226
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.84722891636477e-06,
+      "loss": 15.5778,
+      "step": 15227
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.843637370141579e-06,
+      "loss": 0.6582,
+      "step": 15228
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8400474341165945e-06,
+      "loss": 16.3452,
+      "step": 15229
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.836459108415047e-06,
+      "loss": 18.9783,
+      "step": 15230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.832872393162129e-06,
+      "loss": 0.699,
+      "step": 15231
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.829287288482946e-06,
+      "loss": 0.6821,
+      "step": 15232
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.825703794502588e-06,
+      "loss": 0.638,
+      "step": 15233
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.82212191134606e-06,
+      "loss": 0.6868,
+      "step": 15234
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.818541639138323e-06,
+      "loss": 0.6588,
+      "step": 15235
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8149629780042666e-06,
+      "loss": 17.3691,
+      "step": 15236
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.811385928068739e-06,
+      "loss": 0.7063,
+      "step": 15237
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.807810489456537e-06,
+      "loss": 18.418,
+      "step": 15238
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.804236662292393e-06,
+      "loss": 0.7635,
+      "step": 15239
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8006644467009688e-06,
+      "loss": 0.6763,
+      "step": 15240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7970938428068814e-06,
+      "loss": 16.7067,
+      "step": 15241
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7935248507347087e-06,
+      "loss": 18.4143,
+      "step": 15242
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7899574706089568e-06,
+      "loss": 0.6603,
+      "step": 15243
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.786391702554065e-06,
+      "loss": 17.3294,
+      "step": 15244
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7828275466944397e-06,
+      "loss": 0.7033,
+      "step": 15245
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.779265003154403e-06,
+      "loss": 0.6311,
+      "step": 15246
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.775704072058256e-06,
+      "loss": 0.725,
+      "step": 15247
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.772144753530216e-06,
+      "loss": 17.1374,
+      "step": 15248
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7685870476944496e-06,
+      "loss": 18.6615,
+      "step": 15249
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.765030954675075e-06,
+      "loss": 0.6795,
+      "step": 15250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7614764745961375e-06,
+      "loss": 17.0966,
+      "step": 15251
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7579236075816535e-06,
+      "loss": 0.6604,
+      "step": 15252
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7543723537555585e-06,
+      "loss": 17.0861,
+      "step": 15253
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7508227132417473e-06,
+      "loss": 16.5212,
+      "step": 15254
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7472746861640427e-06,
+      "loss": 0.6826,
+      "step": 15255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7437282726462187e-06,
+      "loss": 18.579,
+      "step": 15256
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.740183472812009e-06,
+      "loss": 16.4764,
+      "step": 15257
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.73664028678507e-06,
+      "loss": 18.4924,
+      "step": 15258
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.733098714689004e-06,
+      "loss": 17.7597,
+      "step": 15259
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.729558756647361e-06,
+      "loss": 18.5427,
+      "step": 15260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7260204127836316e-06,
+      "loss": 19.2492,
+      "step": 15261
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.722483683221267e-06,
+      "loss": 0.6634,
+      "step": 15262
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.718948568083641e-06,
+      "loss": 16.7897,
+      "step": 15263
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.715415067494077e-06,
+      "loss": 16.4109,
+      "step": 15264
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.711883181575848e-06,
+      "loss": 18.1714,
+      "step": 15265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7083529104521564e-06,
+      "loss": 0.6835,
+      "step": 15266
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7048242542461696e-06,
+      "loss": 17.9118,
+      "step": 15267
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7012972130809844e-06,
+      "loss": 18.3143,
+      "step": 15268
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.697771787079646e-06,
+      "loss": 17.2859,
+      "step": 15269
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.694247976365134e-06,
+      "loss": 0.7363,
+      "step": 15270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.690725781060378e-06,
+      "loss": 16.6881,
+      "step": 15271
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6872052012882683e-06,
+      "loss": 17.5257,
+      "step": 15272
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6836862371716075e-06,
+      "loss": 16.6626,
+      "step": 15273
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.680168888833163e-06,
+      "loss": 16.2749,
+      "step": 15274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.676653156395643e-06,
+      "loss": 16.8725,
+      "step": 15275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6731390399816823e-06,
+      "loss": 17.7607,
+      "step": 15276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.669626539713894e-06,
+      "loss": 16.8559,
+      "step": 15277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6661156557148023e-06,
+      "loss": 0.6709,
+      "step": 15278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6626063881068873e-06,
+      "loss": 0.6202,
+      "step": 15279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6590987370125727e-06,
+      "loss": 0.6676,
+      "step": 15280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6555927025542223e-06,
+      "loss": 0.6417,
+      "step": 15281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6520882848541603e-06,
+      "loss": 18.3863,
+      "step": 15282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.648585484034628e-06,
+      "loss": 0.649,
+      "step": 15283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.645084300217827e-06,
+      "loss": 18.63,
+      "step": 15284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6415847335258935e-06,
+      "loss": 18.0483,
+      "step": 15285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.638086784080913e-06,
+      "loss": 0.6512,
+      "step": 15286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6345904520049268e-06,
+      "loss": 18.3163,
+      "step": 15287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6310957374198983e-06,
+      "loss": 17.1217,
+      "step": 15288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.627602640447747e-06,
+      "loss": 18.2061,
+      "step": 15289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6241111612103194e-06,
+      "loss": 17.6677,
+      "step": 15290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6206212998294244e-06,
+      "loss": 0.5996,
+      "step": 15291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.617133056426819e-06,
+      "loss": 0.6547,
+      "step": 15292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6136464311241848e-06,
+      "loss": 17.5421,
+      "step": 15293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6101614240431627e-06,
+      "loss": 18.8684,
+      "step": 15294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.606678035305305e-06,
+      "loss": 16.0591,
+      "step": 15295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6031962650321605e-06,
+      "loss": 0.604,
+      "step": 15296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.599716113345186e-06,
+      "loss": 0.6844,
+      "step": 15297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.59623758036578e-06,
+      "loss": 17.2478,
+      "step": 15298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5927606662153e-06,
+      "loss": 17.695,
+      "step": 15299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.589285371015033e-06,
+      "loss": 16.204,
+      "step": 15300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.585811694886232e-06,
+      "loss": 0.6817,
+      "step": 15301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5823396379500716e-06,
+      "loss": 14.9775,
+      "step": 15302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.578869200327678e-06,
+      "loss": 17.8351,
+      "step": 15303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.575400382140115e-06,
+      "loss": 0.6414,
+      "step": 15304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5719331835083915e-06,
+      "loss": 17.569,
+      "step": 15305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.568467604553477e-06,
+      "loss": 0.7129,
+      "step": 15306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5650036453962697e-06,
+      "loss": 16.2718,
+      "step": 15307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.561541306157601e-06,
+      "loss": 17.4952,
+      "step": 15308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5580805869582623e-06,
+      "loss": 17.5162,
+      "step": 15309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5546214879189744e-06,
+      "loss": 20.5848,
+      "step": 15310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.551164009160429e-06,
+      "loss": 0.7237,
+      "step": 15311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5477081508032304e-06,
+      "loss": 18.1758,
+      "step": 15312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.544253912967949e-06,
+      "loss": 18.2553,
+      "step": 15313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.540801295775076e-06,
+      "loss": 17.4587,
+      "step": 15314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5373502993450546e-06,
+      "loss": 18.0864,
+      "step": 15315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5339009237982945e-06,
+      "loss": 0.6876,
+      "step": 15316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.530453169255116e-06,
+      "loss": 0.6841,
+      "step": 15317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5270070358357997e-06,
+      "loss": 16.7995,
+      "step": 15318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5235625236605728e-06,
+      "loss": 0.7079,
+      "step": 15319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.520119632849578e-06,
+      "loss": 0.6169,
+      "step": 15320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5166783635229515e-06,
+      "loss": 17.6659,
+      "step": 15321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.513238715800732e-06,
+      "loss": 0.6594,
+      "step": 15322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.509800689802911e-06,
+      "loss": 0.6523,
+      "step": 15323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5063642856494273e-06,
+      "loss": 18.0462,
+      "step": 15324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.502929503460156e-06,
+      "loss": 17.9728,
+      "step": 15325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.499496343354941e-06,
+      "loss": 0.628,
+      "step": 15326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4960648054535417e-06,
+      "loss": 16.2046,
+      "step": 15327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4926348898756623e-06,
+      "loss": 18.0195,
+      "step": 15328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.489206596740968e-06,
+      "loss": 0.6754,
+      "step": 15329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4857799261690415e-06,
+      "loss": 0.6933,
+      "step": 15330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4823548782794424e-06,
+      "loss": 16.9548,
+      "step": 15331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.478931453191653e-06,
+      "loss": 17.9113,
+      "step": 15332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4755096510250997e-06,
+      "loss": 17.0786,
+      "step": 15333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.472089471899148e-06,
+      "loss": 0.6872,
+      "step": 15334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.46867091593312e-06,
+      "loss": 0.706,
+      "step": 15335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4652539832462695e-06,
+      "loss": 17.508,
+      "step": 15336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4618386739578067e-06,
+      "loss": 0.6907,
+      "step": 15337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.45842498818687e-06,
+      "loss": 0.6825,
+      "step": 15338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4550129260525534e-06,
+      "loss": 0.671,
+      "step": 15339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4516024876738883e-06,
+      "loss": 0.6721,
+      "step": 15340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4481936731698416e-06,
+      "loss": 17.0135,
+      "step": 15341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4447864826593456e-06,
+      "loss": 0.6518,
+      "step": 15342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4413809162612555e-06,
+      "loss": 0.6318,
+      "step": 15343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.437976974094381e-06,
+      "loss": 17.4693,
+      "step": 15344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.434574656277462e-06,
+      "loss": 16.9364,
+      "step": 15345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.431173962929196e-06,
+      "loss": 16.8607,
+      "step": 15346
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.427774894168223e-06,
+      "loss": 15.7261,
+      "step": 15347
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4243774501131254e-06,
+      "loss": 18.5843,
+      "step": 15348
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.420981630882414e-06,
+      "loss": 0.6373,
+      "step": 15349
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4175874365945492e-06,
+      "loss": 17.9017,
+      "step": 15350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.414194867367959e-06,
+      "loss": 16.8587,
+      "step": 15351
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4108039233209822e-06,
+      "loss": 17.9395,
+      "step": 15352
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4074146045719236e-06,
+      "loss": 16.6446,
+      "step": 15353
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4040269112390164e-06,
+      "loss": 15.5617,
+      "step": 15354
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4006408434404336e-06,
+      "loss": 16.5088,
+      "step": 15355
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3972564012943187e-06,
+      "loss": 0.6512,
+      "step": 15356
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.393873584918733e-06,
+      "loss": 0.7112,
+      "step": 15357
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3904923944316824e-06,
+      "loss": 17.5836,
+      "step": 15358
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3871128299511333e-06,
+      "loss": 17.2454,
+      "step": 15359
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3837348915949694e-06,
+      "loss": 0.7091,
+      "step": 15360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3803585794810466e-06,
+      "loss": 0.6792,
+      "step": 15361
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.376983893727148e-06,
+      "loss": 15.4322,
+      "step": 15362
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3736108344509967e-06,
+      "loss": 0.681,
+      "step": 15363
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3702394017702643e-06,
+      "loss": 19.1478,
+      "step": 15364
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.366869595802563e-06,
+      "loss": 17.9362,
+      "step": 15365
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3635014166654656e-06,
+      "loss": 17.1733,
+      "step": 15366
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3601348644764608e-06,
+      "loss": 17.664,
+      "step": 15367
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.356769939352994e-06,
+      "loss": 17.2846,
+      "step": 15368
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.35340664141246e-06,
+      "loss": 0.6422,
+      "step": 15369
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.350044970772176e-06,
+      "loss": 17.331,
+      "step": 15370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.346684927549426e-06,
+      "loss": 16.9136,
+      "step": 15371
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3433265118614333e-06,
+      "loss": 0.6385,
+      "step": 15372
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3399697238253537e-06,
+      "loss": 18.5955,
+      "step": 15373
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.336614563558288e-06,
+      "loss": 17.4048,
+      "step": 15374
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.333261031177276e-06,
+      "loss": 16.4931,
+      "step": 15375
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.329909126799324e-06,
+      "loss": 17.5055,
+      "step": 15376
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3265588505413558e-06,
+      "loss": 16.7289,
+      "step": 15377
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3232102025202605e-06,
+      "loss": 17.664,
+      "step": 15378
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.319863182852839e-06,
+      "loss": 0.6197,
+      "step": 15379
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3165177916558654e-06,
+      "loss": 0.6532,
+      "step": 15380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3131740290460456e-06,
+      "loss": 0.6528,
+      "step": 15381
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.309831895140031e-06,
+      "loss": 16.7008,
+      "step": 15382
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3064913900544116e-06,
+      "loss": 17.0558,
+      "step": 15383
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3031525139057272e-06,
+      "loss": 0.6426,
+      "step": 15384
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.299815266810441e-06,
+      "loss": 0.6163,
+      "step": 15385
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2964796488849982e-06,
+      "loss": 17.7388,
+      "step": 15386
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.29314566024575e-06,
+      "loss": 0.7531,
+      "step": 15387
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2898133010090147e-06,
+      "loss": 0.6689,
+      "step": 15388
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.286482571291033e-06,
+      "loss": 17.3628,
+      "step": 15389
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2831534712080004e-06,
+      "loss": 16.1592,
+      "step": 15390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2798260008760686e-06,
+      "loss": 16.1631,
+      "step": 15391
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2765001604113053e-06,
+      "loss": 0.5939,
+      "step": 15392
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.273175949929741e-06,
+      "loss": 18.2801,
+      "step": 15393
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2698533695473433e-06,
+      "loss": 16.9327,
+      "step": 15394
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2665324193800086e-06,
+      "loss": 16.6833,
+      "step": 15395
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.263213099543616e-06,
+      "loss": 0.6846,
+      "step": 15396
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2598954101539457e-06,
+      "loss": 0.6508,
+      "step": 15397
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.256579351326744e-06,
+      "loss": 18.3416,
+      "step": 15398
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2532649231776903e-06,
+      "loss": 0.6694,
+      "step": 15399
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2499521258224085e-06,
+      "loss": 15.1894,
+      "step": 15400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2466409593764735e-06,
+      "loss": 0.6487,
+      "step": 15401
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.243331423955398e-06,
+      "loss": 17.3857,
+      "step": 15402
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2400235196746454e-06,
+      "loss": 17.0563,
+      "step": 15403
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2367172466495844e-06,
+      "loss": 0.675,
+      "step": 15404
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2334126049955837e-06,
+      "loss": 0.6954,
+      "step": 15405
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2301095948279235e-06,
+      "loss": 18.6423,
+      "step": 15406
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2268082162618284e-06,
+      "loss": 18.706,
+      "step": 15407
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2235084694124727e-06,
+      "loss": 17.1651,
+      "step": 15408
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.220210354394959e-06,
+      "loss": 16.8954,
+      "step": 15409
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.216913871324356e-06,
+      "loss": 15.8238,
+      "step": 15410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2136190203156667e-06,
+      "loss": 0.7083,
+      "step": 15411
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.210325801483827e-06,
+      "loss": 16.0769,
+      "step": 15412
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.207034214943727e-06,
+      "loss": 0.6724,
+      "step": 15413
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2037442608101874e-06,
+      "loss": 17.3313,
+      "step": 15414
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.200455939197988e-06,
+      "loss": 15.9497,
+      "step": 15415
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1971692502218486e-06,
+      "loss": 17.5214,
+      "step": 15416
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.193884193996427e-06,
+      "loss": 17.0724,
+      "step": 15417
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.190600770636315e-06,
+      "loss": 18.2317,
+      "step": 15418
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1873189802560544e-06,
+      "loss": 16.833,
+      "step": 15419
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1840388229701534e-06,
+      "loss": 18.4018,
+      "step": 15420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1807602988930263e-06,
+      "loss": 16.2789,
+      "step": 15421
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1774834081390536e-06,
+      "loss": 0.6553,
+      "step": 15422
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1742081508225496e-06,
+      "loss": 0.6726,
+      "step": 15423
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1709345270577672e-06,
+      "loss": 18.0475,
+      "step": 15424
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1676625369589207e-06,
+      "loss": 18.4141,
+      "step": 15425
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.164392180640152e-06,
+      "loss": 17.4894,
+      "step": 15426
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1611234582155537e-06,
+      "loss": 19.2373,
+      "step": 15427
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.157856369799145e-06,
+      "loss": 0.6872,
+      "step": 15428
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.154590915504907e-06,
+      "loss": 0.6917,
+      "step": 15429
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.15132709544676e-06,
+      "loss": 17.9439,
+      "step": 15430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1480649097385683e-06,
+      "loss": 0.6355,
+      "step": 15431
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.14480435849413e-06,
+      "loss": 16.2897,
+      "step": 15432
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1415454418271928e-06,
+      "loss": 17.177,
+      "step": 15433
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.138288159851438e-06,
+      "loss": 0.6065,
+      "step": 15434
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1350325126805137e-06,
+      "loss": 18.1338,
+      "step": 15435
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1317785004279843e-06,
+      "loss": 18.4674,
+      "step": 15436
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.128526123207376e-06,
+      "loss": 17.3708,
+      "step": 15437
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1252753811321476e-06,
+      "loss": 16.5083,
+      "step": 15438
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1220262743156982e-06,
+      "loss": 19.1862,
+      "step": 15439
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1187788028713806e-06,
+      "loss": 0.6794,
+      "step": 15440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.115532966912488e-06,
+      "loss": 17.4858,
+      "step": 15441
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1122887665522515e-06,
+      "loss": 0.6495,
+      "step": 15442
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1090462019038426e-06,
+      "loss": 0.7128,
+      "step": 15443
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1058052730803753e-06,
+      "loss": 17.6969,
+      "step": 15444
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1025659801949323e-06,
+      "loss": 0.6865,
+      "step": 15445
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.099328323360506e-06,
+      "loss": 0.6624,
+      "step": 15446
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.09609230269004e-06,
+      "loss": 0.6232,
+      "step": 15447
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.092857918296438e-06,
+      "loss": 16.5527,
+      "step": 15448
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.089625170292515e-06,
+      "loss": 17.8424,
+      "step": 15449
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0863940587910644e-06,
+      "loss": 19.321,
+      "step": 15450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.083164583904802e-06,
+      "loss": 18.2405,
+      "step": 15451
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.079936745746387e-06,
+      "loss": 18.1184,
+      "step": 15452
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0767105444284295e-06,
+      "loss": 17.6525,
+      "step": 15453
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0734859800634676e-06,
+      "loss": 0.5634,
+      "step": 15454
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0702630527640054e-06,
+      "loss": 17.6318,
+      "step": 15455
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.067041762642475e-06,
+      "loss": 16.931,
+      "step": 15456
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.063822109811254e-06,
+      "loss": 0.7047,
+      "step": 15457
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0606040943826465e-06,
+      "loss": 19.7911,
+      "step": 15458
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.057387716468929e-06,
+      "loss": 17.8046,
+      "step": 15459
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0541729761823125e-06,
+      "loss": 18.1065,
+      "step": 15460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0509598736349343e-06,
+      "loss": 17.3388,
+      "step": 15461
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.047748408938894e-06,
+      "loss": 0.5941,
+      "step": 15462
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.044538582206208e-06,
+      "loss": 16.332,
+      "step": 15463
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0413303935488803e-06,
+      "loss": 16.6152,
+      "step": 15464
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.038123843078816e-06,
+      "loss": 0.6691,
+      "step": 15465
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.034918930907882e-06,
+      "loss": 18.7679,
+      "step": 15466
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0317156571478768e-06,
+      "loss": 17.7944,
+      "step": 15467
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.02851402191055e-06,
+      "loss": 0.6993,
+      "step": 15468
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0253140253076017e-06,
+      "loss": 0.6505,
+      "step": 15469
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0221156674506647e-06,
+      "loss": 0.6866,
+      "step": 15470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0189189484513103e-06,
+      "loss": 19.337,
+      "step": 15471
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.015723868421061e-06,
+      "loss": 17.436,
+      "step": 15472
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0125304274713716e-06,
+      "loss": 0.6124,
+      "step": 15473
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0093386257136646e-06,
+      "loss": 16.294,
+      "step": 15474
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0061484632592785e-06,
+      "loss": 0.6536,
+      "step": 15475
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.002959940219502e-06,
+      "loss": 0.6875,
+      "step": 15476
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.999773056705574e-06,
+      "loss": 18.0216,
+      "step": 15477
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9965878128286605e-06,
+      "loss": 17.0821,
+      "step": 15478
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9934042086999017e-06,
+      "loss": 0.6347,
+      "step": 15479
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9902222444303406e-06,
+      "loss": 19.1809,
+      "step": 15480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9870419201309953e-06,
+      "loss": 17.6829,
+      "step": 15481
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9838632359128094e-06,
+      "loss": 18.0789,
+      "step": 15482
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9806861918866615e-06,
+      "loss": 0.6261,
+      "step": 15483
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.977510788163401e-06,
+      "loss": 0.6715,
+      "step": 15484
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9743370248538017e-06,
+      "loss": 17.0658,
+      "step": 15485
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9711649020685796e-06,
+      "loss": 17.1622,
+      "step": 15486
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9679944199183963e-06,
+      "loss": 17.5936,
+      "step": 15487
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9648255785138523e-06,
+      "loss": 16.2788,
+      "step": 15488
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9616583779655037e-06,
+      "loss": 18.3409,
+      "step": 15489
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.95849281838384e-06,
+      "loss": 0.7395,
+      "step": 15490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.955328899879284e-06,
+      "loss": 19.1815,
+      "step": 15491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9521666225622246e-06,
+      "loss": 18.0042,
+      "step": 15492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9490059865429633e-06,
+      "loss": 17.0017,
+      "step": 15493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.945846991931778e-06,
+      "loss": 17.2043,
+      "step": 15494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9426896388388634e-06,
+      "loss": 17.8354,
+      "step": 15495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9395339273743715e-06,
+      "loss": 17.8363,
+      "step": 15496
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9363798576483915e-06,
+      "loss": 0.6442,
+      "step": 15497
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9332274297709406e-06,
+      "loss": 17.6161,
+      "step": 15498
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9300766438520145e-06,
+      "loss": 0.6817,
+      "step": 15499
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9269275000015197e-06,
+      "loss": 0.709,
+      "step": 15500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.923779998329318e-06,
+      "loss": 0.6624,
+      "step": 15501
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9206341389452163e-06,
+      "loss": 17.8798,
+      "step": 15502
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9174899219589436e-06,
+      "loss": 16.3232,
+      "step": 15503
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.914347347480212e-06,
+      "loss": 0.6377,
+      "step": 15504
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9112064156186448e-06,
+      "loss": 16.8622,
+      "step": 15505
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9080671264838046e-06,
+      "loss": 0.7095,
+      "step": 15506
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.90492948018522e-06,
+      "loss": 0.7036,
+      "step": 15507
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9017934768323373e-06,
+      "loss": 16.4061,
+      "step": 15508
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8986591165345746e-06,
+      "loss": 18.6681,
+      "step": 15509
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.895526399401266e-06,
+      "loss": 17.3002,
+      "step": 15510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8923953255417025e-06,
+      "loss": 16.1868,
+      "step": 15511
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8892658950651077e-06,
+      "loss": 16.6986,
+      "step": 15512
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.886138108080666e-06,
+      "loss": 16.4811,
+      "step": 15513
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8830119646974795e-06,
+      "loss": 0.5739,
+      "step": 15514
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.879887465024617e-06,
+      "loss": 18.5196,
+      "step": 15515
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8767646091710676e-06,
+      "loss": 18.3877,
+      "step": 15516
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.873643397245773e-06,
+      "loss": 0.6934,
+      "step": 15517
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.870523829357641e-06,
+      "loss": 17.7364,
+      "step": 15518
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.867405905615478e-06,
+      "loss": 15.6077,
+      "step": 15519
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.864289626128064e-06,
+      "loss": 18.7139,
+      "step": 15520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8611749910041118e-06,
+      "loss": 15.3733,
+      "step": 15521
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8580620003522683e-06,
+      "loss": 17.5132,
+      "step": 15522
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8549506542811465e-06,
+      "loss": 0.6263,
+      "step": 15523
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8518409528992875e-06,
+      "loss": 0.6541,
+      "step": 15524
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8487328963151706e-06,
+      "loss": 0.6749,
+      "step": 15525
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.845626484637215e-06,
+      "loss": 18.2196,
+      "step": 15526
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.842521717973795e-06,
+      "loss": 0.6973,
+      "step": 15527
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8394185964332353e-06,
+      "loss": 17.2777,
+      "step": 15528
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.836317120123777e-06,
+      "loss": 17.7114,
+      "step": 15529
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8332172891536223e-06,
+      "loss": 16.4231,
+      "step": 15530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.830119103630907e-06,
+      "loss": 0.6346,
+      "step": 15531
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8270225636637172e-06,
+      "loss": 16.6755,
+      "step": 15532
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8239276693600712e-06,
+      "loss": 0.6549,
+      "step": 15533
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8208344208279502e-06,
+      "loss": 15.5571,
+      "step": 15534
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.817742818175251e-06,
+      "loss": 17.5539,
+      "step": 15535
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8146528615098367e-06,
+      "loss": 18.6237,
+      "step": 15536
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8115645509394996e-06,
+      "loss": 15.5248,
+      "step": 15537
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.808477886571964e-06,
+      "loss": 0.7184,
+      "step": 15538
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.805392868514933e-06,
+      "loss": 17.0624,
+      "step": 15539
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8023094968760154e-06,
+      "loss": 16.0621,
+      "step": 15540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7992277717627856e-06,
+      "loss": 0.6636,
+      "step": 15541
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.796147693282741e-06,
+      "loss": 0.6503,
+      "step": 15542
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.793069261543335e-06,
+      "loss": 17.7446,
+      "step": 15543
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7899924766519704e-06,
+      "loss": 0.6817,
+      "step": 15544
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7869173387159774e-06,
+      "loss": 0.6606,
+      "step": 15545
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.783843847842632e-06,
+      "loss": 0.6439,
+      "step": 15546
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7807720041391538e-06,
+      "loss": 0.6675,
+      "step": 15547
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.777701807712707e-06,
+      "loss": 17.6152,
+      "step": 15548
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.774633258670406e-06,
+      "loss": 0.6655,
+      "step": 15549
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.771566357119293e-06,
+      "loss": 0.6701,
+      "step": 15550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.76850110316636e-06,
+      "loss": 17.9761,
+      "step": 15551
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7654374969185436e-06,
+      "loss": 0.705,
+      "step": 15552
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7623755384827033e-06,
+      "loss": 0.679,
+      "step": 15553
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.759315227965681e-06,
+      "loss": 15.6266,
+      "step": 15554
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7562565654742246e-06,
+      "loss": 17.5019,
+      "step": 15555
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.753199551115043e-06,
+      "loss": 17.8983,
+      "step": 15556
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7501441849947795e-06,
+      "loss": 0.6167,
+      "step": 15557
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7470904672200204e-06,
+      "loss": 16.0642,
+      "step": 15558
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.744038397897303e-06,
+      "loss": 19.1953,
+      "step": 15559
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7409879771330973e-06,
+      "loss": 17.771,
+      "step": 15560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7379392050338236e-06,
+      "loss": 16.9326,
+      "step": 15561
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.734892081705831e-06,
+      "loss": 17.5584,
+      "step": 15562
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731846607255423e-06,
+      "loss": 16.0593,
+      "step": 15563
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7288027817888586e-06,
+      "loss": 0.5798,
+      "step": 15564
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.725760605412303e-06,
+      "loss": 18.3852,
+      "step": 15565
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7227200782318994e-06,
+      "loss": 0.6237,
+      "step": 15566
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7196812003537133e-06,
+      "loss": 18.1921,
+      "step": 15567
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.716643971883759e-06,
+      "loss": 15.4531,
+      "step": 15568
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7136083929279853e-06,
+      "loss": 0.6526,
+      "step": 15569
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7105744635923023e-06,
+      "loss": 0.7348,
+      "step": 15570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7075421839825423e-06,
+      "loss": 17.0338,
+      "step": 15571
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7045115542044863e-06,
+      "loss": 19.8863,
+      "step": 15572
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7014825743638726e-06,
+      "loss": 0.6574,
+      "step": 15573
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6984552445663613e-06,
+      "loss": 0.6484,
+      "step": 15574
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.695429564917562e-06,
+      "loss": 18.1376,
+      "step": 15575
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6924055355230294e-06,
+      "loss": 0.6882,
+      "step": 15576
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.689383156488251e-06,
+      "loss": 16.9912,
+      "step": 15577
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6863624279186818e-06,
+      "loss": 16.9883,
+      "step": 15578
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6833433499196925e-06,
+      "loss": 0.7466,
+      "step": 15579
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6803259225966104e-06,
+      "loss": 17.7742,
+      "step": 15580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6773101460546903e-06,
+      "loss": 17.8382,
+      "step": 15581
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6742960203991418e-06,
+      "loss": 0.6984,
+      "step": 15582
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.671283545735126e-06,
+      "loss": 16.0624,
+      "step": 15583
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.668272722167725e-06,
+      "loss": 18.0939,
+      "step": 15584
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.665263549801983e-06,
+      "loss": 17.4656,
+      "step": 15585
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6622560287428653e-06,
+      "loss": 18.5368,
+      "step": 15586
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6592501590952944e-06,
+      "loss": 0.6734,
+      "step": 15587
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6562459409641415e-06,
+      "loss": 0.63,
+      "step": 15588
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6532433744542005e-06,
+      "loss": 17.5222,
+      "step": 15589
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6502424596702267e-06,
+      "loss": 18.3835,
+      "step": 15590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.647243196716903e-06,
+      "loss": 0.7301,
+      "step": 15591
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.644245585698857e-06,
+      "loss": 0.6593,
+      "step": 15592
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.641249626720671e-06,
+      "loss": 0.6608,
+      "step": 15593
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.638255319886862e-06,
+      "loss": 0.6807,
+      "step": 15594
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6352626653018796e-06,
+      "loss": 15.0675,
+      "step": 15595
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6322716630701347e-06,
+      "loss": 17.439,
+      "step": 15596
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6292823132959555e-06,
+      "loss": 18.7282,
+      "step": 15597
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6262946160836467e-06,
+      "loss": 17.4728,
+      "step": 15598
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6233085715374308e-06,
+      "loss": 18.5532,
+      "step": 15599
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6203241797614685e-06,
+      "loss": 0.6643,
+      "step": 15600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6173414408598827e-06,
+      "loss": 16.9563,
+      "step": 15601
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.614360354936718e-06,
+      "loss": 18.3976,
+      "step": 15602
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6113809220959796e-06,
+      "loss": 17.7817,
+      "step": 15603
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.608403142441612e-06,
+      "loss": 19.5284,
+      "step": 15604
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6054270160774885e-06,
+      "loss": 0.5718,
+      "step": 15605
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6024525431074364e-06,
+      "loss": 0.6442,
+      "step": 15606
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.599479723635212e-06,
+      "loss": 0.66,
+      "step": 15607
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.596508557764543e-06,
+      "loss": 17.1649,
+      "step": 15608
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.593539045599075e-06,
+      "loss": 16.7843,
+      "step": 15609
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5905711872423967e-06,
+      "loss": 18.5168,
+      "step": 15610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.587604982798042e-06,
+      "loss": 17.7031,
+      "step": 15611
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5846404323694897e-06,
+      "loss": 15.8955,
+      "step": 15612
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5816775360601673e-06,
+      "loss": 16.9877,
+      "step": 15613
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.578716293973438e-06,
+      "loss": 17.5499,
+      "step": 15614
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.575756706212601e-06,
+      "loss": 19.9138,
+      "step": 15615
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5727987728809024e-06,
+      "loss": 0.6758,
+      "step": 15616
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5698424940815313e-06,
+      "loss": 17.5055,
+      "step": 15617
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.566887869917628e-06,
+      "loss": 0.6596,
+      "step": 15618
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.563934900492265e-06,
+      "loss": 0.6057,
+      "step": 15619
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56098358590845e-06,
+      "loss": 17.485,
+      "step": 15620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.558033926269149e-06,
+      "loss": 17.6144,
+      "step": 15621
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.555085921677264e-06,
+      "loss": 17.52,
+      "step": 15622
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5521395722356345e-06,
+      "loss": 17.3527,
+      "step": 15623
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5491948780470454e-06,
+      "loss": 17.0239,
+      "step": 15624
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.546251839214225e-06,
+      "loss": 0.6852,
+      "step": 15625
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5433104558398423e-06,
+      "loss": 16.0813,
+      "step": 15626
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5403707280265142e-06,
+      "loss": 18.0356,
+      "step": 15627
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.537432655876798e-06,
+      "loss": 0.6496,
+      "step": 15628
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5344962394931782e-06,
+      "loss": 16.3067,
+      "step": 15629
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.531561478978106e-06,
+      "loss": 0.633,
+      "step": 15630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5286283744339503e-06,
+      "loss": 0.6677,
+      "step": 15631
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5256969259630447e-06,
+      "loss": 0.5948,
+      "step": 15632
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5227671336676584e-06,
+      "loss": 16.0915,
+      "step": 15633
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.519838997649987e-06,
+      "loss": 0.6738,
+      "step": 15634
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5169125180121933e-06,
+      "loss": 16.9249,
+      "step": 15635
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.513987694856351e-06,
+      "loss": 0.6002,
+      "step": 15636
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5110645282845125e-06,
+      "loss": 0.6886,
+      "step": 15637
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.508143018398651e-06,
+      "loss": 17.5748,
+      "step": 15638
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.505223165300685e-06,
+      "loss": 16.6422,
+      "step": 15639
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5023049690924726e-06,
+      "loss": 17.569,
+      "step": 15640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4993884298758096e-06,
+      "loss": 15.6022,
+      "step": 15641
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.496473547752459e-06,
+      "loss": 0.6107,
+      "step": 15642
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4935603228241012e-06,
+      "loss": 17.7649,
+      "step": 15643
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4906487551923653e-06,
+      "loss": 0.6253,
+      "step": 15644
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4877388449588203e-06,
+      "loss": 18.0121,
+      "step": 15645
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.484830592224979e-06,
+      "loss": 0.7035,
+      "step": 15646
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4819239970923114e-06,
+      "loss": 19.3055,
+      "step": 15647
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.479019059662202e-06,
+      "loss": 18.8014,
+      "step": 15648
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.476115780035998e-06,
+      "loss": 18.6575,
+      "step": 15649
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4732141583149847e-06,
+      "loss": 17.3149,
+      "step": 15650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.470314194600376e-06,
+      "loss": 17.9458,
+      "step": 15651
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.467415888993352e-06,
+      "loss": 16.7649,
+      "step": 15652
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4645192415950204e-06,
+      "loss": 18.1211,
+      "step": 15653
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4616242525064225e-06,
+      "loss": 17.3812,
+      "step": 15654
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.458730921828567e-06,
+      "loss": 0.6853,
+      "step": 15655
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.455839249662373e-06,
+      "loss": 16.7454,
+      "step": 15656
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4529492361087312e-06,
+      "loss": 17.6828,
+      "step": 15657
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4500608812684624e-06,
+      "loss": 17.92,
+      "step": 15658
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4471741852423237e-06,
+      "loss": 17.667,
+      "step": 15659
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4442891481310183e-06,
+      "loss": 17.4685,
+      "step": 15660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4414057700351934e-06,
+      "loss": 18.5157,
+      "step": 15661
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.438524051055441e-06,
+      "loss": 16.5506,
+      "step": 15662
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4356439912922967e-06,
+      "loss": 0.6824,
+      "step": 15663
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.43276559084622e-06,
+      "loss": 18.1641,
+      "step": 15664
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.429888849817641e-06,
+      "loss": 0.6296,
+      "step": 15665
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.427013768306896e-06,
+      "loss": 19.2205,
+      "step": 15666
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4241403464143107e-06,
+      "loss": 0.6692,
+      "step": 15667
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.421268584240105e-06,
+      "loss": 15.4555,
+      "step": 15668
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4183984818844764e-06,
+      "loss": 17.7324,
+      "step": 15669
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4155300394475445e-06,
+      "loss": 18.184,
+      "step": 15670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4126632570293686e-06,
+      "loss": 16.5256,
+      "step": 15671
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4097981347299793e-06,
+      "loss": 0.6437,
+      "step": 15672
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.406934672649308e-06,
+      "loss": 17.4904,
+      "step": 15673
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.404072870887264e-06,
+      "loss": 0.6734,
+      "step": 15674
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4012127295436726e-06,
+      "loss": 16.9878,
+      "step": 15675
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3983542487183206e-06,
+      "loss": 0.6412,
+      "step": 15676
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3954974285109177e-06,
+      "loss": 17.3853,
+      "step": 15677
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3926422690211336e-06,
+      "loss": 17.8418,
+      "step": 15678
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3897887703485724e-06,
+      "loss": 18.1869,
+      "step": 15679
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3869369325927703e-06,
+      "loss": 18.0432,
+      "step": 15680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.384086755853232e-06,
+      "loss": 18.3175,
+      "step": 15681
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3812382402293778e-06,
+      "loss": 18.0015,
+      "step": 15682
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.378391385820583e-06,
+      "loss": 18.6242,
+      "step": 15683
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3755461927261637e-06,
+      "loss": 16.511,
+      "step": 15684
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3727026610453675e-06,
+      "loss": 17.5736,
+      "step": 15685
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.369860790877404e-06,
+      "loss": 0.6883,
+      "step": 15686
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3670205823214163e-06,
+      "loss": 15.6832,
+      "step": 15687
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3641820354764754e-06,
+      "loss": 0.6322,
+      "step": 15688
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3613451504416127e-06,
+      "loss": 16.2824,
+      "step": 15689
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3585099273157883e-06,
+      "loss": 18.2693,
+      "step": 15690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.355676366197923e-06,
+      "loss": 0.7174,
+      "step": 15691
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3528444671868654e-06,
+      "loss": 17.1906,
+      "step": 15692
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.350014230381398e-06,
+      "loss": 15.974,
+      "step": 15693
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.347185655880263e-06,
+      "loss": 0.7243,
+      "step": 15694
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3443587437821323e-06,
+      "loss": 17.8146,
+      "step": 15695
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3415334941856325e-06,
+      "loss": 0.69,
+      "step": 15696
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3387099071893235e-06,
+      "loss": 17.5935,
+      "step": 15697
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.335887982891705e-06,
+      "loss": 16.1316,
+      "step": 15698
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.333067721391219e-06,
+      "loss": 17.7184,
+      "step": 15699
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3302491227862544e-06,
+      "loss": 0.6307,
+      "step": 15700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3274321871751435e-06,
+      "loss": 16.0532,
+      "step": 15701
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3246169146561527e-06,
+      "loss": 17.1372,
+      "step": 15702
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3218033053275023e-06,
+      "loss": 18.2033,
+      "step": 15703
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3189913592873367e-06,
+      "loss": 16.4606,
+      "step": 15704
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3161810766337555e-06,
+      "loss": 0.6434,
+      "step": 15705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3133724574648073e-06,
+      "loss": 16.5129,
+      "step": 15706
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3105655018784588e-06,
+      "loss": 17.1298,
+      "step": 15707
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3077602099726426e-06,
+      "loss": 16.3913,
+      "step": 15708
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3049565818452244e-06,
+      "loss": 18.2718,
+      "step": 15709
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3021546175939935e-06,
+      "loss": 19.4485,
+      "step": 15710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.299354317316721e-06,
+      "loss": 17.1255,
+      "step": 15711
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2965556811110844e-06,
+      "loss": 0.6702,
+      "step": 15712
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2937587090747227e-06,
+      "loss": 18.7497,
+      "step": 15713
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2909634013052074e-06,
+      "loss": 16.7643,
+      "step": 15714
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2881697579000494e-06,
+      "loss": 17.6318,
+      "step": 15715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.28537777895671e-06,
+      "loss": 0.6987,
+      "step": 15716
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.282587464572594e-06,
+      "loss": 17.0741,
+      "step": 15717
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2797988148450466e-06,
+      "loss": 0.7317,
+      "step": 15718
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.277011829871345e-06,
+      "loss": 18.6801,
+      "step": 15719
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2742265097487116e-06,
+      "loss": 17.3029,
+      "step": 15720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.271442854574318e-06,
+      "loss": 0.7571,
+      "step": 15721
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2686608644452766e-06,
+      "loss": 18.0971,
+      "step": 15722
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2658805394586425e-06,
+      "loss": 0.648,
+      "step": 15723
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.263101879711399e-06,
+      "loss": 0.7154,
+      "step": 15724
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.260324885300491e-06,
+      "loss": 0.6593,
+      "step": 15725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2575495563227854e-06,
+      "loss": 18.0093,
+      "step": 15726
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.254775892875116e-06,
+      "loss": 0.6494,
+      "step": 15727
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2520038950542334e-06,
+      "loss": 17.1266,
+      "step": 15728
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.249233562956848e-06,
+      "loss": 0.5868,
+      "step": 15729
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2464648966795954e-06,
+      "loss": 18.4622,
+      "step": 15730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2436978963190746e-06,
+      "loss": 18.9853,
+      "step": 15731
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2409325619718037e-06,
+      "loss": 16.39,
+      "step": 15732
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.238168893734255e-06,
+      "loss": 0.6792,
+      "step": 15733
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.235406891702846e-06,
+      "loss": 18.6388,
+      "step": 15734
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.232646555973922e-06,
+      "loss": 17.7987,
+      "step": 15735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.229887886643789e-06,
+      "loss": 0.6497,
+      "step": 15736
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2271308838086866e-06,
+      "loss": 18.4724,
+      "step": 15737
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2243755475647886e-06,
+      "loss": 0.6488,
+      "step": 15738
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2216218780082175e-06,
+      "loss": 17.3045,
+      "step": 15739
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.21886987523503e-06,
+      "loss": 0.6963,
+      "step": 15740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2161195393412493e-06,
+      "loss": 17.7164,
+      "step": 15741
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.213370870422815e-06,
+      "loss": 16.7696,
+      "step": 15742
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.210623868575612e-06,
+      "loss": 16.8815,
+      "step": 15743
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2078785338954744e-06,
+      "loss": 0.7072,
+      "step": 15744
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2051348664781703e-06,
+      "loss": 0.659,
+      "step": 15745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.202392866419423e-06,
+      "loss": 19.22,
+      "step": 15746
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1996525338148833e-06,
+      "loss": 17.3821,
+      "step": 15747
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1969138687601587e-06,
+      "loss": 17.161,
+      "step": 15748
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.194176871350778e-06,
+      "loss": 0.6612,
+      "step": 15749
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.191441541682221e-06,
+      "loss": 18.0984,
+      "step": 15750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.188707879849927e-06,
+      "loss": 17.4537,
+      "step": 15751
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.185975885949254e-06,
+      "loss": 16.9267,
+      "step": 15752
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.183245560075503e-06,
+      "loss": 0.7082,
+      "step": 15753
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.180516902323937e-06,
+      "loss": 17.2252,
+      "step": 15754
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1777899127897294e-06,
+      "loss": 0.666,
+      "step": 15755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1750645915680323e-06,
+      "loss": 0.746,
+      "step": 15756
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1723409387539084e-06,
+      "loss": 16.9334,
+      "step": 15757
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.169618954442382e-06,
+      "loss": 0.625,
+      "step": 15758
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.166898638728404e-06,
+      "loss": 16.1285,
+      "step": 15759
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1641799917068718e-06,
+      "loss": 18.0893,
+      "step": 15760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1614630134726367e-06,
+      "loss": 16.4552,
+      "step": 15761
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1587477041204786e-06,
+      "loss": 0.6716,
+      "step": 15762
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.156034063745127e-06,
+      "loss": 18.211,
+      "step": 15763
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1533220924412456e-06,
+      "loss": 0.6913,
+      "step": 15764
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.150611790303436e-06,
+      "loss": 17.7749,
+      "step": 15765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.147903157426262e-06,
+      "loss": 16.7878,
+      "step": 15766
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1451961939042143e-06,
+      "loss": 17.1747,
+      "step": 15767
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1424908998317173e-06,
+      "loss": 0.6444,
+      "step": 15768
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1397872753031623e-06,
+      "loss": 17.1848,
+      "step": 15769
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.137085320412846e-06,
+      "loss": 0.6924,
+      "step": 15770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1343850352550487e-06,
+      "loss": 0.7127,
+      "step": 15771
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1316864199239616e-06,
+      "loss": 0.642,
+      "step": 15772
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.128989474513732e-06,
+      "loss": 0.6079,
+      "step": 15773
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1262941991184395e-06,
+      "loss": 17.0401,
+      "step": 15774
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.123600593832109e-06,
+      "loss": 17.6253,
+      "step": 15775
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1209086587487214e-06,
+      "loss": 16.9565,
+      "step": 15776
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.118218393962179e-06,
+      "loss": 17.1181,
+      "step": 15777
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.115529799566335e-06,
+      "loss": 0.6614,
+      "step": 15778
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.11284287565498e-06,
+      "loss": 20.8094,
+      "step": 15779
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1101576223218454e-06,
+      "loss": 18.1333,
+      "step": 15780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.107474039660623e-06,
+      "loss": 0.6602,
+      "step": 15781
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.104792127764921e-06,
+      "loss": 17.0548,
+      "step": 15782
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1021118867282974e-06,
+      "loss": 16.9111,
+      "step": 15783
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0994333166442616e-06,
+      "loss": 0.6603,
+      "step": 15784
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0967564176062547e-06,
+      "loss": 17.4821,
+      "step": 15785
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0940811897076586e-06,
+      "loss": 17.2354,
+      "step": 15786
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0914076330418032e-06,
+      "loss": 0.6927,
+      "step": 15787
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0887357477019643e-06,
+      "loss": 16.6107,
+      "step": 15788
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.086065533781334e-06,
+      "loss": 0.6739,
+      "step": 15789
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.083396991373082e-06,
+      "loss": 19.3741,
+      "step": 15790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.080730120570301e-06,
+      "loss": 0.6547,
+      "step": 15791
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.078064921466022e-06,
+      "loss": 18.9564,
+      "step": 15792
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.075401394153226e-06,
+      "loss": 0.6155,
+      "step": 15793
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.072739538724816e-06,
+      "loss": 15.0869,
+      "step": 15794
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.07007935527368e-06,
+      "loss": 16.4638,
+      "step": 15795
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0674208438926035e-06,
+      "loss": 16.8539,
+      "step": 15796
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.06476400467433e-06,
+      "loss": 16.353,
+      "step": 15797
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0621088377115516e-06,
+      "loss": 17.7007,
+      "step": 15798
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.059455343096889e-06,
+      "loss": 17.1176,
+      "step": 15799
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.056803520922923e-06,
+      "loss": 17.3164,
+      "step": 15800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0541533712821527e-06,
+      "loss": 0.6656,
+      "step": 15801
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.051504894267037e-06,
+      "loss": 17.415,
+      "step": 15802
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0488580899699627e-06,
+      "loss": 16.9695,
+      "step": 15803
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.046212958483268e-06,
+      "loss": 17.1195,
+      "step": 15804
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.043569499899239e-06,
+      "loss": 16.2029,
+      "step": 15805
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.040927714310087e-06,
+      "loss": 18.2472,
+      "step": 15806
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0382876018079697e-06,
+      "loss": 20.3571,
+      "step": 15807
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0356491624849982e-06,
+      "loss": 16.7302,
+      "step": 15808
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.033012396433204e-06,
+      "loss": 0.6882,
+      "step": 15809
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.030377303744585e-06,
+      "loss": 17.1463,
+      "step": 15810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0277438845110684e-06,
+      "loss": 0.7258,
+      "step": 15811
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0251121388245142e-06,
+      "loss": 18.8651,
+      "step": 15812
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0224820667767373e-06,
+      "loss": 0.6561,
+      "step": 15813
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0198536684594815e-06,
+      "loss": 17.6753,
+      "step": 15814
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0172269439644507e-06,
+      "loss": 0.6954,
+      "step": 15815
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0146018933832834e-06,
+      "loss": 16.0772,
+      "step": 15816
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.01197851680755e-06,
+      "loss": 0.6411,
+      "step": 15817
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0093568143287665e-06,
+      "loss": 16.8628,
+      "step": 15818
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.006736786038388e-06,
+      "loss": 0.7008,
+      "step": 15819
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0041184320278296e-06,
+      "loss": 17.745,
+      "step": 15820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.001501752388435e-06,
+      "loss": 0.7091,
+      "step": 15821
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9988867472114757e-06,
+      "loss": 0.6121,
+      "step": 15822
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9962734165881847e-06,
+      "loss": 0.716,
+      "step": 15823
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9936617606097276e-06,
+      "loss": 16.4976,
+      "step": 15824
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.99105177936722e-06,
+      "loss": 0.6753,
+      "step": 15825
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9884434729517123e-06,
+      "loss": 17.8742,
+      "step": 15826
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9858368414541916e-06,
+      "loss": 0.6603,
+      "step": 15827
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9832318849655917e-06,
+      "loss": 17.9477,
+      "step": 15828
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9806286035767895e-06,
+      "loss": 0.6902,
+      "step": 15829
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9780269973786127e-06,
+      "loss": 18.1347,
+      "step": 15830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.975427066461805e-06,
+      "loss": 0.638,
+      "step": 15831
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9728288109170767e-06,
+      "loss": 15.7786,
+      "step": 15832
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9702322308350674e-06,
+      "loss": 18.3681,
+      "step": 15833
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.967637326306354e-06,
+      "loss": 0.6066,
+      "step": 15834
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9650440974214756e-06,
+      "loss": 16.1351,
+      "step": 15835
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9624525442708876e-06,
+      "loss": 17.4998,
+      "step": 15836
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9598626669450003e-06,
+      "loss": 18.5468,
+      "step": 15837
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.95727446553417e-06,
+      "loss": 17.7089,
+      "step": 15838
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9546879401286744e-06,
+      "loss": 0.7253,
+      "step": 15839
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.952103090818763e-06,
+      "loss": 16.732,
+      "step": 15840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.949519917694598e-06,
+      "loss": 18.2488,
+      "step": 15841
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9469384208463002e-06,
+      "loss": 16.1558,
+      "step": 15842
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9443586003639215e-06,
+      "loss": 17.106,
+      "step": 15843
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9417804563374664e-06,
+      "loss": 16.2625,
+      "step": 15844
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9392039888568747e-06,
+      "loss": 0.6766,
+      "step": 15845
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.93662919801203e-06,
+      "loss": 0.6786,
+      "step": 15846
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.934056083892755e-06,
+      "loss": 18.3158,
+      "step": 15847
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.931484646588805e-06,
+      "loss": 16.4944,
+      "step": 15848
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.928914886189903e-06,
+      "loss": 17.3456,
+      "step": 15849
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9263468027856834e-06,
+      "loss": 17.6203,
+      "step": 15850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.923780396465741e-06,
+      "loss": 18.295,
+      "step": 15851
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.921215667319609e-06,
+      "loss": 0.6604,
+      "step": 15852
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9186526154367557e-06,
+      "loss": 17.5859,
+      "step": 15853
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9160912409065977e-06,
+      "loss": 0.6432,
+      "step": 15854
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9135315438184864e-06,
+      "loss": 0.6836,
+      "step": 15855
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9109735242617278e-06,
+      "loss": 15.4724,
+      "step": 15856
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9084171823255504e-06,
+      "loss": 16.0986,
+      "step": 15857
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.905862518099133e-06,
+      "loss": 18.8076,
+      "step": 15858
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9033095316716044e-06,
+      "loss": 0.6628,
+      "step": 15859
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.900758223132032e-06,
+      "loss": 0.6059,
+      "step": 15860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.898208592569406e-06,
+      "loss": 18.9839,
+      "step": 15861
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.895660640072683e-06,
+      "loss": 0.7234,
+      "step": 15862
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.893114365730736e-06,
+      "loss": 17.3281,
+      "step": 15863
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.890569769632411e-06,
+      "loss": 0.6294,
+      "step": 15864
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8880268518664756e-06,
+      "loss": 0.62,
+      "step": 15865
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8854856125216314e-06,
+      "loss": 17.0121,
+      "step": 15866
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.882946051686535e-06,
+      "loss": 0.6839,
+      "step": 15867
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8804081694497822e-06,
+      "loss": 17.7911,
+      "step": 15868
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8778719658999078e-06,
+      "loss": 0.6819,
+      "step": 15869
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8753374411253965e-06,
+      "loss": 17.1016,
+      "step": 15870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8728045952146557e-06,
+      "loss": 17.0526,
+      "step": 15871
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.870273428256053e-06,
+      "loss": 0.6842,
+      "step": 15872
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8677439403378794e-06,
+      "loss": 0.7135,
+      "step": 15873
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8652161315483974e-06,
+      "loss": 17.7976,
+      "step": 15874
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8626900019757754e-06,
+      "loss": 0.6721,
+      "step": 15875
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8601655517081484e-06,
+      "loss": 17.0405,
+      "step": 15876
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.857642780833574e-06,
+      "loss": 18.1848,
+      "step": 15877
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8551216894400648e-06,
+      "loss": 0.6461,
+      "step": 15878
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8526022776155727e-06,
+      "loss": 17.9516,
+      "step": 15879
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8500845454479943e-06,
+      "loss": 0.6706,
+      "step": 15880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8475684930251535e-06,
+      "loss": 16.3508,
+      "step": 15881
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8450541204348304e-06,
+      "loss": 0.6715,
+      "step": 15882
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8425414277647324e-06,
+      "loss": 16.8717,
+      "step": 15883
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8400304151025283e-06,
+      "loss": 0.6911,
+      "step": 15884
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8375210825358091e-06,
+      "loss": 16.3052,
+      "step": 15885
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8350134301521216e-06,
+      "loss": 16.4957,
+      "step": 15886
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8325074580389345e-06,
+      "loss": 16.4472,
+      "step": 15887
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.830003166283678e-06,
+      "loss": 17.1006,
+      "step": 15888
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.827500554973721e-06,
+      "loss": 18.1725,
+      "step": 15889
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8249996241963607e-06,
+      "loss": 16.3176,
+      "step": 15890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8225003740388547e-06,
+      "loss": 0.6563,
+      "step": 15891
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8200028045883776e-06,
+      "loss": 0.6548,
+      "step": 15892
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8175069159320602e-06,
+      "loss": 15.7097,
+      "step": 15893
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.815012708156988e-06,
+      "loss": 0.644,
+      "step": 15894
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.812520181350158e-06,
+      "loss": 16.7534,
+      "step": 15895
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8100293355985343e-06,
+      "loss": 0.7146,
+      "step": 15896
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8075401709889972e-06,
+      "loss": 0.6317,
+      "step": 15897
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8050526876083995e-06,
+      "loss": 17.8626,
+      "step": 15898
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8025668855435107e-06,
+      "loss": 16.8497,
+      "step": 15899
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8000827648810503e-06,
+      "loss": 0.6004,
+      "step": 15900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7976003257076822e-06,
+      "loss": 0.6806,
+      "step": 15901
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.795119568109993e-06,
+      "loss": 16.3999,
+      "step": 15902
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.792640492174552e-06,
+      "loss": 16.842,
+      "step": 15903
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7901630979878237e-06,
+      "loss": 0.6589,
+      "step": 15904
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.787687385636244e-06,
+      "loss": 16.3842,
+      "step": 15905
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7852133552061722e-06,
+      "loss": 17.5769,
+      "step": 15906
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7827410067839167e-06,
+      "loss": 0.6691,
+      "step": 15907
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7802703404557365e-06,
+      "loss": 16.7468,
+      "step": 15908
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7778013563078177e-06,
+      "loss": 18.5962,
+      "step": 15909
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7753340544262863e-06,
+      "loss": 0.5709,
+      "step": 15910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.772868434897229e-06,
+      "loss": 0.6676,
+      "step": 15911
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7704044978066436e-06,
+      "loss": 0.6481,
+      "step": 15912
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7679422432404945e-06,
+      "loss": 16.4154,
+      "step": 15913
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7654816712846911e-06,
+      "loss": 16.6007,
+      "step": 15914
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7630227820250533e-06,
+      "loss": 17.2737,
+      "step": 15915
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7605655755473737e-06,
+      "loss": 15.4155,
+      "step": 15916
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7581100519373672e-06,
+      "loss": 19.9047,
+      "step": 15917
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7556562112806985e-06,
+      "loss": 17.2096,
+      "step": 15918
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7532040536629712e-06,
+      "loss": 0.6063,
+      "step": 15919
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7507535791697339e-06,
+      "loss": 0.6362,
+      "step": 15920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7483047878864678e-06,
+      "loss": 16.9767,
+      "step": 15921
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7458576798986105e-06,
+      "loss": 18.2324,
+      "step": 15922
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.74341225529151e-06,
+      "loss": 19.2594,
+      "step": 15923
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7409685141505038e-06,
+      "loss": 19.0668,
+      "step": 15924
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.738526456560824e-06,
+      "loss": 17.1457,
+      "step": 15925
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.736086082607674e-06,
+      "loss": 0.6513,
+      "step": 15926
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7336473923761864e-06,
+      "loss": 17.4161,
+      "step": 15927
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.731210385951426e-06,
+      "loss": 19.6772,
+      "step": 15928
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7287750634184253e-06,
+      "loss": 0.6746,
+      "step": 15929
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7263414248621324e-06,
+      "loss": 17.4941,
+      "step": 15930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.723909470367452e-06,
+      "loss": 0.6463,
+      "step": 15931
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7214792000192159e-06,
+      "loss": 0.639,
+      "step": 15932
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7190506139022121e-06,
+      "loss": 0.6792,
+      "step": 15933
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7166237121011675e-06,
+      "loss": 18.7965,
+      "step": 15934
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7141984947007418e-06,
+      "loss": 18.9186,
+      "step": 15935
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7117749617855394e-06,
+      "loss": 0.6436,
+      "step": 15936
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7093531134401098e-06,
+      "loss": 17.0833,
+      "step": 15937
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.706932949748935e-06,
+      "loss": 0.6838,
+      "step": 15938
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7045144707964534e-06,
+      "loss": 0.6996,
+      "step": 15939
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7020976766670304e-06,
+      "loss": 18.1113,
+      "step": 15940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6996825674449769e-06,
+      "loss": 18.8932,
+      "step": 15941
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6972691432145471e-06,
+      "loss": 0.6792,
+      "step": 15942
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6948574040599296e-06,
+      "loss": 0.711,
+      "step": 15943
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6924473500652682e-06,
+      "loss": 0.6558,
+      "step": 15944
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.69003898131464e-06,
+      "loss": 17.2709,
+      "step": 15945
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6876322978920556e-06,
+      "loss": 18.1797,
+      "step": 15946
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6852272998814756e-06,
+      "loss": 0.7188,
+      "step": 15947
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6828239873667995e-06,
+      "loss": 17.0124,
+      "step": 15948
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6804223604318825e-06,
+      "loss": 16.1407,
+      "step": 15949
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6780224191604854e-06,
+      "loss": 0.6812,
+      "step": 15950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6756241636363413e-06,
+      "loss": 16.4933,
+      "step": 15951
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6732275939431108e-06,
+      "loss": 17.4148,
+      "step": 15952
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6708327101644106e-06,
+      "loss": 0.6584,
+      "step": 15953
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6684395123837849e-06,
+      "loss": 20.1466,
+      "step": 15954
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6660480006847167e-06,
+      "loss": 15.5511,
+      "step": 15955
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.663658175150634e-06,
+      "loss": 17.7911,
+      "step": 15956
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6612700358649148e-06,
+      "loss": 16.1267,
+      "step": 15957
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.658883582910864e-06,
+      "loss": 0.6714,
+      "step": 15958
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6564988163717432e-06,
+      "loss": 16.5501,
+      "step": 15959
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6541157363307413e-06,
+      "loss": 0.6895,
+      "step": 15960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6517343428709975e-06,
+      "loss": 0.6074,
+      "step": 15961
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6493546360755784e-06,
+      "loss": 17.0019,
+      "step": 15962
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6469766160275124e-06,
+      "loss": 17.1004,
+      "step": 15963
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6446002828097606e-06,
+      "loss": 0.6304,
+      "step": 15964
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6422256365052125e-06,
+      "loss": 16.1631,
+      "step": 15965
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6398526771967127e-06,
+      "loss": 17.0678,
+      "step": 15966
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6374814049670451e-06,
+      "loss": 0.644,
+      "step": 15967
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6351118198989379e-06,
+      "loss": 17.8423,
+      "step": 15968
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6327439220750473e-06,
+      "loss": 16.6487,
+      "step": 15969
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.630377711577985e-06,
+      "loss": 16.1033,
+      "step": 15970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.628013188490296e-06,
+      "loss": 15.7111,
+      "step": 15971
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6256503528944645e-06,
+      "loss": 19.7481,
+      "step": 15972
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6232892048729297e-06,
+      "loss": 0.6749,
+      "step": 15973
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.620929744508054e-06,
+      "loss": 0.6304,
+      "step": 15974
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6185719718821547e-06,
+      "loss": 0.6716,
+      "step": 15975
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6162158870774768e-06,
+      "loss": 0.703,
+      "step": 15976
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.613861490176216e-06,
+      "loss": 15.5225,
+      "step": 15977
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6115087812605123e-06,
+      "loss": 16.0966,
+      "step": 15978
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6091577604124386e-06,
+      "loss": 17.6781,
+      "step": 15979
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6068084277140073e-06,
+      "loss": 0.7473,
+      "step": 15980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6044607832471858e-06,
+      "loss": 16.4549,
+      "step": 15981
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6021148270938646e-06,
+      "loss": 16.4251,
+      "step": 15982
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5997705593358892e-06,
+      "loss": 16.8073,
+      "step": 15983
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.597427980055044e-06,
+      "loss": 17.1369,
+      "step": 15984
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5950870893330473e-06,
+      "loss": 17.0897,
+      "step": 15985
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5927478872515611e-06,
+      "loss": 0.6685,
+      "step": 15986
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.590410373892187e-06,
+      "loss": 19.6922,
+      "step": 15987
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5880745493364823e-06,
+      "loss": 0.5942,
+      "step": 15988
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.585740413665926e-06,
+      "loss": 17.2027,
+      "step": 15989
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5834079669619472e-06,
+      "loss": 16.2885,
+      "step": 15990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5810772093059201e-06,
+      "loss": 0.6512,
+      "step": 15991
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5787481407791405e-06,
+      "loss": 16.2059,
+      "step": 15992
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5764207614628767e-06,
+      "loss": 15.152,
+      "step": 15993
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5740950714383085e-06,
+      "loss": 16.3569,
+      "step": 15994
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571771070786582e-06,
+      "loss": 0.6059,
+      "step": 15995
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5694487595887598e-06,
+      "loss": 0.6227,
+      "step": 15996
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5671281379258552e-06,
+      "loss": 0.6689,
+      "step": 15997
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.564809205878842e-06,
+      "loss": 0.6451,
+      "step": 15998
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5624919635286062e-06,
+      "loss": 15.8439,
+      "step": 15999
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5601764109559824e-06,
+      "loss": 17.1081,
+      "step": 16000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5578625482417619e-06,
+      "loss": 17.9633,
+      "step": 16001
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5555503754666468e-06,
+      "loss": 0.6588,
+      "step": 16002
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5532398927113223e-06,
+      "loss": 0.6756,
+      "step": 16003
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5509311000563798e-06,
+      "loss": 18.7739,
+      "step": 16004
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5486239975823659e-06,
+      "loss": 0.6577,
+      "step": 16005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5463185853697493e-06,
+      "loss": 17.7512,
+      "step": 16006
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5440148634989826e-06,
+      "loss": 17.7896,
+      "step": 16007
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5417128320504126e-06,
+      "loss": 0.6447,
+      "step": 16008
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5394124911043584e-06,
+      "loss": 0.6613,
+      "step": 16009
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5371138407410669e-06,
+      "loss": 18.2287,
+      "step": 16010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5348168810407182e-06,
+      "loss": 17.1133,
+      "step": 16011
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5325216120834596e-06,
+      "loss": 15.5438,
+      "step": 16012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5302280339493547e-06,
+      "loss": 0.6243,
+      "step": 16013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5279361467184172e-06,
+      "loss": 0.666,
+      "step": 16014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5256459504706054e-06,
+      "loss": 17.998,
+      "step": 16015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5233574452858001e-06,
+      "loss": 0.6447,
+      "step": 16016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5210706312438538e-06,
+      "loss": 18.2049,
+      "step": 16017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5187855084245417e-06,
+      "loss": 0.6639,
+      "step": 16018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5165020769075722e-06,
+      "loss": 17.5288,
+      "step": 16019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5142203367726149e-06,
+      "loss": 18.1409,
+      "step": 16020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.511940288099256e-06,
+      "loss": 0.7261,
+      "step": 16021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5096619309670547e-06,
+      "loss": 17.7248,
+      "step": 16022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5073852654554854e-06,
+      "loss": 0.6629,
+      "step": 16023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5051102916439686e-06,
+      "loss": 16.5645,
+      "step": 16024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5028370096118682e-06,
+      "loss": 16.9104,
+      "step": 16025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5005654194384877e-06,
+      "loss": 16.9892,
+      "step": 16026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4982955212030802e-06,
+      "loss": 15.9739,
+      "step": 16027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4960273149848324e-06,
+      "loss": 15.8627,
+      "step": 16028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.49376080086287e-06,
+      "loss": 0.6971,
+      "step": 16029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4914959789162574e-06,
+      "loss": 0.6229,
+      "step": 16030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4892328492240037e-06,
+      "loss": 0.6278,
+      "step": 16031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.486971411865068e-06,
+      "loss": 16.2147,
+      "step": 16032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4847116669183424e-06,
+      "loss": 15.6832,
+      "step": 16033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4824536144626533e-06,
+      "loss": 17.8874,
+      "step": 16034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4801972545767818e-06,
+      "loss": 17.2204,
+      "step": 16035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4779425873394259e-06,
+      "loss": 19.2226,
+      "step": 16036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.475689612829262e-06,
+      "loss": 0.6986,
+      "step": 16037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4734383311248823e-06,
+      "loss": 18.1031,
+      "step": 16038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.471188742304813e-06,
+      "loss": 18.2983,
+      "step": 16039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.468940846447542e-06,
+      "loss": 0.6429,
+      "step": 16040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4666946436314832e-06,
+      "loss": 16.6586,
+      "step": 16041
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.464450133935008e-06,
+      "loss": 0.6794,
+      "step": 16042
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4622073174364038e-06,
+      "loss": 18.8643,
+      "step": 16043
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4599661942139243e-06,
+      "loss": 16.0927,
+      "step": 16044
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4577267643457459e-06,
+      "loss": 18.6677,
+      "step": 16045
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4554890279099897e-06,
+      "loss": 17.894,
+      "step": 16046
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.453252984984732e-06,
+      "loss": 0.6572,
+      "step": 16047
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4510186356479716e-06,
+      "loss": 16.4063,
+      "step": 16048
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4487859799776514e-06,
+      "loss": 17.4725,
+      "step": 16049
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4465550180516706e-06,
+      "loss": 0.6359,
+      "step": 16050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4443257499478447e-06,
+      "loss": 0.6515,
+      "step": 16051
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.44209817574395e-06,
+      "loss": 17.2996,
+      "step": 16052
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4398722955177025e-06,
+      "loss": 17.2385,
+      "step": 16053
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4376481093467454e-06,
+      "loss": 17.9559,
+      "step": 16054
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4354256173086778e-06,
+      "loss": 17.3917,
+      "step": 16055
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.433204819481021e-06,
+      "loss": 17.2625,
+      "step": 16056
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4309857159412631e-06,
+      "loss": 19.0975,
+      "step": 16057
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4287683067668145e-06,
+      "loss": 0.6223,
+      "step": 16058
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4265525920350297e-06,
+      "loss": 0.6399,
+      "step": 16059
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4243385718231971e-06,
+      "loss": 16.6006,
+      "step": 16060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4221262462085715e-06,
+      "loss": 17.1597,
+      "step": 16061
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.419915615268319e-06,
+      "loss": 18.9708,
+      "step": 16062
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4177066790795669e-06,
+      "loss": 15.2131,
+      "step": 16063
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4154994377193641e-06,
+      "loss": 17.9588,
+      "step": 16064
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.413293891264722e-06,
+      "loss": 16.7417,
+      "step": 16065
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.411090039792584e-06,
+      "loss": 17.7965,
+      "step": 16066
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.408887883379828e-06,
+      "loss": 0.6449,
+      "step": 16067
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4066874221032755e-06,
+      "loss": 18.2175,
+      "step": 16068
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4044886560396985e-06,
+      "loss": 0.6447,
+      "step": 16069
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4022915852657915e-06,
+      "loss": 17.0033,
+      "step": 16070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.400096209858215e-06,
+      "loss": 0.6601,
+      "step": 16071
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3979025298935467e-06,
+      "loss": 17.4978,
+      "step": 16072
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.39571054544832e-06,
+      "loss": 0.6495,
+      "step": 16073
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3935202565989957e-06,
+      "loss": 0.6833,
+      "step": 16074
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3913316634219909e-06,
+      "loss": 0.6275,
+      "step": 16075
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3891447659936552e-06,
+      "loss": 0.6679,
+      "step": 16076
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3869595643902832e-06,
+      "loss": 0.6128,
+      "step": 16077
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3847760586880976e-06,
+      "loss": 19.0346,
+      "step": 16078
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3825942489632814e-06,
+      "loss": 16.9835,
+      "step": 16079
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3804141352919408e-06,
+      "loss": 17.5602,
+      "step": 16080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378235717750137e-06,
+      "loss": 0.651,
+      "step": 16081
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.376058996413865e-06,
+      "loss": 17.5978,
+      "step": 16082
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.373883971359058e-06,
+      "loss": 17.9061,
+      "step": 16083
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3717106426615943e-06,
+      "loss": 0.6906,
+      "step": 16084
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.369539010397286e-06,
+      "loss": 16.3135,
+      "step": 16085
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3673690746419054e-06,
+      "loss": 0.6432,
+      "step": 16086
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3652008354711477e-06,
+      "loss": 18.8956,
+      "step": 16087
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3630342929606465e-06,
+      "loss": 16.312,
+      "step": 16088
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3608694471859917e-06,
+      "loss": 18.8686,
+      "step": 16089
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3587062982226951e-06,
+      "loss": 16.4108,
+      "step": 16090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3565448461462293e-06,
+      "loss": 0.684,
+      "step": 16091
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.354385091032001e-06,
+      "loss": 16.3509,
+      "step": 16092
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3522270329553443e-06,
+      "loss": 17.8229,
+      "step": 16093
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.350070671991549e-06,
+      "loss": 18.6841,
+      "step": 16094
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3479160082158383e-06,
+      "loss": 18.5136,
+      "step": 16095
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.345763041703385e-06,
+      "loss": 17.701,
+      "step": 16096
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3436117725293019e-06,
+      "loss": 0.6743,
+      "step": 16097
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3414622007686228e-06,
+      "loss": 15.0679,
+      "step": 16098
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3393143264963436e-06,
+      "loss": 0.6269,
+      "step": 16099
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.337168149787399e-06,
+      "loss": 0.6945,
+      "step": 16100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3350236707166509e-06,
+      "loss": 17.2668,
+      "step": 16101
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3328808893589228e-06,
+      "loss": 0.6533,
+      "step": 16102
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.330739805788961e-06,
+      "loss": 18.0884,
+      "step": 16103
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.328600420081455e-06,
+      "loss": 16.5644,
+      "step": 16104
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.32646273231104e-06,
+      "loss": 0.687,
+      "step": 16105
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3243267425522953e-06,
+      "loss": 17.9249,
+      "step": 16106
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3221924508797335e-06,
+      "loss": 17.0383,
+      "step": 16107
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.320059857367817e-06,
+      "loss": 17.5637,
+      "step": 16108
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3179289620909308e-06,
+      "loss": 0.6786,
+      "step": 16109
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.315799765123421e-06,
+      "loss": 18.1453,
+      "step": 16110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3136722665395617e-06,
+      "loss": 19.5377,
+      "step": 16111
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3115464664135823e-06,
+      "loss": 18.6215,
+      "step": 16112
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3094223648196403e-06,
+      "loss": 0.6507,
+      "step": 16113
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3072999618318204e-06,
+      "loss": 17.7881,
+      "step": 16114
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.305179257524175e-06,
+      "loss": 16.8694,
+      "step": 16115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3030602519706947e-06,
+      "loss": 18.332,
+      "step": 16116
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3009429452452926e-06,
+      "loss": 0.6998,
+      "step": 16117
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2988273374218374e-06,
+      "loss": 19.7163,
+      "step": 16118
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2967134285741255e-06,
+      "loss": 18.2149,
+      "step": 16119
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2946012187759038e-06,
+      "loss": 17.1983,
+      "step": 16120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2924907081008686e-06,
+      "loss": 0.7162,
+      "step": 16121
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2903818966226443e-06,
+      "loss": 17.2191,
+      "step": 16122
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.288274784414789e-06,
+      "loss": 0.7437,
+      "step": 16123
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.286169371550816e-06,
+      "loss": 16.3318,
+      "step": 16124
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2840656581041722e-06,
+      "loss": 17.0245,
+      "step": 16125
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.281963644148254e-06,
+      "loss": 17.1998,
+      "step": 16126
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2798633297563867e-06,
+      "loss": 0.6894,
+      "step": 16127
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.277764715001839e-06,
+      "loss": 0.6678,
+      "step": 16128
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2756677999578247e-06,
+      "loss": 0.7124,
+      "step": 16129
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.273572584697491e-06,
+      "loss": 0.6304,
+      "step": 16130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.271479069293946e-06,
+      "loss": 18.2592,
+      "step": 16131
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2693872538202145e-06,
+      "loss": 18.6615,
+      "step": 16132
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.267297138349266e-06,
+      "loss": 16.5625,
+      "step": 16133
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2652087229540198e-06,
+      "loss": 17.3908,
+      "step": 16134
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2631220077073292e-06,
+      "loss": 17.0127,
+      "step": 16135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.261036992682002e-06,
+      "loss": 16.5849,
+      "step": 16136
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2589536779507637e-06,
+      "loss": 16.9274,
+      "step": 16137
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2568720635862952e-06,
+      "loss": 0.6829,
+      "step": 16138
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2547921496612158e-06,
+      "loss": 17.4545,
+      "step": 16139
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2527139362480843e-06,
+      "loss": 18.4888,
+      "step": 16140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2506374234193985e-06,
+      "loss": 15.49,
+      "step": 16141
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2485626112476057e-06,
+      "loss": 0.6572,
+      "step": 16142
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2464894998050868e-06,
+      "loss": 16.9925,
+      "step": 16143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2444180891641567e-06,
+      "loss": 16.682,
+      "step": 16144
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2423483793970736e-06,
+      "loss": 15.8628,
+      "step": 16145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.240280370576058e-06,
+      "loss": 0.6417,
+      "step": 16146
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2382140627732407e-06,
+      "loss": 0.7022,
+      "step": 16147
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2361494560607146e-06,
+      "loss": 0.6425,
+      "step": 16148
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2340865505104992e-06,
+      "loss": 18.05,
+      "step": 16149
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2320253461945597e-06,
+      "loss": 0.6567,
+      "step": 16150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.229965843184805e-06,
+      "loss": 0.622,
+      "step": 16151
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.227908041553083e-06,
+      "loss": 17.1977,
+      "step": 16152
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2258519413711866e-06,
+      "loss": 0.7468,
+      "step": 16153
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2237975427108305e-06,
+      "loss": 17.4956,
+      "step": 16154
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2217448456436909e-06,
+      "loss": 17.1646,
+      "step": 16155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.219693850241388e-06,
+      "loss": 0.6139,
+      "step": 16156
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.217644556575459e-06,
+      "loss": 0.6347,
+      "step": 16157
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.215596964717397e-06,
+      "loss": 18.5043,
+      "step": 16158
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.213551074738639e-06,
+      "loss": 17.8777,
+      "step": 16159
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2115068867105505e-06,
+      "loss": 18.526,
+      "step": 16160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.209464400704452e-06,
+      "loss": 16.8751,
+      "step": 16161
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.207423616791592e-06,
+      "loss": 16.0073,
+      "step": 16162
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2053845350431636e-06,
+      "loss": 15.2092,
+      "step": 16163
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2033471555303045e-06,
+      "loss": 0.6705,
+      "step": 16164
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2013114783240909e-06,
+      "loss": 16.5043,
+      "step": 16165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1992775034955383e-06,
+      "loss": 17.8696,
+      "step": 16166
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1972452311156069e-06,
+      "loss": 0.6096,
+      "step": 16167
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.195214661255184e-06,
+      "loss": 0.7191,
+      "step": 16168
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1931857939851132e-06,
+      "loss": 17.5887,
+      "step": 16169
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1911586293761767e-06,
+      "loss": 18.3968,
+      "step": 16170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1891331674990846e-06,
+      "loss": 17.3569,
+      "step": 16171
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1871094084245082e-06,
+      "loss": 0.6303,
+      "step": 16172
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.185087352223041e-06,
+      "loss": 0.6812,
+      "step": 16173
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1830669989652155e-06,
+      "loss": 0.6373,
+      "step": 16174
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1810483487215307e-06,
+      "loss": 16.3088,
+      "step": 16175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1790314015624027e-06,
+      "loss": 18.0767,
+      "step": 16176
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1770161575581862e-06,
+      "loss": 0.6408,
+      "step": 16177
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1750026167791917e-06,
+      "loss": 16.8424,
+      "step": 16178
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1729907792956574e-06,
+      "loss": 0.6291,
+      "step": 16179
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1709806451777773e-06,
+      "loss": 0.674,
+      "step": 16180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1689722144956671e-06,
+      "loss": 14.9964,
+      "step": 16181
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.166965487319399e-06,
+      "loss": 16.4195,
+      "step": 16182
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1649604637189782e-06,
+      "loss": 16.989,
+      "step": 16183
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1629571437643371e-06,
+      "loss": 0.7291,
+      "step": 16184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.160955527525387e-06,
+      "loss": 17.6853,
+      "step": 16185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1589556150719439e-06,
+      "loss": 16.6466,
+      "step": 16186
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1569574064737688e-06,
+      "loss": 0.7202,
+      "step": 16187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1549609018005837e-06,
+      "loss": 0.6545,
+      "step": 16188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1529661011220271e-06,
+      "loss": 18.292,
+      "step": 16189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1509730045076995e-06,
+      "loss": 16.5835,
+      "step": 16190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1489816120271279e-06,
+      "loss": 0.6525,
+      "step": 16191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1469919237497794e-06,
+      "loss": 0.6681,
+      "step": 16192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1450039397450708e-06,
+      "loss": 18.6004,
+      "step": 16193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1430176600823462e-06,
+      "loss": 17.3881,
+      "step": 16194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1410330848309058e-06,
+      "loss": 0.6297,
+      "step": 16195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1390502140599835e-06,
+      "loss": 0.636,
+      "step": 16196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1370690478387514e-06,
+      "loss": 0.6995,
+      "step": 16197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.135089586236321e-06,
+      "loss": 0.6902,
+      "step": 16198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.133111829321748e-06,
+      "loss": 16.9128,
+      "step": 16199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.131135777164033e-06,
+      "loss": 17.6142,
+      "step": 16200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1291614298321096e-06,
+      "loss": 16.242,
+      "step": 16201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1271887873948561e-06,
+      "loss": 0.7046,
+      "step": 16202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1252178499210785e-06,
+      "loss": 17.1747,
+      "step": 16203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1232486174795443e-06,
+      "loss": 17.7901,
+      "step": 16204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1212810901389536e-06,
+      "loss": 0.6631,
+      "step": 16205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1193152679679408e-06,
+      "loss": 16.8003,
+      "step": 16206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1173511510350899e-06,
+      "loss": 17.1646,
+      "step": 16207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1153887394089124e-06,
+      "loss": 18.6922,
+      "step": 16208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.113428033157865e-06,
+      "loss": 0.6313,
+      "step": 16209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1114690323503652e-06,
+      "loss": 0.6518,
+      "step": 16210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1095117370547415e-06,
+      "loss": 17.0476,
+      "step": 16211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1075561473392783e-06,
+      "loss": 0.6349,
+      "step": 16212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.105602263272204e-06,
+      "loss": 17.8711,
+      "step": 16213
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1036500849216647e-06,
+      "loss": 17.5504,
+      "step": 16214
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1016996123557833e-06,
+      "loss": 15.3799,
+      "step": 16215
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0997508456425888e-06,
+      "loss": 16.4661,
+      "step": 16216
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0978037848500766e-06,
+      "loss": 0.6165,
+      "step": 16217
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.095858430046165e-06,
+      "loss": 0.6827,
+      "step": 16218
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0939147812987106e-06,
+      "loss": 18.6557,
+      "step": 16219
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0919728386755368e-06,
+      "loss": 18.2957,
+      "step": 16220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090032602244384e-06,
+      "loss": 17.0566,
+      "step": 16221
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.088094072072937e-06,
+      "loss": 16.8942,
+      "step": 16222
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0861572482288086e-06,
+      "loss": 15.2052,
+      "step": 16223
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0842221307795885e-06,
+      "loss": 0.6432,
+      "step": 16224
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0822887197927788e-06,
+      "loss": 17.4613,
+      "step": 16225
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.08035701533582e-06,
+      "loss": 0.6625,
+      "step": 16226
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0784270174761079e-06,
+      "loss": 16.8694,
+      "step": 16227
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.076498726280961e-06,
+      "loss": 17.9111,
+      "step": 16228
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0745721418176647e-06,
+      "loss": 0.6979,
+      "step": 16229
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0726472641534257e-06,
+      "loss": 19.6388,
+      "step": 16230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.070724093355391e-06,
+      "loss": 0.6743,
+      "step": 16231
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0688026294906506e-06,
+      "loss": 17.1133,
+      "step": 16232
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0668828726262348e-06,
+      "loss": 18.8487,
+      "step": 16233
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0649648228291176e-06,
+      "loss": 0.6998,
+      "step": 16234
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0630484801662176e-06,
+      "loss": 16.9908,
+      "step": 16235
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0611338447043816e-06,
+      "loss": 0.6043,
+      "step": 16236
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0592209165104062e-06,
+      "loss": 16.5921,
+      "step": 16237
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.057309695651021e-06,
+      "loss": 0.6768,
+      "step": 16238
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.055400182192906e-06,
+      "loss": 0.672,
+      "step": 16239
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0534923762026694e-06,
+      "loss": 15.6443,
+      "step": 16240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0515862777468689e-06,
+      "loss": 0.6715,
+      "step": 16241
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0496818868920066e-06,
+      "loss": 0.6839,
+      "step": 16242
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0477792037045076e-06,
+      "loss": 0.7172,
+      "step": 16243
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0458782282507574e-06,
+      "loss": 0.6687,
+      "step": 16244
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.043978960597064e-06,
+      "loss": 18.3043,
+      "step": 16245
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0420814008096968e-06,
+      "loss": 18.3102,
+      "step": 16246
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0401855489548417e-06,
+      "loss": 16.6203,
+      "step": 16247
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0382914050986403e-06,
+      "loss": 0.7095,
+      "step": 16248
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0363989693071785e-06,
+      "loss": 0.6573,
+      "step": 16249
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0345082416464703e-06,
+      "loss": 16.1535,
+      "step": 16250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0326192221824737e-06,
+      "loss": 0.7112,
+      "step": 16251
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0307319109810864e-06,
+      "loss": 16.3922,
+      "step": 16252
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.02884630810815e-06,
+      "loss": 0.6702,
+      "step": 16253
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0269624136294454e-06,
+      "loss": 17.5898,
+      "step": 16254
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.025080227610703e-06,
+      "loss": 17.9989,
+      "step": 16255
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0231997501175704e-06,
+      "loss": 16.6928,
+      "step": 16256
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0213209812156565e-06,
+      "loss": 18.1744,
+      "step": 16257
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0194439209704975e-06,
+      "loss": 0.6302,
+      "step": 16258
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0175685694475856e-06,
+      "loss": 0.6115,
+      "step": 16259
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0156949267123406e-06,
+      "loss": 18.9211,
+      "step": 16260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0138229928301212e-06,
+      "loss": 0.6528,
+      "step": 16261
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.011952767866231e-06,
+      "loss": 17.8956,
+      "step": 16262
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0100842518859177e-06,
+      "loss": 16.3984,
+      "step": 16263
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.008217444954368e-06,
+      "loss": 16.3712,
+      "step": 16264
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0063523471367076e-06,
+      "loss": 16.7421,
+      "step": 16265
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0044889584979955e-06,
+      "loss": 17.053,
+      "step": 16266
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.002627279103241e-06,
+      "loss": 15.8749,
+      "step": 16267
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0007673090173808e-06,
+      "loss": 17.2826,
+      "step": 16268
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.989090483053188e-07,
+      "loss": 0.6904,
+      "step": 16269
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.970524970318695e-07,
+      "loss": 0.6604,
+      "step": 16270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.951976552618093e-07,
+      "loss": 0.6406,
+      "step": 16271
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.93344523059836e-07,
+      "loss": 0.6425,
+      "step": 16272
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.914931004905925e-07,
+      "loss": 0.6285,
+      "step": 16273
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.896433876186883e-07,
+      "loss": 0.6742,
+      "step": 16274
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.87795384508633e-07,
+      "loss": 0.6286,
+      "step": 16275
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.859490912249025e-07,
+      "loss": 0.6347,
+      "step": 16276
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.841045078319123e-07,
+      "loss": 19.017,
+      "step": 16277
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.822616343939993e-07,
+      "loss": 0.6774,
+      "step": 16278
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.804204709754572e-07,
+      "loss": 0.6416,
+      "step": 16279
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.78581017640523e-07,
+      "loss": 18.2323,
+      "step": 16280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.767432744533623e-07,
+      "loss": 15.7328,
+      "step": 16281
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.749072414780848e-07,
+      "loss": 0.6825,
+      "step": 16282
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.730729187787446e-07,
+      "loss": 16.6486,
+      "step": 16283
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.712403064193298e-07,
+      "loss": 18.7147,
+      "step": 16284
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69409404463778e-07,
+      "loss": 17.2991,
+      "step": 16285
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.675802129759603e-07,
+      "loss": 0.731,
+      "step": 16286
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.657527320196812e-07,
+      "loss": 16.947,
+      "step": 16287
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.639269616587065e-07,
+      "loss": 14.9923,
+      "step": 16288
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.62102901956724e-07,
+      "loss": 0.651,
+      "step": 16289
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.602805529773662e-07,
+      "loss": 18.6757,
+      "step": 16290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.584599147842044e-07,
+      "loss": 18.2,
+      "step": 16291
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.566409874407544e-07,
+      "loss": 18.5712,
+      "step": 16292
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.548237710104768e-07,
+      "loss": 17.8934,
+      "step": 16293
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.530082655567596e-07,
+      "loss": 19.2961,
+      "step": 16294
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.51194471142941e-07,
+      "loss": 0.6745,
+      "step": 16295
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.493823878322983e-07,
+      "loss": 16.5787,
+      "step": 16296
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.475720156880419e-07,
+      "loss": 18.3702,
+      "step": 16297
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.457633547733325e-07,
+      "loss": 17.936,
+      "step": 16298
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.439564051512639e-07,
+      "loss": 0.714,
+      "step": 16299
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.421511668848748e-07,
+      "loss": 17.8672,
+      "step": 16300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.403476400371424e-07,
+      "loss": 18.3748,
+      "step": 16301
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.385458246709889e-07,
+      "loss": 0.6831,
+      "step": 16302
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.367457208492525e-07,
+      "loss": 16.552,
+      "step": 16303
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.349473286347555e-07,
+      "loss": 0.6349,
+      "step": 16304
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.331506480902252e-07,
+      "loss": 17.4074,
+      "step": 16305
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.313556792783395e-07,
+      "loss": 18.5644,
+      "step": 16306
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.29562422261715e-07,
+      "loss": 0.7402,
+      "step": 16307
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.277708771029126e-07,
+      "loss": 0.6236,
+      "step": 16308
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.259810438644378e-07,
+      "loss": 18.0982,
+      "step": 16309
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.241929226087243e-07,
+      "loss": 16.992,
+      "step": 16310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.224065133981496e-07,
+      "loss": 0.5976,
+      "step": 16311
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.20621816295042e-07,
+      "loss": 17.7182,
+      "step": 16312
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.188388313616513e-07,
+      "loss": 0.6985,
+      "step": 16313
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.170575586601893e-07,
+      "loss": 18.1963,
+      "step": 16314
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.152779982527948e-07,
+      "loss": 18.3295,
+      "step": 16315
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.135001502015405e-07,
+      "loss": 18.4644,
+      "step": 16316
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.117240145684547e-07,
+      "loss": 18.7764,
+      "step": 16317
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.099495914154987e-07,
+      "loss": 17.2235,
+      "step": 16318
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.081768808045732e-07,
+      "loss": 17.1293,
+      "step": 16319
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.06405882797523e-07,
+      "loss": 16.3675,
+      "step": 16320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.046365974561266e-07,
+      "loss": 0.6611,
+      "step": 16321
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.028690248421124e-07,
+      "loss": 17.6002,
+      "step": 16322
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.011031650171364e-07,
+      "loss": 17.0724,
+      "step": 16323
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.993390180428052e-07,
+      "loss": 16.8929,
+      "step": 16324
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.975765839806694e-07,
+      "loss": 16.6509,
+      "step": 16325
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.958158628922019e-07,
+      "loss": 16.6814,
+      "step": 16326
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.940568548388317e-07,
+      "loss": 17.4863,
+      "step": 16327
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.92299559881915e-07,
+      "loss": 18.2636,
+      "step": 16328
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.905439780827751e-07,
+      "loss": 15.9218,
+      "step": 16329
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.887901095026408e-07,
+      "loss": 16.8062,
+      "step": 16330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.870379542027018e-07,
+      "loss": 20.6408,
+      "step": 16331
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.852875122440818e-07,
+      "loss": 20.264,
+      "step": 16332
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.835387836878484e-07,
+      "loss": 0.6595,
+      "step": 16333
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.817917685950083e-07,
+      "loss": 18.095,
+      "step": 16334
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.800464670265018e-07,
+      "loss": 0.6884,
+      "step": 16335
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.783028790432191e-07,
+      "loss": 0.5582,
+      "step": 16336
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.765610047059891e-07,
+      "loss": 16.5523,
+      "step": 16337
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.748208440755745e-07,
+      "loss": 16.1406,
+      "step": 16338
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.730823972126823e-07,
+      "loss": 17.3113,
+      "step": 16339
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.713456641779583e-07,
+      "loss": 0.6661,
+      "step": 16340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.696106450319929e-07,
+      "loss": 18.6697,
+      "step": 16341
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.678773398353102e-07,
+      "loss": 17.6215,
+      "step": 16342
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.661457486483837e-07,
+      "loss": 17.8303,
+      "step": 16343
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.644158715316153e-07,
+      "loss": 18.5517,
+      "step": 16344
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.626877085453567e-07,
+      "loss": 0.6814,
+      "step": 16345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.609612597498928e-07,
+      "loss": 0.6261,
+      "step": 16346
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.592365252054535e-07,
+      "loss": 0.6723,
+      "step": 16347
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.575135049722072e-07,
+      "loss": 17.3803,
+      "step": 16348
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.557921991102613e-07,
+      "loss": 0.6731,
+      "step": 16349
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.540726076796735e-07,
+      "loss": 0.6624,
+      "step": 16350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.523547307404178e-07,
+      "loss": 0.6704,
+      "step": 16351
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.506385683524354e-07,
+      "loss": 0.6611,
+      "step": 16352
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.489241205755949e-07,
+      "loss": 17.1465,
+      "step": 16353
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.472113874696985e-07,
+      "loss": 0.6018,
+      "step": 16354
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.455003690945041e-07,
+      "loss": 17.184,
+      "step": 16355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.43791065509697e-07,
+      "loss": 0.6286,
+      "step": 16356
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.420834767749075e-07,
+      "loss": 0.7121,
+      "step": 16357
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.403776029497157e-07,
+      "loss": 18.3845,
+      "step": 16358
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.386734440936183e-07,
+      "loss": 18.4469,
+      "step": 16359
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.369710002660735e-07,
+      "loss": 16.6125,
+      "step": 16360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.352702715264726e-07,
+      "loss": 17.7382,
+      "step": 16361
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.335712579341404e-07,
+      "loss": 18.1775,
+      "step": 16362
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.318739595483571e-07,
+      "loss": 0.6284,
+      "step": 16363
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.30178376428331e-07,
+      "loss": 17.122,
+      "step": 16364
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.284845086332093e-07,
+      "loss": 0.7216,
+      "step": 16365
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.267923562220892e-07,
+      "loss": 0.6619,
+      "step": 16366
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.251019192540011e-07,
+      "loss": 15.7244,
+      "step": 16367
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.234131977879144e-07,
+      "loss": 0.5909,
+      "step": 16368
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.21726191882749e-07,
+      "loss": 0.6491,
+      "step": 16369
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.20040901597352e-07,
+      "loss": 17.2162,
+      "step": 16370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.183573269905098e-07,
+      "loss": 20.306,
+      "step": 16371
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.166754681209642e-07,
+      "loss": 15.3361,
+      "step": 16372
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.149953250473852e-07,
+      "loss": 0.6691,
+      "step": 16373
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.133168978283867e-07,
+      "loss": 16.8988,
+      "step": 16374
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.116401865225221e-07,
+      "loss": 0.7513,
+      "step": 16375
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.099651911882833e-07,
+      "loss": 20.105,
+      "step": 16376
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.082919118841015e-07,
+      "loss": 18.844,
+      "step": 16377
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.066203486683521e-07,
+      "loss": 17.0445,
+      "step": 16378
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.049505015993553e-07,
+      "loss": 16.82,
+      "step": 16379
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.032823707353587e-07,
+      "loss": 17.3656,
+      "step": 16380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.016159561345548e-07,
+      "loss": 16.5746,
+      "step": 16381
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.99951257855075e-07,
+      "loss": 16.537,
+      "step": 16382
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.982882759550003e-07,
+      "loss": 0.6788,
+      "step": 16383
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.966270104923457e-07,
+      "loss": 0.7219,
+      "step": 16384
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.949674615250646e-07,
+      "loss": 0.6781,
+      "step": 16385
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.933096291110442e-07,
+      "loss": 0.6765,
+      "step": 16386
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.916535133081271e-07,
+      "loss": 17.6671,
+      "step": 16387
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.899991141740892e-07,
+      "loss": 17.6697,
+      "step": 16388
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.8834643176664e-07,
+      "loss": 17.4082,
+      "step": 16389
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.866954661434334e-07,
+      "loss": 16.1259,
+      "step": 16390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.850462173620677e-07,
+      "loss": 0.6635,
+      "step": 16391
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.833986854800802e-07,
+      "loss": 16.5203,
+      "step": 16392
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.817528705549471e-07,
+      "loss": 17.9309,
+      "step": 16393
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.801087726440781e-07,
+      "loss": 0.7021,
+      "step": 16394
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.784663918048274e-07,
+      "loss": 16.7907,
+      "step": 16395
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.768257280944991e-07,
+      "loss": 16.0669,
+      "step": 16396
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.751867815703196e-07,
+      "loss": 0.6011,
+      "step": 16397
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.735495522894764e-07,
+      "loss": 18.1656,
+      "step": 16398
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.719140403090796e-07,
+      "loss": 0.6442,
+      "step": 16399
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.70280245686178e-07,
+      "loss": 16.4055,
+      "step": 16400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.686481684777758e-07,
+      "loss": 0.6719,
+      "step": 16401
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.670178087408108e-07,
+      "loss": 0.6316,
+      "step": 16402
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.653891665321544e-07,
+      "loss": 16.049,
+      "step": 16403
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.637622419086221e-07,
+      "loss": 16.1458,
+      "step": 16404
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.621370349269741e-07,
+      "loss": 17.9272,
+      "step": 16405
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.605135456439038e-07,
+      "loss": 16.5002,
+      "step": 16406
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.58891774116055e-07,
+      "loss": 0.6534,
+      "step": 16407
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.572717203999935e-07,
+      "loss": 16.5422,
+      "step": 16408
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.556533845522462e-07,
+      "loss": 16.1153,
+      "step": 16409
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.540367666292624e-07,
+      "loss": 0.6338,
+      "step": 16410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.524218666874417e-07,
+      "loss": 0.569,
+      "step": 16411
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.508086847831219e-07,
+      "loss": 0.6937,
+      "step": 16412
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.491972209725806e-07,
+      "loss": 0.6222,
+      "step": 16413
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.475874753120337e-07,
+      "loss": 0.611,
+      "step": 16414
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.459794478576365e-07,
+      "loss": 0.5948,
+      "step": 16415
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.443731386654884e-07,
+      "loss": 0.6627,
+      "step": 16416
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.427685477916279e-07,
+      "loss": 17.7266,
+      "step": 16417
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.411656752920271e-07,
+      "loss": 17.6054,
+      "step": 16418
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.395645212226132e-07,
+      "loss": 18.3378,
+      "step": 16419
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.379650856392306e-07,
+      "loss": 17.6882,
+      "step": 16420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.363673685976846e-07,
+      "loss": 17.3026,
+      "step": 16421
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.34771370153714e-07,
+      "loss": 18.2272,
+      "step": 16422
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.331770903629908e-07,
+      "loss": 0.6598,
+      "step": 16423
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.31584529281143e-07,
+      "loss": 0.68,
+      "step": 16424
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.299936869637148e-07,
+      "loss": 0.6306,
+      "step": 16425
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.284045634662062e-07,
+      "loss": 16.9722,
+      "step": 16426
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.268171588440675e-07,
+      "loss": 0.7002,
+      "step": 16427
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.252314731526655e-07,
+      "loss": 17.4191,
+      "step": 16428
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.236475064473169e-07,
+      "loss": 16.3757,
+      "step": 16429
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.220652587832888e-07,
+      "loss": 16.5037,
+      "step": 16430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.204847302157647e-07,
+      "loss": 0.6748,
+      "step": 16431
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.189059207998949e-07,
+      "loss": 0.7282,
+      "step": 16432
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.173288305907577e-07,
+      "loss": 17.6376,
+      "step": 16433
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.157534596433646e-07,
+      "loss": 0.683,
+      "step": 16434
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.141798080126771e-07,
+      "loss": 18.2717,
+      "step": 16435
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.126078757535848e-07,
+      "loss": 18.5503,
+      "step": 16436
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.110376629209381e-07,
+      "loss": 16.0854,
+      "step": 16437
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.094691695695099e-07,
+      "loss": 0.6454,
+      "step": 16438
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.079023957540177e-07,
+      "loss": 16.1424,
+      "step": 16439
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.063373415291175e-07,
+      "loss": 17.731,
+      "step": 16440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.047740069494102e-07,
+      "loss": 18.2145,
+      "step": 16441
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.032123920694356e-07,
+      "loss": 16.6847,
+      "step": 16442
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.016524969436666e-07,
+      "loss": 16.0983,
+      "step": 16443
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.000943216265266e-07,
+      "loss": 0.6685,
+      "step": 16444
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.985378661723662e-07,
+      "loss": 17.634,
+      "step": 16445
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.96983130635498e-07,
+      "loss": 16.6103,
+      "step": 16446
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.954301150701447e-07,
+      "loss": 0.6289,
+      "step": 16447
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.938788195304913e-07,
+      "loss": 18.6364,
+      "step": 16448
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.923292440706497e-07,
+      "loss": 0.6662,
+      "step": 16449
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.907813887446879e-07,
+      "loss": 0.6571,
+      "step": 16450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.892352536065961e-07,
+      "loss": 17.9639,
+      "step": 16451
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.876908387103198e-07,
+      "loss": 17.5889,
+      "step": 16452
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.861481441097328e-07,
+      "loss": 0.6522,
+      "step": 16453
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.846071698586476e-07,
+      "loss": 18.0009,
+      "step": 16454
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.830679160108266e-07,
+      "loss": 16.067,
+      "step": 16455
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.81530382619977e-07,
+      "loss": 17.7168,
+      "step": 16456
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.799945697397225e-07,
+      "loss": 16.4366,
+      "step": 16457
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.78460477423648e-07,
+      "loss": 16.5721,
+      "step": 16458
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.769281057252719e-07,
+      "loss": 0.688,
+      "step": 16459
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.753974546980457e-07,
+      "loss": 17.5428,
+      "step": 16460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.738685243953769e-07,
+      "loss": 17.2161,
+      "step": 16461
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.723413148706004e-07,
+      "loss": 18.0742,
+      "step": 16462
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.708158261769903e-07,
+      "loss": 17.1437,
+      "step": 16463
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.692920583677653e-07,
+      "loss": 16.1271,
+      "step": 16464
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.677700114960828e-07,
+      "loss": 0.6216,
+      "step": 16465
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.662496856150446e-07,
+      "loss": 0.6519,
+      "step": 16466
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.647310807776863e-07,
+      "loss": 0.6772,
+      "step": 16467
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.632141970369821e-07,
+      "loss": 16.6314,
+      "step": 16468
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.616990344458563e-07,
+      "loss": 16.0928,
+      "step": 16469
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.601855930571554e-07,
+      "loss": 16.8131,
+      "step": 16470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.58673872923693e-07,
+      "loss": 0.6712,
+      "step": 16471
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.571638740981933e-07,
+      "loss": 17.184,
+      "step": 16472
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.556555966333367e-07,
+      "loss": 0.6192,
+      "step": 16473
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.541490405817474e-07,
+      "loss": 0.6647,
+      "step": 16474
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.52644205995967e-07,
+      "loss": 18.0292,
+      "step": 16475
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.511410929285089e-07,
+      "loss": 17.7981,
+      "step": 16476
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.496397014318089e-07,
+      "loss": 0.6764,
+      "step": 16477
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.481400315582365e-07,
+      "loss": 17.5242,
+      "step": 16478
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.466420833601105e-07,
+      "loss": 17.8986,
+      "step": 16479
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.451458568896895e-07,
+      "loss": 0.7075,
+      "step": 16480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.43651352199165e-07,
+      "loss": 17.4495,
+      "step": 16481
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.42158569340684e-07,
+      "loss": 0.647,
+      "step": 16482
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.406675083663161e-07,
+      "loss": 0.689,
+      "step": 16483
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.391781693280863e-07,
+      "loss": 18.1225,
+      "step": 16484
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.37690552277942e-07,
+      "loss": 0.6615,
+      "step": 16485
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.362046572677805e-07,
+      "loss": 16.9438,
+      "step": 16486
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.347204843494437e-07,
+      "loss": 14.9332,
+      "step": 16487
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.332380335747012e-07,
+      "loss": 17.1114,
+      "step": 16488
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.317573049952785e-07,
+      "loss": 0.694,
+      "step": 16489
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.30278298662823e-07,
+      "loss": 18.5369,
+      "step": 16490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.288010146289324e-07,
+      "loss": 0.6881,
+      "step": 16491
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.273254529451544e-07,
+      "loss": 14.8102,
+      "step": 16492
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.258516136629477e-07,
+      "loss": 18.3867,
+      "step": 16493
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.243794968337435e-07,
+      "loss": 17.5706,
+      "step": 16494
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.22909102508884e-07,
+      "loss": 16.0627,
+      "step": 16495
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.214404307396726e-07,
+      "loss": 17.9696,
+      "step": 16496
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.199734815773462e-07,
+      "loss": 18.9052,
+      "step": 16497
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.185082550730803e-07,
+      "loss": 0.6041,
+      "step": 16498
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.170447512779842e-07,
+      "loss": 0.6738,
+      "step": 16499
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.15582970243117e-07,
+      "loss": 17.5995,
+      "step": 16500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.141229120194714e-07,
+      "loss": 18.163,
+      "step": 16501
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.126645766579897e-07,
+      "loss": 0.6197,
+      "step": 16502
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.112079642095426e-07,
+      "loss": 18.1486,
+      "step": 16503
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.097530747249447e-07,
+      "loss": 15.4145,
+      "step": 16504
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.082999082549501e-07,
+      "loss": 18.1963,
+      "step": 16505
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.068484648502571e-07,
+      "loss": 0.7154,
+      "step": 16506
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.053987445614972e-07,
+      "loss": 16.6284,
+      "step": 16507
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.03950747439247e-07,
+      "loss": 0.6956,
+      "step": 16508
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.025044735340157e-07,
+      "loss": 16.8656,
+      "step": 16509
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.010599228962632e-07,
+      "loss": 0.603,
+      "step": 16510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.996170955763824e-07,
+      "loss": 18.5959,
+      "step": 16511
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.981759916247109e-07,
+      "loss": 18.3425,
+      "step": 16512
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.96736611091514e-07,
+      "loss": 17.6502,
+      "step": 16513
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.952989540270127e-07,
+      "loss": 18.8087,
+      "step": 16514
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.938630204813555e-07,
+      "loss": 0.6128,
+      "step": 16515
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.924288105046416e-07,
+      "loss": 16.1377,
+      "step": 16516
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.90996324146903e-07,
+      "loss": 0.6338,
+      "step": 16517
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.895655614581108e-07,
+      "loss": 17.795,
+      "step": 16518
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.881365224881807e-07,
+      "loss": 16.3001,
+      "step": 16519
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.867092072869562e-07,
+      "loss": 0.6844,
+      "step": 16520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.852836159042474e-07,
+      "loss": 17.344,
+      "step": 16521
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.838597483897757e-07,
+      "loss": 0.6575,
+      "step": 16522
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.824376047932123e-07,
+      "loss": 16.9427,
+      "step": 16523
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.810171851641788e-07,
+      "loss": 16.6635,
+      "step": 16524
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.795984895522188e-07,
+      "loss": 16.2035,
+      "step": 16525
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.781815180068261e-07,
+      "loss": 17.8931,
+      "step": 16526
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.767662705774391e-07,
+      "loss": 19.3583,
+      "step": 16527
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.753527473134235e-07,
+      "loss": 0.7098,
+      "step": 16528
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.739409482640956e-07,
+      "loss": 0.6629,
+      "step": 16529
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.725308734786994e-07,
+      "loss": 15.8372,
+      "step": 16530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.711225230064344e-07,
+      "loss": 0.669,
+      "step": 16531
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.697158968964278e-07,
+      "loss": 15.7531,
+      "step": 16532
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.683109951977517e-07,
+      "loss": 15.9058,
+      "step": 16533
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.669078179594168e-07,
+      "loss": 18.0293,
+      "step": 16534
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.655063652303727e-07,
+      "loss": 0.6648,
+      "step": 16535
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.641066370595139e-07,
+      "loss": 17.4051,
+      "step": 16536
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.627086334956677e-07,
+      "loss": 16.992,
+      "step": 16537
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.613123545876064e-07,
+      "loss": 17.7503,
+      "step": 16538
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.59917800384041e-07,
+      "loss": 16.9918,
+      "step": 16539
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.58524970933616e-07,
+      "loss": 17.8231,
+      "step": 16540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.571338662849257e-07,
+      "loss": 16.7549,
+      "step": 16541
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.557444864865036e-07,
+      "loss": 17.123,
+      "step": 16542
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.543568315868108e-07,
+      "loss": 16.6668,
+      "step": 16543
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.529709016342644e-07,
+      "loss": 16.1702,
+      "step": 16544
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.515866966772032e-07,
+      "loss": 0.6357,
+      "step": 16545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.502042167639277e-07,
+      "loss": 19.5458,
+      "step": 16546
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.488234619426602e-07,
+      "loss": 17.1712,
+      "step": 16547
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.474444322615735e-07,
+      "loss": 16.6548,
+      "step": 16548
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.460671277687735e-07,
+      "loss": 18.416,
+      "step": 16549
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.446915485123105e-07,
+      "loss": 15.7452,
+      "step": 16550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.43317694540163e-07,
+      "loss": 16.8662,
+      "step": 16551
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.419455659002759e-07,
+      "loss": 17.4075,
+      "step": 16552
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.405751626404998e-07,
+      "loss": 16.6088,
+      "step": 16553
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.39206484808652e-07,
+      "loss": 17.8413,
+      "step": 16554
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.378395324524777e-07,
+      "loss": 0.7016,
+      "step": 16555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.364743056196664e-07,
+      "loss": 17.981,
+      "step": 16556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.351108043578412e-07,
+      "loss": 0.6999,
+      "step": 16557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.337490287145696e-07,
+      "loss": 18.5551,
+      "step": 16558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.323889787373582e-07,
+      "loss": 0.7007,
+      "step": 16559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.31030654473652e-07,
+      "loss": 17.7621,
+      "step": 16560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.296740559708413e-07,
+      "loss": 17.2297,
+      "step": 16561
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.283191832762546e-07,
+      "loss": 17.1926,
+      "step": 16562
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.269660364371487e-07,
+      "loss": 0.6078,
+      "step": 16563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.256146155007302e-07,
+      "loss": 0.5997,
+      "step": 16564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.242649205141503e-07,
+      "loss": 0.6706,
+      "step": 16565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.229169515244936e-07,
+      "loss": 0.6896,
+      "step": 16566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.21570708578778e-07,
+      "loss": 0.6878,
+      "step": 16567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.20226191723977e-07,
+      "loss": 18.258,
+      "step": 16568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.188834010069865e-07,
+      "loss": 19.0454,
+      "step": 16569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.175423364746579e-07,
+      "loss": 0.6429,
+      "step": 16570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.162029981737759e-07,
+      "loss": 0.6734,
+      "step": 16571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.148653861510532e-07,
+      "loss": 18.5057,
+      "step": 16572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.135295004531693e-07,
+      "loss": 16.4176,
+      "step": 16573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.121953411267089e-07,
+      "loss": 16.6615,
+      "step": 16574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.108629082182348e-07,
+      "loss": 0.7026,
+      "step": 16575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.095322017742154e-07,
+      "loss": 18.2268,
+      "step": 16576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.082032218410804e-07,
+      "loss": 0.6679,
+      "step": 16577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.068759684651924e-07,
+      "loss": 18.7774,
+      "step": 16578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.055504416928424e-07,
+      "loss": 0.6521,
+      "step": 16579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.042266415702879e-07,
+      "loss": 15.9731,
+      "step": 16580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.029045681437084e-07,
+      "loss": 0.6597,
+      "step": 16581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.01584221459217e-07,
+      "loss": 0.6217,
+      "step": 16582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.002656015628771e-07,
+      "loss": 0.6334,
+      "step": 16583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.989487085006906e-07,
+      "loss": 0.6911,
+      "step": 16584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.976335423186041e-07,
+      "loss": 0.6039,
+      "step": 16585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.96320103062492e-07,
+      "loss": 0.6917,
+      "step": 16586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.950083907781733e-07,
+      "loss": 17.7337,
+      "step": 16587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.936984055114169e-07,
+      "loss": 0.7124,
+      "step": 16588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.923901473079084e-07,
+      "loss": 17.4176,
+      "step": 16589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.910836162133004e-07,
+      "loss": 0.7045,
+      "step": 16590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.897788122731672e-07,
+      "loss": 15.3802,
+      "step": 16591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.884757355330282e-07,
+      "loss": 0.7305,
+      "step": 16592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.87174386038336e-07,
+      "loss": 0.6647,
+      "step": 16593
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.858747638344985e-07,
+      "loss": 17.4242,
+      "step": 16594
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.845768689668517e-07,
+      "loss": 18.4986,
+      "step": 16595
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.832807014806706e-07,
+      "loss": 0.6921,
+      "step": 16596
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.819862614211746e-07,
+      "loss": 19.2367,
+      "step": 16597
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.806935488335218e-07,
+      "loss": 0.682,
+      "step": 16598
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.794025637628097e-07,
+      "loss": 0.687,
+      "step": 16599
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.781133062540743e-07,
+      "loss": 0.6726,
+      "step": 16600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.76825776352291e-07,
+      "loss": 0.6283,
+      "step": 16601
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7553997410237917e-07,
+      "loss": 17.127,
+      "step": 16602
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7425589954919193e-07,
+      "loss": 17.6152,
+      "step": 16603
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7297355273752673e-07,
+      "loss": 18.0518,
+      "step": 16604
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.716929337121201e-07,
+      "loss": 18.1758,
+      "step": 16605
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.704140425176418e-07,
+      "loss": 0.7139,
+      "step": 16606
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.691368791987172e-07,
+      "loss": 16.6578,
+      "step": 16607
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.678614437998885e-07,
+      "loss": 15.6766,
+      "step": 16608
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6658773636566454e-07,
+      "loss": 0.6685,
+      "step": 16609
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6531575694046534e-07,
+      "loss": 0.644,
+      "step": 16610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.640455055686777e-07,
+      "loss": 18.928,
+      "step": 16611
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6277698229460486e-07,
+      "loss": 15.8825,
+      "step": 16612
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.615101871625005e-07,
+      "loss": 17.0263,
+      "step": 16613
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6024512021656805e-07,
+      "loss": 17.995,
+      "step": 16614
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5898178150092787e-07,
+      "loss": 18.0471,
+      "step": 16615
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.577201710596612e-07,
+      "loss": 18.0364,
+      "step": 16616
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5646028893678306e-07,
+      "loss": 16.8947,
+      "step": 16617
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5520213517623036e-07,
+      "loss": 17.9935,
+      "step": 16618
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.539457098219069e-07,
+      "loss": 17.665,
+      "step": 16619
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5269101291764424e-07,
+      "loss": 18.2254,
+      "step": 16620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.514380445072075e-07,
+      "loss": 17.9031,
+      "step": 16621
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5018680463430606e-07,
+      "loss": 18.7133,
+      "step": 16622
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.489372933425995e-07,
+      "loss": 17.4945,
+      "step": 16623
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4768951067566953e-07,
+      "loss": 17.5781,
+      "step": 16624
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4644345667705367e-07,
+      "loss": 19.1703,
+      "step": 16625
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.451991313902115e-07,
+      "loss": 0.5685,
+      "step": 16626
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.439565348585639e-07,
+      "loss": 16.8045,
+      "step": 16627
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4271566712544843e-07,
+      "loss": 17.8981,
+      "step": 16628
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.414765282341582e-07,
+      "loss": 16.1989,
+      "step": 16629
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.402391182279253e-07,
+      "loss": 18.1308,
+      "step": 16630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3900343714990967e-07,
+      "loss": 17.2877,
+      "step": 16631
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.377694850432323e-07,
+      "loss": 16.0822,
+      "step": 16632
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3653726195092004e-07,
+      "loss": 19.2093,
+      "step": 16633
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.353067679159772e-07,
+      "loss": 16.2641,
+      "step": 16634
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3407800298133075e-07,
+      "loss": 17.1876,
+      "step": 16635
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.328509671898351e-07,
+      "loss": 0.6985,
+      "step": 16636
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.316256605843061e-07,
+      "loss": 0.6054,
+      "step": 16637
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.304020832074762e-07,
+      "loss": 15.4209,
+      "step": 16638
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.291802351020502e-07,
+      "loss": 0.6056,
+      "step": 16639
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2796011631063835e-07,
+      "loss": 0.6319,
+      "step": 16640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2674172687581225e-07,
+      "loss": 18.1162,
+      "step": 16641
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2552506684007676e-07,
+      "loss": 17.2117,
+      "step": 16642
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2431013624587014e-07,
+      "loss": 0.6466,
+      "step": 16643
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2309693513558067e-07,
+      "loss": 0.6836,
+      "step": 16644
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2188546355153013e-07,
+      "loss": 16.3657,
+      "step": 16645
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.206757215359847e-07,
+      "loss": 0.6915,
+      "step": 16646
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.194677091311383e-07,
+      "loss": 18.1458,
+      "step": 16647
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.182614263791462e-07,
+      "loss": 17.0508,
+      "step": 16648
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.170568733220803e-07,
+      "loss": 17.2298,
+      "step": 16649
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1585405000196806e-07,
+      "loss": 0.6709,
+      "step": 16650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1465295646076485e-07,
+      "loss": 0.6568,
+      "step": 16651
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.13453592740376e-07,
+      "loss": 18.3728,
+      "step": 16652
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.122559588826402e-07,
+      "loss": 16.4333,
+      "step": 16653
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.110600549293464e-07,
+      "loss": 17.5191,
+      "step": 16654
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.098658809222e-07,
+      "loss": 15.7649,
+      "step": 16655
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.086734369028733e-07,
+      "loss": 16.5827,
+      "step": 16656
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0748272291296077e-07,
+      "loss": 16.148,
+      "step": 16657
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.062937389939958e-07,
+      "loss": 17.5824,
+      "step": 16658
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.051064851874675e-07,
+      "loss": 17.5024,
+      "step": 16659
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0392096153478163e-07,
+      "loss": 17.1071,
+      "step": 16660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0273716807731067e-07,
+      "loss": 18.5051,
+      "step": 16661
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.015551048563326e-07,
+      "loss": 0.7027,
+      "step": 16662
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.00374771913109e-07,
+      "loss": 18.1166,
+      "step": 16663
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.991961692887958e-07,
+      "loss": 0.6641,
+      "step": 16664
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9801929702452113e-07,
+      "loss": 0.6722,
+      "step": 16665
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.968441551613411e-07,
+      "loss": 17.7701,
+      "step": 16666
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.956707437402396e-07,
+      "loss": 18.7912,
+      "step": 16667
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.944990628021616e-07,
+      "loss": 18.438,
+      "step": 16668
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9332911238798566e-07,
+      "loss": 18.3623,
+      "step": 16669
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9216089253851786e-07,
+      "loss": 0.6366,
+      "step": 16670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.909944032945201e-07,
+      "loss": 17.0389,
+      "step": 16671
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8982964469667647e-07,
+      "loss": 18.3554,
+      "step": 16672
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8866661678562677e-07,
+      "loss": 17.4189,
+      "step": 16673
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8750531960194404e-07,
+      "loss": 0.6905,
+      "step": 16674
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.863457531861403e-07,
+      "loss": 0.6467,
+      "step": 16675
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.851879175786666e-07,
+      "loss": 0.7051,
+      "step": 16676
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.840318128199183e-07,
+      "loss": 15.7991,
+      "step": 16677
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.828774389502188e-07,
+      "loss": 0.6902,
+      "step": 16678
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8172479600985246e-07,
+      "loss": 0.6655,
+      "step": 16679
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.805738840390205e-07,
+      "loss": 0.6605,
+      "step": 16680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7942470307787415e-07,
+      "loss": 18.6245,
+      "step": 16681
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7827725316650355e-07,
+      "loss": 0.61,
+      "step": 16682
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7713153434493775e-07,
+      "loss": 0.6271,
+      "step": 16683
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7598754665315596e-07,
+      "loss": 0.6522,
+      "step": 16684
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7484529013105396e-07,
+      "loss": 0.6984,
+      "step": 16685
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7370476481848327e-07,
+      "loss": 0.6509,
+      "step": 16686
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7256597075523427e-07,
+      "loss": 0.6318,
+      "step": 16687
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7142890798103625e-07,
+      "loss": 0.6599,
+      "step": 16688
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7029357653555195e-07,
+      "loss": 16.4083,
+      "step": 16689
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.691599764583942e-07,
+      "loss": 17.7716,
+      "step": 16690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6802810778910347e-07,
+      "loss": 0.6701,
+      "step": 16691
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6689797056716493e-07,
+      "loss": 17.1884,
+      "step": 16692
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.657695648320081e-07,
+      "loss": 18.6499,
+      "step": 16693
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.646428906230015e-07,
+      "loss": 16.2907,
+      "step": 16694
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6351794797944705e-07,
+      "loss": 0.6478,
+      "step": 16695
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6239473694058554e-07,
+      "loss": 19.0782,
+      "step": 16696
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6127325754560236e-07,
+      "loss": 0.6709,
+      "step": 16697
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6015350983362174e-07,
+      "loss": 0.6677,
+      "step": 16698
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.590354938437124e-07,
+      "loss": 17.6015,
+      "step": 16699
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.579192096148709e-07,
+      "loss": 0.7159,
+      "step": 16700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5680465718603837e-07,
+      "loss": 18.8807,
+      "step": 16701
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.556918365961004e-07,
+      "loss": 17.6854,
+      "step": 16702
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5458074788387586e-07,
+      "loss": 17.3023,
+      "step": 16703
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5347139108813374e-07,
+      "loss": 17.5572,
+      "step": 16704
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.523637662475654e-07,
+      "loss": 0.6508,
+      "step": 16705
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5125787340081207e-07,
+      "loss": 0.7001,
+      "step": 16706
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.501537125864596e-07,
+      "loss": 17.6944,
+      "step": 16707
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4905128384302157e-07,
+      "loss": 0.6146,
+      "step": 16708
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4795058720896165e-07,
+      "loss": 16.8505,
+      "step": 16709
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.468516227226715e-07,
+      "loss": 19.1771,
+      "step": 16710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.45754390422498e-07,
+      "loss": 0.6421,
+      "step": 16711
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.446588903467163e-07,
+      "loss": 0.735,
+      "step": 16712
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.435651225335401e-07,
+      "loss": 0.6624,
+      "step": 16713
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.424730870211279e-07,
+      "loss": 16.1055,
+      "step": 16714
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4138278384757694e-07,
+      "loss": 0.7311,
+      "step": 16715
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.402942130509179e-07,
+      "loss": 15.8311,
+      "step": 16716
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.392073746691371e-07,
+      "loss": 18.3242,
+      "step": 16717
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.381222687401431e-07,
+      "loss": 17.1385,
+      "step": 16718
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.370388953017889e-07,
+      "loss": 14.9991,
+      "step": 16719
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.359572543918721e-07,
+      "loss": 17.5312,
+      "step": 16720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3487734604812915e-07,
+      "loss": 0.6115,
+      "step": 16721
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.337991703082244e-07,
+      "loss": 17.7654,
+      "step": 16722
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.327227272097777e-07,
+      "loss": 0.6311,
+      "step": 16723
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.316480167903424e-07,
+      "loss": 16.7093,
+      "step": 16724
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.305750390874107e-07,
+      "loss": 17.751,
+      "step": 16725
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.295037941384083e-07,
+      "loss": 0.7156,
+      "step": 16726
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.284342819807107e-07,
+      "loss": 16.9088,
+      "step": 16727
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2736650265162707e-07,
+      "loss": 17.4362,
+      "step": 16728
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2630045618841085e-07,
+      "loss": 17.1663,
+      "step": 16729
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.252361426282491e-07,
+      "loss": 17.7132,
+      "step": 16730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.241735620082731e-07,
+      "loss": 17.6857,
+      "step": 16731
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2311271436554214e-07,
+      "loss": 0.6765,
+      "step": 16732
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.220535997370822e-07,
+      "loss": 0.7042,
+      "step": 16733
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.209962181598247e-07,
+      "loss": 17.6292,
+      "step": 16734
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.199405696706681e-07,
+      "loss": 17.2422,
+      "step": 16735
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1888665430643837e-07,
+      "loss": 0.6599,
+      "step": 16736
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.178344721038895e-07,
+      "loss": 0.6525,
+      "step": 16737
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1678402309974765e-07,
+      "loss": 19.2786,
+      "step": 16738
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.157353073306446e-07,
+      "loss": 16.9875,
+      "step": 16739
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.146883248331678e-07,
+      "loss": 0.6781,
+      "step": 16740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1364307564384354e-07,
+      "loss": 0.6659,
+      "step": 16741
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.125995597991371e-07,
+      "loss": 16.7759,
+      "step": 16742
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.115577773354528e-07,
+      "loss": 0.646,
+      "step": 16743
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1051772828912805e-07,
+      "loss": 19.096,
+      "step": 16744
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.094794126964562e-07,
+      "loss": 16.9659,
+      "step": 16745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.084428305936471e-07,
+      "loss": 0.6248,
+      "step": 16746
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.074079820168718e-07,
+      "loss": 16.7131,
+      "step": 16747
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0637486700222927e-07,
+      "loss": 16.9567,
+      "step": 16748
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.053434855857629e-07,
+      "loss": 0.6819,
+      "step": 16749
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0431383780344934e-07,
+      "loss": 17.4528,
+      "step": 16750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0328592369120445e-07,
+      "loss": 0.5767,
+      "step": 16751
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0225974328489395e-07,
+      "loss": 18.1284,
+      "step": 16752
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0123529662032246e-07,
+      "loss": 18.3721,
+      "step": 16753
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0021258373321704e-07,
+      "loss": 0.718,
+      "step": 16754
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9919160465926577e-07,
+      "loss": 0.6504,
+      "step": 16755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9817235943407904e-07,
+      "loss": 0.6474,
+      "step": 16756
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.971548480932118e-07,
+      "loss": 18.0101,
+      "step": 16757
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9613907067216894e-07,
+      "loss": 16.5462,
+      "step": 16758
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9512502720638327e-07,
+      "loss": 16.4126,
+      "step": 16759
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.941127177312264e-07,
+      "loss": 18.5307,
+      "step": 16760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9310214228202013e-07,
+      "loss": 17.1229,
+      "step": 16761
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.920933008940141e-07,
+      "loss": 0.6501,
+      "step": 16762
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.910861936024079e-07,
+      "loss": 0.6306,
+      "step": 16763
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.900808204423289e-07,
+      "loss": 0.6456,
+      "step": 16764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.890771814488491e-07,
+      "loss": 0.6585,
+      "step": 16765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.880752766569905e-07,
+      "loss": 16.8462,
+      "step": 16766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.870751061016974e-07,
+      "loss": 15.9321,
+      "step": 16767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.860766698178696e-07,
+      "loss": 0.6469,
+      "step": 16768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.850799678403293e-07,
+      "loss": 18.6347,
+      "step": 16769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8408500020384866e-07,
+      "loss": 16.4744,
+      "step": 16770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8309176694313876e-07,
+      "loss": 16.5684,
+      "step": 16771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.821002680928497e-07,
+      "loss": 0.6165,
+      "step": 16772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8111050368757606e-07,
+      "loss": 15.9491,
+      "step": 16773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.801224737618402e-07,
+      "loss": 0.615,
+      "step": 16774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7913617835010895e-07,
+      "loss": 17.8786,
+      "step": 16775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7815161748679373e-07,
+      "loss": 0.6741,
+      "step": 16776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.771687912062393e-07,
+      "loss": 17.7289,
+      "step": 16777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7618769954273483e-07,
+      "loss": 18.1292,
+      "step": 16778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.75208342530503e-07,
+      "loss": 18.4584,
+      "step": 16779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.742307202037164e-07,
+      "loss": 17.4107,
+      "step": 16780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.732548325964701e-07,
+      "loss": 17.5929,
+      "step": 16781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7228067974281457e-07,
+      "loss": 0.6456,
+      "step": 16782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7130826167673375e-07,
+      "loss": 0.6547,
+      "step": 16783
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7033757843215047e-07,
+      "loss": 17.3632,
+      "step": 16784
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.693686300429321e-07,
+      "loss": 16.9802,
+      "step": 16785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6840141654286834e-07,
+      "loss": 0.6501,
+      "step": 16786
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6743593796571544e-07,
+      "loss": 0.6849,
+      "step": 16787
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.664721943451465e-07,
+      "loss": 17.795,
+      "step": 16788
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6551018571478456e-07,
+      "loss": 15.123,
+      "step": 16789
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.645499121081918e-07,
+      "loss": 0.6953,
+      "step": 16790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6359137355886355e-07,
+      "loss": 0.6608,
+      "step": 16791
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6263457010024526e-07,
+      "loss": 17.0203,
+      "step": 16792
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.616795017657103e-07,
+      "loss": 19.2338,
+      "step": 16793
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.60726168588582e-07,
+      "loss": 16.3204,
+      "step": 16794
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5977457060211153e-07,
+      "loss": 0.6451,
+      "step": 16795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.588247078395001e-07,
+      "loss": 16.0342,
+      "step": 16796
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.578765803338823e-07,
+      "loss": 0.7029,
+      "step": 16797
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.569301881183428e-07,
+      "loss": 0.6433,
+      "step": 16798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.55985531225883e-07,
+      "loss": 0.6422,
+      "step": 16799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.550426096894709e-07,
+      "loss": 0.6218,
+      "step": 16800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.541014235419914e-07,
+      "loss": 0.6871,
+      "step": 16801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5316197281628484e-07,
+      "loss": 16.7041,
+      "step": 16802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.522242575451195e-07,
+      "loss": 0.6078,
+      "step": 16803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5128827776121355e-07,
+      "loss": 16.6386,
+      "step": 16804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.503540334972132e-07,
+      "loss": 18.3553,
+      "step": 16805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.494215247857146e-07,
+      "loss": 17.0796,
+      "step": 16806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.484907516592527e-07,
+      "loss": 19.243,
+      "step": 16807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.475617141502906e-07,
+      "loss": 0.6479,
+      "step": 16808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4663441229124116e-07,
+      "loss": 16.6638,
+      "step": 16809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.457088461144563e-07,
+      "loss": 18.0938,
+      "step": 16810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4478501565221577e-07,
+      "loss": 18.0068,
+      "step": 16811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4386292093676044e-07,
+      "loss": 18.837,
+      "step": 16812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.42942562000259e-07,
+      "loss": 15.9026,
+      "step": 16813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.42023938874808e-07,
+      "loss": 16.4645,
+      "step": 16814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.411070515924596e-07,
+      "loss": 0.6728,
+      "step": 16815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.401919001851938e-07,
+      "loss": 17.1768,
+      "step": 16816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.392784846849516e-07,
+      "loss": 17.9248,
+      "step": 16817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3836680512358546e-07,
+      "loss": 0.6529,
+      "step": 16818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3745686153290313e-07,
+      "loss": 16.8597,
+      "step": 16819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3654865394464597e-07,
+      "loss": 17.4161,
+      "step": 16820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3564218239049972e-07,
+      "loss": 0.6803,
+      "step": 16821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3473744690209464e-07,
+      "loss": 17.2461,
+      "step": 16822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3383444751098327e-07,
+      "loss": 0.6447,
+      "step": 16823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3293318424867372e-07,
+      "loss": 0.6302,
+      "step": 16824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3203365714659641e-07,
+      "loss": 15.7427,
+      "step": 16825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3113586623614292e-07,
+      "loss": 0.6885,
+      "step": 16826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3023981154863262e-07,
+      "loss": 16.9903,
+      "step": 16827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2934549311532384e-07,
+      "loss": 17.4006,
+      "step": 16828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2845291096740827e-07,
+      "loss": 16.9001,
+      "step": 16829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.275620651360333e-07,
+      "loss": 0.6615,
+      "step": 16830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2667295565227398e-07,
+      "loss": 0.6253,
+      "step": 16831
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2578558254714443e-07,
+      "loss": 17.8039,
+      "step": 16832
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2489994585160877e-07,
+      "loss": 0.6583,
+      "step": 16833
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2401604559655344e-07,
+      "loss": 15.6967,
+      "step": 16834
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2313388181282034e-07,
+      "loss": 17.2287,
+      "step": 16835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2225345453118495e-07,
+      "loss": 17.0567,
+      "step": 16836
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2137476378236155e-07,
+      "loss": 18.2379,
+      "step": 16837
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2049780959699784e-07,
+      "loss": 17.6733,
+      "step": 16838
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1962259200569157e-07,
+      "loss": 18.6358,
+      "step": 16839
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.187491110389739e-07,
+      "loss": 0.6457,
+      "step": 16840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.178773667273204e-07,
+      "loss": 16.4136,
+      "step": 16841
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1700735910113455e-07,
+      "loss": 0.667,
+      "step": 16842
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.16139088190781e-07,
+      "loss": 16.7741,
+      "step": 16843
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1527255402653545e-07,
+      "loss": 19.1866,
+      "step": 16844
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1440775663863488e-07,
+      "loss": 18.5125,
+      "step": 16845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1354469605724403e-07,
+      "loss": 17.7603,
+      "step": 16846
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.126833723124777e-07,
+      "loss": 0.7235,
+      "step": 16847
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1182378543438408e-07,
+      "loss": 0.6654,
+      "step": 16848
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1096593545294474e-07,
+      "loss": 17.2683,
+      "step": 16849
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1010982239808575e-07,
+      "loss": 16.4987,
+      "step": 16850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0925544629967764e-07,
+      "loss": 0.7066,
+      "step": 16851
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0840280718752436e-07,
+      "loss": 17.0176,
+      "step": 16852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0755190509137435e-07,
+      "loss": 0.6483,
+      "step": 16853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0670274004090383e-07,
+      "loss": 0.6258,
+      "step": 16854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0585531206574472e-07,
+      "loss": 0.6966,
+      "step": 16855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0500962119545665e-07,
+      "loss": 15.7965,
+      "step": 16856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0416566745953826e-07,
+      "loss": 18.4729,
+      "step": 16857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.033234508874382e-07,
+      "loss": 18.2619,
+      "step": 16858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.02482971508533e-07,
+      "loss": 0.6337,
+      "step": 16859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0164422935214922e-07,
+      "loss": 0.6731,
+      "step": 16860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0080722444754118e-07,
+      "loss": 15.5895,
+      "step": 16861
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.999719568239078e-07,
+      "loss": 16.4481,
+      "step": 16862
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9913842651039794e-07,
+      "loss": 17.5305,
+      "step": 16863
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9830663353607725e-07,
+      "loss": 0.7439,
+      "step": 16864
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.97476577929967e-07,
+      "loss": 0.644,
+      "step": 16865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9664825972102728e-07,
+      "loss": 16.8949,
+      "step": 16866
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.958216789381573e-07,
+      "loss": 17.5366,
+      "step": 16867
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.949968356101839e-07,
+      "loss": 18.972,
+      "step": 16868
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.941737297658952e-07,
+      "loss": 18.834,
+      "step": 16869
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.933523614339905e-07,
+      "loss": 18.733,
+      "step": 16870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9253273064313017e-07,
+      "loss": 18.2897,
+      "step": 16871
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9171483742191353e-07,
+      "loss": 18.2971,
+      "step": 16872
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9089868179886782e-07,
+      "loss": 17.5176,
+      "step": 16873
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.900842638024647e-07,
+      "loss": 0.6728,
+      "step": 16874
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.892715834611203e-07,
+      "loss": 18.029,
+      "step": 16875
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.884606408031786e-07,
+      "loss": 17.6622,
+      "step": 16876
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8765143585693922e-07,
+      "loss": 18.8359,
+      "step": 16877
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8684396865062402e-07,
+      "loss": 17.6882,
+      "step": 16878
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8603823921239937e-07,
+      "loss": 0.6117,
+      "step": 16879
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8523424757038166e-07,
+      "loss": 15.5135,
+      "step": 16880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.844319937526151e-07,
+      "loss": 17.4349,
+      "step": 16881
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.83631477787094e-07,
+      "loss": 17.0991,
+      "step": 16882
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8283269970172934e-07,
+      "loss": 18.105,
+      "step": 16883
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8203565952440438e-07,
+      "loss": 19.2326,
+      "step": 16884
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.81240357282908e-07,
+      "loss": 0.66,
+      "step": 16885
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8044679300500132e-07,
+      "loss": 0.6464,
+      "step": 16886
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.796549667183567e-07,
+      "loss": 17.4277,
+      "step": 16887
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7886487845060195e-07,
+      "loss": 0.6293,
+      "step": 16888
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7807652822929843e-07,
+      "loss": 0.7238,
+      "step": 16889
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7728991608194634e-07,
+      "loss": 0.587,
+      "step": 16890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7650504203599593e-07,
+      "loss": 0.6697,
+      "step": 16891
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.757219061188198e-07,
+      "loss": 18.3027,
+      "step": 16892
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7494050835774044e-07,
+      "loss": 17.447,
+      "step": 16893
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7416084878001393e-07,
+      "loss": 19.2022,
+      "step": 16894
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7338292741284623e-07,
+      "loss": 0.6381,
+      "step": 16895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.726067442833712e-07,
+      "loss": 0.6322,
+      "step": 16896
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7183229941866718e-07,
+      "loss": 17.2803,
+      "step": 16897
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7105959284575147e-07,
+      "loss": 0.6806,
+      "step": 16898
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.702886245915858e-07,
+      "loss": 17.9355,
+      "step": 16899
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6951939468305422e-07,
+      "loss": 19.2887,
+      "step": 16900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6875190314700195e-07,
+      "loss": 0.6755,
+      "step": 16901
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.67986150010202e-07,
+      "loss": 17.5836,
+      "step": 16902
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6722213529936082e-07,
+      "loss": 17.7126,
+      "step": 16903
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6645985904114035e-07,
+      "loss": 17.4511,
+      "step": 16904
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.656993212621305e-07,
+      "loss": 18.1601,
+      "step": 16905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6494052198886555e-07,
+      "loss": 0.6019,
+      "step": 16906
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.641834612478077e-07,
+      "loss": 16.8715,
+      "step": 16907
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.634281390653747e-07,
+      "loss": 18.5949,
+      "step": 16908
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6267455546791764e-07,
+      "loss": 0.6308,
+      "step": 16909
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6192271048172113e-07,
+      "loss": 0.6723,
+      "step": 16910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.611726041330197e-07,
+      "loss": 16.7332,
+      "step": 16911
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6042423644798132e-07,
+      "loss": 0.645,
+      "step": 16912
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5967760745270177e-07,
+      "loss": 18.3031,
+      "step": 16913
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5893271717324353e-07,
+      "loss": 18.4205,
+      "step": 16914
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.581895656355803e-07,
+      "loss": 17.5867,
+      "step": 16915
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.574481528656413e-07,
+      "loss": 15.797,
+      "step": 16916
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5670847888929475e-07,
+      "loss": 0.6119,
+      "step": 16917
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5597054373234222e-07,
+      "loss": 0.6125,
+      "step": 16918
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5523434742052977e-07,
+      "loss": 16.7365,
+      "step": 16919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5449988997953135e-07,
+      "loss": 16.2005,
+      "step": 16920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5376717143497642e-07,
+      "loss": 0.6474,
+      "step": 16921
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.530361918124279e-07,
+      "loss": 18.0683,
+      "step": 16922
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5230695113738758e-07,
+      "loss": 0.635,
+      "step": 16923
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5157944943528512e-07,
+      "loss": 0.698,
+      "step": 16924
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.508536867315058e-07,
+      "loss": 17.7876,
+      "step": 16925
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5012966305137377e-07,
+      "loss": 18.5958,
+      "step": 16926
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4940737842014108e-07,
+      "loss": 17.2277,
+      "step": 16927
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4868683286300977e-07,
+      "loss": 19.3241,
+      "step": 16928
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4796802640510977e-07,
+      "loss": 0.7606,
+      "step": 16929
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4725095907151542e-07,
+      "loss": 0.6682,
+      "step": 16930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4653563088725676e-07,
+      "loss": 0.6631,
+      "step": 16931
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.458220418772749e-07,
+      "loss": 16.5583,
+      "step": 16932
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4511019206647214e-07,
+      "loss": 16.7445,
+      "step": 16933
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.444000814796731e-07,
+      "loss": 16.0348,
+      "step": 16934
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4369171014165794e-07,
+      "loss": 0.6662,
+      "step": 16935
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4298507807713468e-07,
+      "loss": 0.639,
+      "step": 16936
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4228018531076136e-07,
+      "loss": 19.523,
+      "step": 16937
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4157703186711835e-07,
+      "loss": 17.8252,
+      "step": 16938
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4087561777074155e-07,
+      "loss": 0.6994,
+      "step": 16939
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4017594304610027e-07,
+      "loss": 18.9216,
+      "step": 16940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3947800771760276e-07,
+      "loss": 0.6342,
+      "step": 16941
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3878181180959627e-07,
+      "loss": 0.7223,
+      "step": 16942
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3808735534636686e-07,
+      "loss": 0.6368,
+      "step": 16943
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3739463835214516e-07,
+      "loss": 16.3034,
+      "step": 16944
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.367036608510952e-07,
+      "loss": 17.7367,
+      "step": 16945
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.360144228673199e-07,
+      "loss": 0.6507,
+      "step": 16946
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3532692442486672e-07,
+      "loss": 0.6287,
+      "step": 16947
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3464116554772198e-07,
+      "loss": 0.5924,
+      "step": 16948
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3395714625979995e-07,
+      "loss": 17.7648,
+      "step": 16949
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3327486658497035e-07,
+      "loss": 18.396,
+      "step": 16950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.325943265470364e-07,
+      "loss": 17.1235,
+      "step": 16951
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3191552616973468e-07,
+      "loss": 0.6128,
+      "step": 16952
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3123846547674624e-07,
+      "loss": 17.525,
+      "step": 16953
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3056314449169105e-07,
+      "loss": 17.6531,
+      "step": 16954
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2988956323812808e-07,
+      "loss": 18.2557,
+      "step": 16955
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.292177217395607e-07,
+      "loss": 0.6711,
+      "step": 16956
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2854762001942023e-07,
+      "loss": 0.6528,
+      "step": 16957
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2787925810108236e-07,
+      "loss": 0.668,
+      "step": 16958
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2721263600787293e-07,
+      "loss": 0.6328,
+      "step": 16959
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2654775376303442e-07,
+      "loss": 16.2699,
+      "step": 16960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2588461138977604e-07,
+      "loss": 16.0312,
+      "step": 16961
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.252232089112182e-07,
+      "loss": 0.7193,
+      "step": 16962
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2456354635044797e-07,
+      "loss": 16.0999,
+      "step": 16963
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2390562373046367e-07,
+      "loss": 15.9131,
+      "step": 16964
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.232494410742302e-07,
+      "loss": 16.3678,
+      "step": 16965
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.225949984046293e-07,
+      "loss": 17.0198,
+      "step": 16966
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2194229574449824e-07,
+      "loss": 16.2714,
+      "step": 16967
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2129133311660768e-07,
+      "loss": 17.9415,
+      "step": 16968
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2064211054365614e-07,
+      "loss": 0.6599,
+      "step": 16969
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1999462804830887e-07,
+      "loss": 0.6906,
+      "step": 16970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1934888565313662e-07,
+      "loss": 15.4828,
+      "step": 16971
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1870488338068254e-07,
+      "loss": 17.2758,
+      "step": 16972
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1806262125340084e-07,
+      "loss": 17.6941,
+      "step": 16973
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.174220992937014e-07,
+      "loss": 17.4577,
+      "step": 16974
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1678331752393302e-07,
+      "loss": 0.6199,
+      "step": 16975
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1614627596637784e-07,
+      "loss": 18.4898,
+      "step": 16976
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.15510974643257e-07,
+      "loss": 18.4076,
+      "step": 16977
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.148774135767361e-07,
+      "loss": 0.6158,
+      "step": 16978
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1424559278890856e-07,
+      "loss": 17.0806,
+      "step": 16979
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1361551230183453e-07,
+      "loss": 17.7521,
+      "step": 16980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1298717213747978e-07,
+      "loss": 18.0497,
+      "step": 16981
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1236057231776565e-07,
+      "loss": 0.6905,
+      "step": 16982
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1173571286455797e-07,
+      "loss": 17.043,
+      "step": 16983
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1111259379965044e-07,
+      "loss": 0.6839,
+      "step": 16984
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1049121514478123e-07,
+      "loss": 0.6888,
+      "step": 16985
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0987157692162742e-07,
+      "loss": 18.5057,
+      "step": 16986
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0925367915181617e-07,
+      "loss": 0.7131,
+      "step": 16987
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.086375218568858e-07,
+      "loss": 19.782,
+      "step": 16988
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0802310505834135e-07,
+      "loss": 17.2507,
+      "step": 16989
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0741042877761564e-07,
+      "loss": 0.6435,
+      "step": 16990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.067994930360805e-07,
+      "loss": 0.6651,
+      "step": 16991
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0619029785505218e-07,
+      "loss": 0.6746,
+      "step": 16992
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0558284325578038e-07,
+      "loss": 15.9375,
+      "step": 16993
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0497712925945923e-07,
+      "loss": 0.6556,
+      "step": 16994
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0437315588722185e-07,
+      "loss": 0.7302,
+      "step": 16995
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0377092316012915e-07,
+      "loss": 18.9905,
+      "step": 16996
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0317043109919212e-07,
+      "loss": 16.6374,
+      "step": 16997
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.025716797253662e-07,
+      "loss": 0.7306,
+      "step": 16998
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0197466905954022e-07,
+      "loss": 16.8828,
+      "step": 16999
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0137939912253091e-07,
+      "loss": 0.6641,
+      "step": 17000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0078586993511052e-07,
+      "loss": 0.6584,
+      "step": 17001
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0019408151799026e-07,
+      "loss": 0.7447,
+      "step": 17002
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.960403389180361e-08,
+      "loss": 19.3571,
+      "step": 17003
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.901572707714524e-08,
+      "loss": 0.6658,
+      "step": 17004
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.842916109453204e-08,
+      "loss": 0.698,
+      "step": 17005
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.784433596443098e-08,
+      "loss": 0.7018,
+      "step": 17006
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.726125170724243e-08,
+      "loss": 18.9566,
+      "step": 17007
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.667990834330564e-08,
+      "loss": 16.0549,
+      "step": 17008
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.610030589290442e-08,
+      "loss": 17.8417,
+      "step": 17009
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.552244437626145e-08,
+      "loss": 16.6719,
+      "step": 17010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.49463238135273e-08,
+      "loss": 0.7095,
+      "step": 17011
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.437194422480256e-08,
+      "loss": 18.1283,
+      "step": 17012
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.379930563012674e-08,
+      "loss": 0.6945,
+      "step": 17013
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.322840804948385e-08,
+      "loss": 16.9952,
+      "step": 17014
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.265925150278021e-08,
+      "loss": 17.2278,
+      "step": 17015
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.209183600987214e-08,
+      "loss": 17.3861,
+      "step": 17016
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.152616159056049e-08,
+      "loss": 16.8745,
+      "step": 17017
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.096222826456836e-08,
+      "loss": 16.4008,
+      "step": 17018
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.040003605157998e-08,
+      "loss": 15.8906,
+      "step": 17019
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.983958497120748e-08,
+      "loss": 16.678,
+      "step": 17020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.928087504299632e-08,
+      "loss": 0.676,
+      "step": 17021
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.872390628643646e-08,
+      "loss": 17.1201,
+      "step": 17022
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.816867872096235e-08,
+      "loss": 0.6764,
+      "step": 17023
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.761519236594739e-08,
+      "loss": 0.652,
+      "step": 17024
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.706344724069281e-08,
+      "loss": 16.8936,
+      "step": 17025
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.651344336444988e-08,
+      "loss": 18.0306,
+      "step": 17026
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.59651807564088e-08,
+      "loss": 16.9826,
+      "step": 17027
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.541865943568761e-08,
+      "loss": 16.4951,
+      "step": 17028
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.48738794213655e-08,
+      "loss": 18.1727,
+      "step": 17029
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.433084073243281e-08,
+      "loss": 16.4433,
+      "step": 17030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.378954338784107e-08,
+      "loss": 0.6557,
+      "step": 17031
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.324998740647516e-08,
+      "loss": 0.6301,
+      "step": 17032
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.271217280715893e-08,
+      "loss": 0.6509,
+      "step": 17033
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.217609960864403e-08,
+      "loss": 19.3965,
+      "step": 17034
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.164176782964328e-08,
+      "loss": 0.6461,
+      "step": 17035
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.110917748879731e-08,
+      "loss": 0.6727,
+      "step": 17036
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.057832860467462e-08,
+      "loss": 16.5099,
+      "step": 17037
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.004922119580482e-08,
+      "loss": 17.3576,
+      "step": 17038
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.952185528064538e-08,
+      "loss": 17.5081,
+      "step": 17039
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.899623087758712e-08,
+      "loss": 17.1181,
+      "step": 17040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.84723480049765e-08,
+      "loss": 0.6488,
+      "step": 17041
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.795020668107667e-08,
+      "loss": 0.6595,
+      "step": 17042
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.74298069241175e-08,
+      "loss": 18.2408,
+      "step": 17043
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.691114875224003e-08,
+      "loss": 0.6079,
+      "step": 17044
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.639423218355202e-08,
+      "loss": 17.918,
+      "step": 17045
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.587905723607235e-08,
+      "loss": 0.6976,
+      "step": 17046
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.536562392778112e-08,
+      "loss": 0.6553,
+      "step": 17047
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.485393227658622e-08,
+      "loss": 17.4811,
+      "step": 17048
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.434398230034556e-08,
+      "loss": 17.7923,
+      "step": 17049
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.383577401684493e-08,
+      "loss": 17.83,
+      "step": 17050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.332930744380906e-08,
+      "loss": 18.4769,
+      "step": 17051
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.282458259890712e-08,
+      "loss": 17.0661,
+      "step": 17052
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.23215994997528e-08,
+      "loss": 0.697,
+      "step": 17053
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.18203581638932e-08,
+      "loss": 0.6839,
+      "step": 17054
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.132085860880878e-08,
+      "loss": 17.6787,
+      "step": 17055
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.082310085193e-08,
+      "loss": 17.3718,
+      "step": 17056
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.032708491061524e-08,
+      "loss": 18.2669,
+      "step": 17057
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.983281080217285e-08,
+      "loss": 17.7642,
+      "step": 17058
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.934027854384461e-08,
+      "loss": 0.543,
+      "step": 17059
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.884948815281678e-08,
+      "loss": 0.6452,
+      "step": 17060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.836043964620342e-08,
+      "loss": 0.6912,
+      "step": 17061
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.787313304107424e-08,
+      "loss": 0.6791,
+      "step": 17062
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.738756835442117e-08,
+      "loss": 16.6878,
+      "step": 17063
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.690374560318625e-08,
+      "loss": 16.1516,
+      "step": 17064
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.642166480425038e-08,
+      "loss": 0.6521,
+      "step": 17065
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.594132597442793e-08,
+      "loss": 0.6631,
+      "step": 17066
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.546272913047769e-08,
+      "loss": 0.6356,
+      "step": 17067
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.498587428909742e-08,
+      "loss": 16.9457,
+      "step": 17068
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.451076146692381e-08,
+      "loss": 17.1651,
+      "step": 17069
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.403739068052139e-08,
+      "loss": 16.2963,
+      "step": 17070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.356576194641583e-08,
+      "loss": 18.5598,
+      "step": 17071
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.30958752810551e-08,
+      "loss": 0.667,
+      "step": 17072
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.262773070083162e-08,
+      "loss": 0.6583,
+      "step": 17073
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.216132822207676e-08,
+      "loss": 0.7004,
+      "step": 17074
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.169666786106088e-08,
+      "loss": 16.7375,
+      "step": 17075
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.123374963399875e-08,
+      "loss": 17.9028,
+      "step": 17076
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.077257355703303e-08,
+      "loss": 18.7456,
+      "step": 17077
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.031313964625641e-08,
+      "loss": 16.4362,
+      "step": 17078
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.985544791768937e-08,
+      "loss": 0.6386,
+      "step": 17079
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.9399498387313625e-08,
+      "loss": 15.5401,
+      "step": 17080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.894529107102198e-08,
+      "loss": 0.6683,
+      "step": 17081
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.849282598466288e-08,
+      "loss": 17.1885,
+      "step": 17082
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.804210314402369e-08,
+      "loss": 16.3113,
+      "step": 17083
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7593122564825184e-08,
+      "loss": 17.4028,
+      "step": 17084
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.71458842627326e-08,
+      "loss": 0.6211,
+      "step": 17085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.670038825335011e-08,
+      "loss": 17.3858,
+      "step": 17086
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.62566345522153e-08,
+      "loss": 20.1603,
+      "step": 17087
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.581462317480468e-08,
+      "loss": 18.2903,
+      "step": 17088
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.5374354136550344e-08,
+      "loss": 17.1351,
+      "step": 17089
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4935827452801127e-08,
+      "loss": 18.2073,
+      "step": 17090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.449904313886145e-08,
+      "loss": 17.4631,
+      "step": 17091
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4064001209963576e-08,
+      "loss": 0.6431,
+      "step": 17092
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3630701681284256e-08,
+      "loss": 17.2703,
+      "step": 17093
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.319914456794473e-08,
+      "loss": 0.5729,
+      "step": 17094
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.276932988499961e-08,
+      "loss": 0.6434,
+      "step": 17095
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.234125764744246e-08,
+      "loss": 0.6502,
+      "step": 17096
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.191492787020025e-08,
+      "loss": 17.3168,
+      "step": 17097
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.149034056815549e-08,
+      "loss": 0.6263,
+      "step": 17098
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1067495756113025e-08,
+      "loss": 0.6634,
+      "step": 17099
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.064639344882771e-08,
+      "loss": 16.9452,
+      "step": 17100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0227033660987796e-08,
+      "loss": 18.2291,
+      "step": 17101
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.980941640722603e-08,
+      "loss": 17.7741,
+      "step": 17102
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.939354170210852e-08,
+      "loss": 17.9949,
+      "step": 17103
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8979409560145905e-08,
+      "loss": 0.7277,
+      "step": 17104
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.856701999578772e-08,
+      "loss": 15.9764,
+      "step": 17105
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.815637302341136e-08,
+      "loss": 0.6471,
+      "step": 17106
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7747468657349806e-08,
+      "loss": 0.6593,
+      "step": 17107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.734030691186386e-08,
+      "loss": 17.2783,
+      "step": 17108
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6934887801164396e-08,
+      "loss": 17.2057,
+      "step": 17109
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6531211339390094e-08,
+      "loss": 0.6433,
+      "step": 17110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6129277540618575e-08,
+      "loss": 0.6317,
+      "step": 17111
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.572908641887752e-08,
+      "loss": 0.6528,
+      "step": 17112
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5330637988133526e-08,
+      "loss": 16.8935,
+      "step": 17113
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.493393226227549e-08,
+      "loss": 0.6943,
+      "step": 17114
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.453896925514789e-08,
+      "loss": 0.6602,
+      "step": 17115
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.414574898052859e-08,
+      "loss": 15.606,
+      "step": 17116
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.375427145213995e-08,
+      "loss": 17.9659,
+      "step": 17117
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.336453668362661e-08,
+      "loss": 0.6571,
+      "step": 17118
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.297654468859991e-08,
+      "loss": 0.6501,
+      "step": 17119
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2590295480587903e-08,
+      "loss": 17.0245,
+      "step": 17120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.220578907306316e-08,
+      "loss": 17.1986,
+      "step": 17121
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1823025479442725e-08,
+      "loss": 16.5694,
+      "step": 17122
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.144200471307702e-08,
+      "loss": 0.6727,
+      "step": 17123
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.106272678726653e-08,
+      "loss": 17.8452,
+      "step": 17124
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.068519171522844e-08,
+      "loss": 17.7011,
+      "step": 17125
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0309399510146674e-08,
+      "loss": 18.821,
+      "step": 17126
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.993535018512184e-08,
+      "loss": 0.6245,
+      "step": 17127
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.956304375320463e-08,
+      "loss": 17.1479,
+      "step": 17128
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9192480227390195e-08,
+      "loss": 0.6467,
+      "step": 17129
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.882365962059598e-08,
+      "loss": 16.4363,
+      "step": 17130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8456581945695014e-08,
+      "loss": 16.0626,
+      "step": 17131
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.809124721548818e-08,
+      "loss": 16.8339,
+      "step": 17132
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7727655442726385e-08,
+      "loss": 16.7926,
+      "step": 17133
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.736580664009393e-08,
+      "loss": 16.3517,
+      "step": 17134
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.700570082020849e-08,
+      "loss": 16.2012,
+      "step": 17135
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.66473379956378e-08,
+      "loss": 0.6618,
+      "step": 17136
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.629071817887742e-08,
+      "loss": 0.6574,
+      "step": 17137
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.593584138237294e-08,
+      "loss": 17.1316,
+      "step": 17138
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.558270761850335e-08,
+      "loss": 18.2996,
+      "step": 17139
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.523131689959214e-08,
+      "loss": 16.8627,
+      "step": 17140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4881669237890604e-08,
+      "loss": 18.1094,
+      "step": 17141
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4533764645600096e-08,
+      "loss": 16.7907,
+      "step": 17142
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.418760313485536e-08,
+      "loss": 16.5676,
+      "step": 17143
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3843184717735624e-08,
+      "loss": 20.0257,
+      "step": 17144
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.350050940625349e-08,
+      "loss": 0.6231,
+      "step": 17145
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3159577212366064e-08,
+      "loss": 0.6977,
+      "step": 17146
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.282038814796939e-08,
+      "loss": 17.5021,
+      "step": 17147
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.248294222488735e-08,
+      "loss": 0.6091,
+      "step": 17148
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.214723945489939e-08,
+      "loss": 0.6084,
+      "step": 17149
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.181327984971838e-08,
+      "loss": 19.7125,
+      "step": 17150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1481063420985e-08,
+      "loss": 0.6555,
+      "step": 17151
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1150590180295536e-08,
+      "loss": 17.8427,
+      "step": 17152
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.082186013917965e-08,
+      "loss": 19.1414,
+      "step": 17153
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0494873309105945e-08,
+      "loss": 17.245,
+      "step": 17154
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.016962970147641e-08,
+      "loss": 17.3962,
+      "step": 17155
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.984612932764308e-08,
+      "loss": 15.8906,
+      "step": 17156
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.952437219888582e-08,
+      "loss": 0.6288,
+      "step": 17157
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9204358326434533e-08,
+      "loss": 0.6287,
+      "step": 17158
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8886087721452513e-08,
+      "loss": 0.7172,
+      "step": 17159
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8569560395041994e-08,
+      "loss": 17.4243,
+      "step": 17160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8254776358238588e-08,
+      "loss": 16.6603,
+      "step": 17161
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7941735622033506e-08,
+      "loss": 15.3723,
+      "step": 17162
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7630438197345787e-08,
+      "loss": 16.3254,
+      "step": 17163
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7320884095033417e-08,
+      "loss": 17.3392,
+      "step": 17164
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.701307332589331e-08,
+      "loss": 18.2258,
+      "step": 17165
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.670700590066133e-08,
+      "loss": 0.6503,
+      "step": 17166
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6402681830023367e-08,
+      "loss": 0.6741,
+      "step": 17167
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6100101124593158e-08,
+      "loss": 0.6705,
+      "step": 17168
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5799263794923367e-08,
+      "loss": 16.6127,
+      "step": 17169
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.550016985151116e-08,
+      "loss": 0.6812,
+      "step": 17170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5202819304787074e-08,
+      "loss": 18.7324,
+      "step": 17171
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4907212165131698e-08,
+      "loss": 0.6683,
+      "step": 17172
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4613348442847905e-08,
+      "loss": 0.6433,
+      "step": 17173
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4321228148199703e-08,
+      "loss": 20.2698,
+      "step": 17174
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4030851291362288e-08,
+      "loss": 0.6768,
+      "step": 17175
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3742217882483097e-08,
+      "loss": 0.6845,
+      "step": 17176
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.34553279316152e-08,
+      "loss": 0.6333,
+      "step": 17177
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.317018144877836e-08,
+      "loss": 17.3277,
+      "step": 17178
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2886778443914626e-08,
+      "loss": 17.2028,
+      "step": 17179
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.260511892691053e-08,
+      "loss": 0.6554,
+      "step": 17180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2325202907597098e-08,
+      "loss": 18.4422,
+      "step": 17181
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2047030395733193e-08,
+      "loss": 0.6017,
+      "step": 17182
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.177060140102771e-08,
+      "loss": 0.6666,
+      "step": 17183
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1495915933122946e-08,
+      "loss": 0.6377,
+      "step": 17184
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1222974001594565e-08,
+      "loss": 0.6999,
+      "step": 17185
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0951775615973833e-08,
+      "loss": 17.371,
+      "step": 17186
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0682320785719854e-08,
+      "loss": 18.5996,
+      "step": 17187
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0414609520230665e-08,
+      "loss": 17.2302,
+      "step": 17188
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.014864182884324e-08,
+      "loss": 0.6618,
+      "step": 17189
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9884417720844594e-08,
+      "loss": 0.7112,
+      "step": 17190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.962193720543848e-08,
+      "loss": 17.5706,
+      "step": 17191
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9361200291795334e-08,
+      "loss": 0.7475,
+      "step": 17192
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9102206989007886e-08,
+      "loss": 18.2648,
+      "step": 17193
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8844957306102253e-08,
+      "loss": 0.6507,
+      "step": 17194
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8589451252065683e-08,
+      "loss": 16.5067,
+      "step": 17195
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8335688835802167e-08,
+      "loss": 0.6328,
+      "step": 17196
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8083670066171288e-08,
+      "loss": 18.7888,
+      "step": 17197
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7833394951954907e-08,
+      "loss": 0.667,
+      "step": 17198
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7584863501896033e-08,
+      "loss": 0.6875,
+      "step": 17199
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7338075724659954e-08,
+      "loss": 17.5321,
+      "step": 17200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7093031628850898e-08,
+      "loss": 0.6203,
+      "step": 17201
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6849731223028687e-08,
+      "loss": 0.6325,
+      "step": 17202
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.660817451566987e-08,
+      "loss": 0.649,
+      "step": 17203
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.63683615152066e-08,
+      "loss": 17.4937,
+      "step": 17204
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6130292230004395e-08,
+      "loss": 18.1775,
+      "step": 17205
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.589396666836773e-08,
+      "loss": 17.2423,
+      "step": 17206
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5659384838545566e-08,
+      "loss": 0.6689,
+      "step": 17207
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5426546748709135e-08,
+      "loss": 0.666,
+      "step": 17208
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.519545240699638e-08,
+      "loss": 0.6379,
+      "step": 17209
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4966101821461964e-08,
+      "loss": 0.705,
+      "step": 17210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4738495000105045e-08,
+      "loss": 19.0239,
+      "step": 17211
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4512631950869271e-08,
+      "loss": 0.6994,
+      "step": 17212
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4288512681631672e-08,
+      "loss": 0.689,
+      "step": 17213
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4066137200213769e-08,
+      "loss": 16.9544,
+      "step": 17214
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3845505514364921e-08,
+      "loss": 0.6434,
+      "step": 17215
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3626617631795624e-08,
+      "loss": 0.6182,
+      "step": 17216
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3409473560133112e-08,
+      "loss": 17.2892,
+      "step": 17217
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3194073306954657e-08,
+      "loss": 0.6449,
+      "step": 17218
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.298041687977647e-08,
+      "loss": 15.8691,
+      "step": 17219
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2768504286042594e-08,
+      "loss": 0.6161,
+      "step": 17220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2558335533158217e-08,
+      "loss": 18.0992,
+      "step": 17221
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2349910628450812e-08,
+      "loss": 17.9961,
+      "step": 17222
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2143229579192338e-08,
+      "loss": 16.1529,
+      "step": 17223
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1938292392588147e-08,
+      "loss": 0.7232,
+      "step": 17224
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.173509907579362e-08,
+      "loss": 17.8939,
+      "step": 17225
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1533649635891986e-08,
+      "loss": 0.6925,
+      "step": 17226
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1333944079916503e-08,
+      "loss": 18.8728,
+      "step": 17227
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1135982414828272e-08,
+      "loss": 0.6799,
+      "step": 17228
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.093976464753843e-08,
+      "loss": 17.9808,
+      "step": 17229
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0745290784891504e-08,
+      "loss": 0.6934,
+      "step": 17230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0552560833665403e-08,
+      "loss": 17.1262,
+      "step": 17231
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0361574800593632e-08,
+      "loss": 0.6422,
+      "step": 17232
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0172332692331976e-08,
+      "loss": 16.6462,
+      "step": 17233
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.984834515480712e-09,
+      "loss": 17.9553,
+      "step": 17234
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.799080276590156e-09,
+      "loss": 18.3923,
+      "step": 17235
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.615069982132907e-09,
+      "loss": 18.3243,
+      "step": 17236
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.432803638526056e-09,
+      "loss": 0.7845,
+      "step": 17237
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.25228125213673e-09,
+      "loss": 17.9969,
+      "step": 17238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.073502829254344e-09,
+      "loss": 0.6469,
+      "step": 17239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.896468376123901e-09,
+      "loss": 18.6314,
+      "step": 17240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.72117789891269e-09,
+      "loss": 0.678,
+      "step": 17241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.547631403738044e-09,
+      "loss": 16.4661,
+      "step": 17242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.375828896661775e-09,
+      "loss": 16.3888,
+      "step": 17243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.205770383667988e-09,
+      "loss": 20.3842,
+      "step": 17244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.037455870696375e-09,
+      "loss": 17.2595,
+      "step": 17245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.870885363608915e-09,
+      "loss": 17.1593,
+      "step": 17246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.706058868228727e-09,
+      "loss": 0.6472,
+      "step": 17247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.542976390295663e-09,
+      "loss": 17.4584,
+      "step": 17248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.381637935510721e-09,
+      "loss": 16.7976,
+      "step": 17249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.222043509486076e-09,
+      "loss": 18.1365,
+      "step": 17250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.06419311780615e-09,
+      "loss": 15.6843,
+      "step": 17251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.908086765966548e-09,
+      "loss": 0.6971,
+      "step": 17252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.753724459412914e-09,
+      "loss": 16.0362,
+      "step": 17253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.601106203535379e-09,
+      "loss": 16.8728,
+      "step": 17254
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.450232003657464e-09,
+      "loss": 19.8385,
+      "step": 17255
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.301101865041625e-09,
+      "loss": 0.6276,
+      "step": 17256
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.153715792894809e-09,
+      "loss": 15.5701,
+      "step": 17257
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.008073792346247e-09,
+      "loss": 0.6237,
+      "step": 17258
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.864175868491861e-09,
+      "loss": 16.2544,
+      "step": 17259
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.722022026338758e-09,
+      "loss": 16.3671,
+      "step": 17260
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.581612270855186e-09,
+      "loss": 17.8146,
+      "step": 17261
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.442946606937227e-09,
+      "loss": 16.3005,
+      "step": 17262
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.306025039414353e-09,
+      "loss": 18.8417,
+      "step": 17263
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.170847573077176e-09,
+      "loss": 16.096,
+      "step": 17264
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.03741421263304e-09,
+      "loss": 17.1405,
+      "step": 17265
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.905724962733782e-09,
+      "loss": 17.0862,
+      "step": 17266
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7757798279812745e-09,
+      "loss": 0.6417,
+      "step": 17267
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.647578812905229e-09,
+      "loss": 18.4783,
+      "step": 17268
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.521121921974292e-09,
+      "loss": 16.3041,
+      "step": 17269
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.396409159607151e-09,
+      "loss": 16.583,
+      "step": 17270
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.273440530155881e-09,
+      "loss": 0.6723,
+      "step": 17271
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1522160379003915e-09,
+      "loss": 0.6487,
+      "step": 17272
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.032735687070632e-09,
+      "loss": 0.6756,
+      "step": 17273
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.914999481846593e-09,
+      "loss": 16.5107,
+      "step": 17274
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.799007426324996e-09,
+      "loss": 0.7047,
+      "step": 17275
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.6847595245526056e-09,
+      "loss": 16.5512,
+      "step": 17276
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5722557805206726e-09,
+      "loss": 18.6104,
+      "step": 17277
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.461496198153835e-09,
+      "loss": 17.0315,
+      "step": 17278
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.352480781310119e-09,
+      "loss": 0.6436,
+      "step": 17279
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2452095337920374e-09,
+      "loss": 0.6706,
+      "step": 17280
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1396824593521446e-09,
+      "loss": 16.0376,
+      "step": 17281
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0358995616597275e-09,
+      "loss": 16.7329,
+      "step": 17282
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9338608443452154e-09,
+      "loss": 16.0606,
+      "step": 17283
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.833566310961322e-09,
+      "loss": 0.7049,
+      "step": 17284
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.735015965010801e-09,
+      "loss": 0.6737,
+      "step": 17285
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6382098099242413e-09,
+      "loss": 17.1328,
+      "step": 17286
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5431478490933745e-09,
+      "loss": 18.4772,
+      "step": 17287
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4498300858211144e-09,
+      "loss": 0.6811,
+      "step": 17288
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.358256523371516e-09,
+      "loss": 0.6908,
+      "step": 17289
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2684271649309195e-09,
+      "loss": 0.6408,
+      "step": 17290
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1803420136412565e-09,
+      "loss": 18.5008,
+      "step": 17291
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0940010725722935e-09,
+      "loss": 18.5768,
+      "step": 17292
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.009404344732735e-09,
+      "loss": 0.6841,
+      "step": 17293
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9265518330757737e-09,
+      "loss": 17.466,
+      "step": 17294
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.845443540493541e-09,
+      "loss": 17.0001,
+      "step": 17295
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7660794698171057e-09,
+      "loss": 19.1955,
+      "step": 17296
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.688459623810923e-09,
+      "loss": 0.6813,
+      "step": 17297
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.612584005183937e-09,
+      "loss": 0.6804,
+      "step": 17298
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5384526165895807e-09,
+      "loss": 0.6617,
+      "step": 17299
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4660654606035717e-09,
+      "loss": 0.6743,
+      "step": 17300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3954225397516674e-09,
+      "loss": 17.7936,
+      "step": 17301
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.326523856509665e-09,
+      "loss": 0.6733,
+      "step": 17302
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.259369413270095e-09,
+      "loss": 17.7897,
+      "step": 17303
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1939592123810794e-09,
+      "loss": 17.7515,
+      "step": 17304
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1302932561241264e-09,
+      "loss": 18.8721,
+      "step": 17305
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0683715467141309e-09,
+      "loss": 18.4014,
+      "step": 17306
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.008194086321579e-09,
+      "loss": 0.6466,
+      "step": 17307
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.497608770392408e-10,
+      "loss": 17.2043,
+      "step": 17308
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.930719209043759e-10,
+      "loss": 0.6509,
+      "step": 17309
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.381272198987322e-10,
+      "loss": 0.66,
+      "step": 17310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.849267759374446e-10,
+      "loss": 0.6674,
+      "step": 17311
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.334705908745854e-10,
+      "loss": 16.6204,
+      "step": 17312
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.837586665087159e-10,
+      "loss": 16.7623,
+      "step": 17313
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.357910045773352e-10,
+      "loss": 0.6392,
+      "step": 17314
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.895676067402268e-10,
+      "loss": 0.6265,
+      "step": 17315
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.450884746238672e-10,
+      "loss": 18.0863,
+      "step": 17316
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.023536097770176e-10,
+      "loss": 0.6179,
+      "step": 17317
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.613630136818259e-10,
+      "loss": 18.708,
+      "step": 17318
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2211668777603075e-10,
+      "loss": 18.1145,
+      "step": 17319
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8461463343075767e-10,
+      "loss": 0.6154,
+      "step": 17320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4885685195051867e-10,
+      "loss": 18.7293,
+      "step": 17321
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1484334457876353e-10,
+      "loss": 19.5781,
+      "step": 17322
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8257411251453316e-10,
+      "loss": 18.0081,
+      "step": 17323
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5204915686805055e-10,
+      "loss": 18.9541,
+      "step": 17324
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2326847871623202e-10,
+      "loss": 15.4918,
+      "step": 17325
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9623207905272723e-10,
+      "loss": 18.557,
+      "step": 17326
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7093995883232795e-10,
+      "loss": 20.5589,
+      "step": 17327
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4739211893211035e-10,
+      "loss": 17.5825,
+      "step": 17328
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.255885601680884e-10,
+      "loss": 15.9578,
+      "step": 17329
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0552928330631596e-10,
+      "loss": 0.6019,
+      "step": 17330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.721428905178464e-11,
+      "loss": 15.8343,
+      "step": 17331
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.064357803177047e-11,
+      "loss": 16.6508,
+      "step": 17332
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.5817150834691636e-11,
+      "loss": 0.64,
+      "step": 17333
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.273500797125074e-11,
+      "loss": 17.2291,
+      "step": 17334
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.139714990219034e-11,
+      "loss": 18.2996,
+      "step": 17335
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1803577016088484e-11,
+      "loss": 0.5999,
+      "step": 17336
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3954289657114316e-11,
+      "loss": 16.3506,
+      "step": 17337
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.849288091721363e-12,
+      "loss": 17.0988,
+      "step": 17338
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.488572536403112e-12,
+      "loss": 0.6391,
+      "step": 17339
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.721431410396719e-13,
+      "loss": 19.1748,
+      "step": 17340
+    },
+    {
+      "epoch": 1.0,
+      "step": 17340,
+      "total_flos": 4.933059514333009e+20,
+      "train_loss": 105.25300236873699,
+      "train_runtime": 113142.7736,
+      "train_samples_per_second": 38.008,
+      "train_steps_per_second": 0.153
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 17341,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 2000,
+  "total_flos": 4.933059514333009e+20,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}